Add files using upload-large-folder tool

.gitattributes

@@ -140,3 +140,95 @@ Data/Tsukuyomi/wavs/VOICEACTRESS100_008.wav filter=lfs diff=lfs merge=lfs -text
 Data/Tsukuyomi/wavs/VOICEACTRESS100_012.wav filter=lfs diff=lfs merge=lfs -text
 Data/Tsukuyomi/wavs/VOICEACTRESS100_015.wav filter=lfs diff=lfs merge=lfs -text
 Data/Tsukuyomi/wavs/VOICEACTRESS100_010.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_013.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_016.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_017.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_018.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_021.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_023.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_025.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_024.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_020.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_022.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_019.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_027.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_030.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_041.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_043.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_042.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_026.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_028.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_045.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_044.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_046.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_047.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_029.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_049.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_050.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_048.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_052.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_053.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_055.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_054.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_057.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_058.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_056.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_060.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_031.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_051.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_033.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_059.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_034.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_032.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_037.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_036.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_040.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_061.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_038.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_035.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_062.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_064.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_063.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_067.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_068.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_069.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_065.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_039.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_070.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_066.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_083.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_082.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_081.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_084.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_087.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_085.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_089.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_086.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_088.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_090.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_071.wav filter=lfs diff=lfs merge=lfs -text
+.venv/Lib/site-packages/unidic_lite/dicdir/sys.dic filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_074.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_075.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_073.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_076.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_078.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_077.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_072.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_080.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_079.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_091.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_095.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_092.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_096.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_098.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_097.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_094.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_093.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_099.wav filter=lfs diff=lfs merge=lfs -text
+Data/Tsukuyomi/wavs/VOICEACTRESS100_100.wav filter=lfs diff=lfs merge=lfs -text
+.venv/Lib/site-packages/torch/lib/cublasLt64_12.dll filter=lfs diff=lfs merge=lfs -text
+.venv/Lib/site-packages/unidic/dicdir/sys.dic filter=lfs diff=lfs merge=lfs -text
+.venv/Lib/site-packages/torch/lib/cudnn_engines_precompiled64_9.dll filter=lfs diff=lfs merge=lfs -text
+.venv/Lib/site-packages/torch/lib/dnnl.lib filter=lfs diff=lfs merge=lfs -text
+.venv/Lib/site-packages/torch/lib/torch_cuda.dll filter=lfs diff=lfs merge=lfs -text

.venv/Lib/site-packages/scipy/sparse/tests/data/csc_py2.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bac27f1a3eb1fdd102dae39b7dd61ce83e82f096388e344e14285071984d01fa
+size 846

.venv/Lib/site-packages/scipy/sparse/tests/data/csc_py3.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b1b84315c7077417e720512d086a5a6217c2875b818d27704ae9b7237c69dfe
+size 851

.venv/Lib/site-packages/scipy/special/_ufuncs_cxx.pxd

@@ -0,0 +1,60 @@
+from . cimport sf_error
+cdef void _set_action(sf_error.sf_error_t, sf_error.sf_action_t) noexcept nogil
+cdef void *_export_ccospi
+cdef void *_export_lambertw_scalar
+cdef void *_export_csinpi
+cdef void *_export__stirling2_inexact
+cdef void *_export_ibeta_float
+cdef void *_export_ibeta_double
+cdef void *_export_ibetac_float
+cdef void *_export_ibetac_double
+cdef void *_export_ibetac_inv_float
+cdef void *_export_ibetac_inv_double
+cdef void *_export_ibeta_inv_float
+cdef void *_export_ibeta_inv_double
+cdef void *_export_binom
+cdef void *_export_faddeeva_dawsn
+cdef void *_export_faddeeva_dawsn_complex
+cdef void *_export_fellint_RC
+cdef void *_export_cellint_RC
+cdef void *_export_fellint_RD
+cdef void *_export_cellint_RD
+cdef void *_export_fellint_RF
+cdef void *_export_cellint_RF
+cdef void *_export_fellint_RG
+cdef void *_export_cellint_RG
+cdef void *_export_fellint_RJ
+cdef void *_export_cellint_RJ
+cdef void *_export_faddeeva_erf
+cdef void *_export_faddeeva_erfc_complex
+cdef void *_export_faddeeva_erfcx
+cdef void *_export_faddeeva_erfcx_complex
+cdef void *_export_faddeeva_erfi
+cdef void *_export_faddeeva_erfi_complex
+cdef void *_export_erfinv_float
+cdef void *_export_erfinv_double
+cdef void *_export_expit
+cdef void *_export_expitf
+cdef void *_export_expitl
+cdef void *_export_cgamma
+cdef void *_export_hyp1f1_double
+cdef void *_export_log_expit
+cdef void *_export_log_expitf
+cdef void *_export_log_expitl
+cdef void *_export_faddeeva_log_ndtr
+cdef void *_export_faddeeva_log_ndtr_complex
+cdef void *_export_loggamma_real
+cdef void *_export_loggamma
+cdef void *_export_logit
+cdef void *_export_logitf
+cdef void *_export_logitl
+cdef void *_export_faddeeva_ndtr
+cdef void *_export_powm1_float
+cdef void *_export_powm1_double
+cdef void *_export_cdigamma
+cdef void *_export_digamma
+cdef void *_export_crgamma
+cdef void *_export_faddeeva_voigt_profile
+cdef void *_export_faddeeva_w
+cdef void *_export_wrightomega
+cdef void *_export_wrightomega_real

.venv/Lib/site-packages/scipy/special/special/cephes/const.h

@@ -0,0 +1,77 @@
+/* Translated into C++ by SciPy developers in 2024.
+ * Original header with Copyright information appears below.
+ *
+ * Since we support only IEEE-754 floating point numbers, conditional logic
+ * supporting other arithmetic types has been removed.
+ */
+
+/*
+ *
+ *
+ *                                                   const.c
+ *
+ *     Globally declared constants
+ *
+ *
+ *
+ * SYNOPSIS:
+ *
+ * extern double nameofconstant;
+ *
+ *
+ *
+ *
+ * DESCRIPTION:
+ *
+ * This file contains a number of mathematical constants and
+ * also some needed size parameters of the computer arithmetic.
+ * The values are supplied as arrays of hexadecimal integers
+ * for IEEE arithmetic, and in a normal decimal scientific notation for
+ * other machines.  The particular notation used is determined
+ * by a symbol (IBMPC, or UNK) defined in the include file
+ * mconf.h.
+ *
+ * The default size parameters are as follows.
+ *
+ * For UNK mode:
+ * MACHEP =  1.38777878078144567553E-17       2**-56
+ * MAXLOG =  8.8029691931113054295988E1       log(2**127)
+ * MINLOG = -8.872283911167299960540E1        log(2**-128)
+ *
+ * For IEEE arithmetic (IBMPC):
+ * MACHEP =  1.11022302462515654042E-16       2**-53
+ * MAXLOG =  7.09782712893383996843E2         log(2**1024)
+ * MINLOG = -7.08396418532264106224E2         log(2**-1022)
+ *
+ * The global symbols for mathematical constants are
+ * SQ2OPI =  7.9788456080286535587989E-1      sqrt( 2/pi )
+ * LOGSQ2 =  3.46573590279972654709E-1        log(2)/2
+ * THPIO4 =  2.35619449019234492885           3*pi/4
+ *
+ * These lists are subject to change.
+ */
+/*                                                     const.c */
+
+/*
+ * Cephes Math Library Release 2.3:  March, 1995
+ * Copyright 1984, 1995 by Stephen L. Moshier
+ */
+#pragma once
+
+namespace special {
+namespace cephes {
+    namespace detail {
+        constexpr double MACHEP = 1.11022302462515654042E-16;  // 2**-53
+        constexpr double MAXLOG = 7.09782712893383996732E2;    // log(DBL_MAX)
+        constexpr double MINLOG = -7.451332191019412076235E2;  // log 2**-1022
+        constexpr double SQ2OPI = 7.9788456080286535587989E-1; // sqrt( 2/pi )
+        constexpr double LOGSQ2 = 3.46573590279972654709E-1;   // log(2)/2
+        constexpr double THPIO4 = 2.35619449019234492885;      // 3*pi/4
+        // Following two added by SciPy developers.
+        // Euler's constant
+        constexpr double SCIPY_EULER = 0.577215664901532860606512090082402431;
+        // e as long double
+        constexpr long double SCIPY_El = 2.718281828459045235360287471352662498L;
+    } // namespace detail
+} // namespace cephes
+} // namespace special

.venv/Lib/site-packages/scipy/special/special/cephes/gamma.h

@@ -0,0 +1,343 @@
+/* Translated into C++ by SciPy developers in 2024.
+ * Original header with Copyright information appears below.
+ */
+
+/*
+ *     Gamma function
+ *
+ *
+ *
+ * SYNOPSIS:
+ *
+ * double x, y, Gamma();
+ *
+ * y = Gamma( x );
+ *
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns Gamma function of the argument.  The result is
+ * correctly signed.
+ *
+ * Arguments |x| <= 34 are reduced by recurrence and the function
+ * approximated by a rational function of degree 6/7 in the
+ * interval (2,3).  Large arguments are handled by Stirling's
+ * formula. Large negative arguments are made positive using
+ * a reflection formula.
+ *
+ *
+ * ACCURACY:
+ *
+ *                      Relative error:
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE    -170,-33      20000       2.3e-15     3.3e-16
+ *    IEEE     -33,  33     20000       9.4e-16     2.2e-16
+ *    IEEE      33, 171.6   20000       2.3e-15     3.2e-16
+ *
+ * Error for arguments outside the test range will be larger
+ * owing to error amplification by the exponential function.
+ *
+ */
+
+/*                                                     lgam()
+ *
+ *     Natural logarithm of Gamma function
+ *
+ *
+ *
+ * SYNOPSIS:
+ *
+ * double x, y, lgam();
+ *
+ * y = lgam( x );
+ *
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Returns the base e (2.718...) logarithm of the absolute
+ * value of the Gamma function of the argument.
+ *
+ * For arguments greater than 13, the logarithm of the Gamma
+ * function is approximated by the logarithmic version of
+ * Stirling's formula using a polynomial approximation of
+ * degree 4. Arguments between -33 and +33 are reduced by
+ * recurrence to the interval [2,3] of a rational approximation.
+ * The cosecant reflection formula is employed for arguments
+ * less than -33.
+ *
+ * Arguments greater than MAXLGM return INFINITY and an error
+ * message.  MAXLGM = 2.556348e305 for IEEE arithmetic.
+ *
+ *
+ *
+ * ACCURACY:
+ *
+ *
+ * arithmetic      domain        # trials     peak         rms
+ *    IEEE    0, 3                 28000     5.4e-16     1.1e-16
+ *    IEEE    2.718, 2.556e305     40000     3.5e-16     8.3e-17
+ * The error criterion was relative when the function magnitude
+ * was greater than one but absolute when it was less than one.
+ *
+ * The following test used the relative error criterion, though
+ * at certain points the relative error could be much higher than
+ * indicated.
+ *    IEEE    -200, -4             10000     4.8e-16     1.3e-16
+ *
+ */
+
+/*
+ * Cephes Math Library Release 2.2:  July, 1992
+ * Copyright 1984, 1987, 1989, 1992 by Stephen L. Moshier
+ * Direct inquiries to 30 Frost Street, Cambridge, MA 02140
+ */
+#pragma once
+
+#include "../config.h"
+#include "../error.h"
+#include "polevl.h"
+
+namespace special {
+namespace cephes {
+    namespace detail {
+        constexpr double gamma_P[] = {1.60119522476751861407E-4, 1.19135147006586384913E-3, 1.04213797561761569935E-2,
+                                      4.76367800457137231464E-2, 2.07448227648435975150E-1, 4.94214826801497100753E-1,
+                                      9.99999999999999996796E-1};
+
+        constexpr double gamma_Q[] = {-2.31581873324120129819E-5, 5.39605580493303397842E-4, -4.45641913851797240494E-3,
+                                      1.18139785222060435552E-2,  3.58236398605498653373E-2, -2.34591795718243348568E-1,
+                                      7.14304917030273074085E-2,  1.00000000000000000320E0};
+
+        constexpr double MAXGAM = 171.624376956302725;
+        constexpr double LOGPI = 1.14472988584940017414;
+
+        /* Stirling's formula for the Gamma function */
+        constexpr double gamma_STIR[5] = {
+            7.87311395793093628397E-4, -2.29549961613378126380E-4, -2.68132617805781232825E-3,
+            3.47222221605458667310E-3, 8.33333333333482257126E-2,
+        };
+
+        constexpr double MAXSTIR = 143.01608;
+        constexpr double SQTPI = 2.50662827463100050242E0;
+
+        /* Gamma function computed by Stirling's formula.
+         * The polynomial STIR is valid for 33 <= x <= 172.
+         */
+        SPECFUN_HOST_DEVICE inline double stirf(double x) {
+            double y, w, v;
+
+            if (x >= MAXGAM) {
+                return (std::numeric_limits<double>::infinity());
+            }
+            w = 1.0 / x;
+            w = 1.0 + w * special::cephes::polevl(w, gamma_STIR, 4);
+            y = std::exp(x);
+            if (x > MAXSTIR) { /* Avoid overflow in pow() */
+                v = std::pow(x, 0.5 * x - 0.25);
+                y = v * (v / y);
+            } else {
+                y = std::pow(x, x - 0.5) / y;
+            }
+            y = SQTPI * y * w;
+            return (y);
+        }
+    } // namespace detail
+
+    SPECFUN_HOST_DEVICE inline double Gamma(double x) {
+        double p, q, z;
+        int i;
+        int sgngam = 1;
+
+        if (!std::isfinite(x)) {
+            return x;
+        }
+        q = std::abs(x);
+
+        if (q > 33.0) {
+            if (x < 0.0) {
+                p = floor(q);
+                if (p == q) {
+                gamnan:
+                    set_error("Gamma", SF_ERROR_OVERFLOW, NULL);
+                    return (std::numeric_limits<double>::infinity());
+                }
+                i = p;
+                if ((i & 1) == 0) {
+                    sgngam = -1;
+                }
+                z = q - p;
+                if (z > 0.5) {
+                    p += 1.0;
+                    z = q - p;
+                }
+                z = q * std::sin(M_PI * z);
+                if (z == 0.0) {
+                    return (sgngam * std::numeric_limits<double>::infinity());
+                }
+                z = std::abs(z);
+                z = M_PI / (z * detail::stirf(q));
+            } else {
+                z = detail::stirf(x);
+            }
+            return (sgngam * z);
+        }
+
+        z = 1.0;
+        while (x >= 3.0) {
+            x -= 1.0;
+            z *= x;
+        }
+
+        while (x < 0.0) {
+            if (x > -1.E-9) {
+                goto small;
+            }
+            z /= x;
+            x += 1.0;
+        }
+
+        while (x < 2.0) {
+            if (x < 1.e-9) {
+                goto small;
+            }
+            z /= x;
+            x += 1.0;
+        }
+
+        if (x == 2.0) {
+            return (z);
+        }
+
+        x -= 2.0;
+        p = polevl(x, detail::gamma_P, 6);
+        q = polevl(x, detail::gamma_Q, 7);
+        return (z * p / q);
+
+    small:
+        if (x == 0.0) {
+            goto gamnan;
+        } else
+            return (z / ((1.0 + 0.5772156649015329 * x) * x));
+    }
+
+    namespace detail {
+        /* A[]: Stirling's formula expansion of log Gamma
+         * B[], C[]: log Gamma function between 2 and 3
+         */
+        constexpr double gamma_A[] = {8.11614167470508450300E-4, -5.95061904284301438324E-4, 7.93650340457716943945E-4,
+                                      -2.77777777730099687205E-3, 8.33333333333331927722E-2};
+
+        constexpr double gamma_B[] = {-1.37825152569120859100E3, -3.88016315134637840924E4, -3.31612992738871184744E5,
+                                      -1.16237097492762307383E6, -1.72173700820839662146E6, -8.53555664245765465627E5};
+
+        constexpr double gamma_C[] = {
+            /* 1.00000000000000000000E0, */
+            -3.51815701436523470549E2, -1.70642106651881159223E4, -2.20528590553854454839E5,
+            -1.13933444367982507207E6, -2.53252307177582951285E6, -2.01889141433532773231E6};
+
+        /* log( sqrt( 2*pi ) ) */
+        constexpr double LS2PI = 0.91893853320467274178;
+
+        constexpr double MAXLGM = 2.556348e305;
+
+        SPECFUN_HOST_DEVICE double lgam_sgn(double x, int *sign) {
+            double p, q, u, w, z;
+            int i;
+
+            *sign = 1;
+
+            if (!std::isfinite(x)) {
+                return x;
+            }
+
+            if (x < -34.0) {
+                q = -x;
+                w = lgam_sgn(q, sign);
+                p = floor(q);
+                if (p == q) {
+                lgsing:
+                    set_error("lgam", SF_ERROR_SINGULAR, NULL);
+                    return (std::numeric_limits<double>::infinity());
+                }
+                i = p;
+                if ((i & 1) == 0) {
+                    *sign = -1;
+                } else {
+                    *sign = 1;
+                }
+                z = q - p;
+                if (z > 0.5) {
+                    p += 1.0;
+                    z = p - q;
+                }
+                z = q * std::sin(M_PI * z);
+                if (z == 0.0) {
+                    goto lgsing;
+                }
+                /*     z = log(M_PI) - log( z ) - w; */
+                z = LOGPI - std::log(z) - w;
+                return (z);
+            }
+
+            if (x < 13.0) {
+                z = 1.0;
+                p = 0.0;
+                u = x;
+                while (u >= 3.0) {
+                    p -= 1.0;
+                    u = x + p;
+                    z *= u;
+                }
+                while (u < 2.0) {
+                    if (u == 0.0) {
+                        goto lgsing;
+                    }
+                    z /= u;
+                    p += 1.0;
+                    u = x + p;
+                }
+                if (z < 0.0) {
+                    *sign = -1;
+                    z = -z;
+                } else {
+                    *sign = 1;
+                }
+                if (u == 2.0) {
+                    return (std::log(z));
+                }
+                p -= 2.0;
+                x = x + p;
+                p = x * polevl(x, gamma_B, 5) / p1evl(x, gamma_C, 6);
+                return (std::log(z) + p);
+            }
+
+            if (x > MAXLGM) {
+                return (*sign * std::numeric_limits<double>::infinity());
+            }
+
+            q = (x - 0.5) * std::log(x) - x + LS2PI;
+            if (x > 1.0e8) {
+                return (q);
+            }
+
+            p = 1.0 / (x * x);
+            if (x >= 1000.0) {
+                q += ((7.9365079365079365079365e-4 * p - 2.7777777777777777777778e-3) * p + 0.0833333333333333333333) /
+                     x;
+            } else {
+                q += polevl(p, gamma_A, 4) / x;
+            }
+            return (q);
+        }
+    } // namespace detail
+
+    /* Logarithm of Gamma function */
+    SPECFUN_HOST_DEVICE double lgam(double x) {
+        int sign;
+        return detail::lgam_sgn(x, &sign);
+    }
+
+} // namespace cephes
+} // namespace special

.venv/Lib/site-packages/scipy/special/special/cephes/polevl.h

@@ -0,0 +1,165 @@
+/* Translated into C++ by SciPy developers in 2024.
+ * Original header with Copyright information appears below.
+ */
+
+/*                                                     polevl.c
+ *                                                     p1evl.c
+ *
+ *     Evaluate polynomial
+ *
+ *
+ *
+ * SYNOPSIS:
+ *
+ * int N;
+ * double x, y, coef[N+1], polevl[];
+ *
+ * y = polevl( x, coef, N );
+ *
+ *
+ *
+ * DESCRIPTION:
+ *
+ * Evaluates polynomial of degree N:
+ *
+ *                     2          N
+ * y  =  C  + C x + C x  +...+ C x
+ *        0    1     2          N
+ *
+ * Coefficients are stored in reverse order:
+ *
+ * coef[0] = C  , ..., coef[N] = C  .
+ *            N                   0
+ *
+ * The function p1evl() assumes that c_N = 1.0 so that coefficent
+ * is omitted from the array.  Its calling arguments are
+ * otherwise the same as polevl().
+ *
+ *
+ * SPEED:
+ *
+ * In the interest of speed, there are no checks for out
+ * of bounds arithmetic.  This routine is used by most of
+ * the functions in the library.  Depending on available
+ * equipment features, the user may wish to rewrite the
+ * program in microcode or assembly language.
+ *
+ */
+
+/*
+ * Cephes Math Library Release 2.1:  December, 1988
+ * Copyright 1984, 1987, 1988 by Stephen L. Moshier
+ * Direct inquiries to 30 Frost Street, Cambridge, MA 02140
+ */
+
+/* Sources:
+ * [1] Holin et. al., "Polynomial and Rational Function Evaluation",
+ *     https://www.boost.org/doc/libs/1_61_0/libs/math/doc/html/math_toolkit/roots/rational.html
+ */
+
+/* Scipy changes:
+ * - 06-23-2016: add code for evaluating rational functions
+ */
+#pragma once
+
+#include "../config.h"
+
+namespace special {
+namespace cephes {
+    SPECFUN_HOST_DEVICE inline double polevl(double x, const double coef[], int N) {
+        double ans;
+        int i;
+        const double *p;
+
+        p = coef;
+        ans = *p++;
+        i = N;
+
+        do {
+            ans = ans * x + *p++;
+        } while (--i);
+
+        return (ans);
+    }
+
+    /*                                                     p1evl() */
+    /*                                          N
+     * Evaluate polynomial when coefficient of x  is 1.0.
+     * That is, C_{N} is assumed to be 1, and that coefficient
+     * is not included in the input array coef.
+     * coef must have length N and contain the polynomial coefficients
+     * stored as
+     *     coef[0] = C_{N-1}
+     *     coef[1] = C_{N-2}
+     *          ...
+     *     coef[N-2] = C_1
+     *     coef[N-1] = C_0
+     * Otherwise same as polevl.
+     */
+
+    SPECFUN_HOST_DEVICE inline double p1evl(double x, const double coef[], int N) {
+        double ans;
+        const double *p;
+        int i;
+
+        p = coef;
+        ans = x + *p++;
+        i = N - 1;
+
+        do
+            ans = ans * x + *p++;
+        while (--i);
+
+        return (ans);
+    }
+
+    /* Evaluate a rational function. See [1]. */
+
+    SPECFUN_HOST_DEVICE inline double ratevl(double x, const double num[], int M, const double denom[], int N) {
+        int i, dir;
+        double y, num_ans, denom_ans;
+        double absx = std::abs(x);
+        const double *p;
+
+        if (absx > 1) {
+            /* Evaluate as a polynomial in 1/x. */
+            dir = -1;
+            p = num + M;
+            y = 1 / x;
+        } else {
+            dir = 1;
+            p = num;
+            y = x;
+        }
+
+        /* Evaluate the numerator */
+        num_ans = *p;
+        p += dir;
+        for (i = 1; i <= M; i++) {
+            num_ans = num_ans * y + *p;
+            p += dir;
+        }
+
+        /* Evaluate the denominator */
+        if (absx > 1) {
+            p = denom + N;
+        } else {
+            p = denom;
+        }
+
+        denom_ans = *p;
+        p += dir;
+        for (i = 1; i <= N; i++) {
+            denom_ans = denom_ans * y + *p;
+            p += dir;
+        }
+
+        if (absx > 1) {
+            i = N - M;
+            return std::pow(x, i) * num_ans / denom_ans;
+        } else {
+            return num_ans / denom_ans;
+        }
+    }
+} // namespace cephes
+} // namespace special

.venv/Lib/site-packages/scipy/special/special/cephes/psi.h

@@ -0,0 +1,194 @@
+/* Translated into C++ by SciPy developers in 2024.
+ * Original header with Copyright information appears below.
+ */
+
+/*                                                     psi.c
+ *
+ *     Psi (digamma) function
+ *
+ *
+ * SYNOPSIS:
+ *
+ * double x, y, psi();
+ *
+ * y = psi( x );
+ *
+ *
+ * DESCRIPTION:
+ *
+ *              d      -
+ *   psi(x)  =  -- ln | (x)
+ *              dx
+ *
+ * is the logarithmic derivative of the gamma function.
+ * For integer x,
+ *                   n-1
+ *                    -
+ * psi(n) = -EUL  +   >  1/k.
+ *                    -
+ *                   k=1
+ *
+ * This formula is used for 0 < n <= 10.  If x is negative, it
+ * is transformed to a positive argument by the reflection
+ * formula  psi(1-x) = psi(x) + pi cot(pi x).
+ * For general positive x, the argument is made greater than 10
+ * using the recurrence  psi(x+1) = psi(x) + 1/x.
+ * Then the following asymptotic expansion is applied:
+ *
+ *                           inf.   B
+ *                            -      2k
+ * psi(x) = log(x) - 1/2x -   >   -------
+ *                            -        2k
+ *                           k=1   2k x
+ *
+ * where the B2k are Bernoulli numbers.
+ *
+ * ACCURACY:
+ *    Relative error (except absolute when |psi| < 1):
+ * arithmetic   domain     # trials      peak         rms
+ *    IEEE      0,30        30000       1.3e-15     1.4e-16
+ *    IEEE      -30,0       40000       1.5e-15     2.2e-16
+ *
+ * ERROR MESSAGES:
+ *     message         condition      value returned
+ * psi singularity    x integer <=0      INFINITY
+ */
+
+/*
+ * Cephes Math Library Release 2.8:  June, 2000
+ * Copyright 1984, 1987, 1992, 2000 by Stephen L. Moshier
+ */
+
+/*
+ * Code for the rational approximation on [1, 2] is:
+ *
+ * (C) Copyright John Maddock 2006.
+ * Use, modification and distribution are subject to the
+ * Boost Software License, Version 1.0. (See accompanying file
+ * LICENSE_1_0.txt or copy at https://www.boost.org/LICENSE_1_0.txt)
+ */
+#pragma once
+
+#include "../config.h"
+#include "../error.h"
+#include "const.h"
+#include "polevl.h"
+
+namespace special {
+namespace cephes {
+    namespace detail {
+        constexpr double psi_A[] = {8.33333333333333333333E-2,  -2.10927960927960927961E-2, 7.57575757575757575758E-3,
+                                    -4.16666666666666666667E-3, 3.96825396825396825397E-3,  -8.33333333333333333333E-3,
+                                    8.33333333333333333333E-2};
+
+        constexpr float psi_Y = 0.99558162689208984f;
+
+        constexpr double psi_root1 = 1569415565.0 / 1073741824.0;
+        constexpr double psi_root2 = (381566830.0 / 1073741824.0) / 1073741824.0;
+        constexpr double psi_root3 = 0.9016312093258695918615325266959189453125e-19;
+
+        constexpr double psi_P[] = {-0.0020713321167745952, -0.045251321448739056, -0.28919126444774784,
+                                    -0.65031853770896507,   -0.32555031186804491,  0.25479851061131551};
+        constexpr double psi_Q[] = {-0.55789841321675513e-6,
+                                    0.0021284987017821144,
+                                    0.054151797245674225,
+                                    0.43593529692665969,
+                                    1.4606242909763515,
+                                    2.0767117023730469,
+                                    1.0};
+
+        SPECFUN_HOST_DEVICE double digamma_imp_1_2(double x) {
+            /*
+             * Rational approximation on [1, 2] taken from Boost.
+             *
+             * Now for the approximation, we use the form:
+             *
+             * digamma(x) = (x - root) * (Y + R(x-1))
+             *
+             * Where root is the location of the positive root of digamma,
+             * Y is a constant, and R is optimised for low absolute error
+             * compared to Y.
+             *
+             * Maximum Deviation Found:               1.466e-18
+             * At double precision, max error found:  2.452e-17
+             */
+            double r, g;
+
+            g = x - psi_root1;
+            g -= psi_root2;
+            g -= psi_root3;
+            r = special::cephes::polevl(x - 1.0, psi_P, 5) / special::cephes::polevl(x - 1.0, psi_Q, 6);
+
+            return g * psi_Y + g * r;
+        }
+
+        SPECFUN_HOST_DEVICE double psi_asy(double x) {
+            double y, z;
+
+            if (x < 1.0e17) {
+                z = 1.0 / (x * x);
+                y = z * special::cephes::polevl(z, psi_A, 6);
+            } else {
+                y = 0.0;
+            }
+
+            return std::log(x) - (0.5 / x) - y;
+        }
+    } // namespace detail
+
+    SPECFUN_HOST_DEVICE double psi(double x) {
+        double y = 0.0;
+        double q, r;
+        int i, n;
+
+        if (std::isnan(x)) {
+            return x;
+        } else if (x == std::numeric_limits<double>::infinity()) {
+            return x;
+        } else if (x == -std::numeric_limits<double>::infinity()) {
+            return std::numeric_limits<double>::quiet_NaN();
+        } else if (x == 0) {
+            set_error("psi", SF_ERROR_SINGULAR, NULL);
+            return std::copysign(std::numeric_limits<double>::infinity(), -x);
+        } else if (x < 0.0) {
+            /* argument reduction before evaluating tan(pi * x) */
+            r = std::modf(x, &q);
+            if (r == 0.0) {
+                set_error("psi", SF_ERROR_SINGULAR, NULL);
+                return std::numeric_limits<double>::quiet_NaN();
+            }
+            y = -M_PI / std::tan(M_PI * r);
+            x = 1.0 - x;
+        }
+
+        /* check for positive integer up to 10 */
+        if ((x <= 10.0) && (x == std::floor(x))) {
+            n = static_cast<int>(x);
+            for (i = 1; i < n; i++) {
+                y += 1.0 / i;
+            }
+            y -= detail::SCIPY_EULER;
+            return y;
+        }
+
+        /* use the recurrence relation to move x into [1, 2] */
+        if (x < 1.0) {
+            y -= 1.0 / x;
+            x += 1.0;
+        } else if (x < 10.0) {
+            while (x > 2.0) {
+                x -= 1.0;
+                y += 1.0 / x;
+            }
+        }
+        if ((1.0 <= x) && (x <= 2.0)) {
+            y += detail::digamma_imp_1_2(x);
+            return y;
+        }
+
+        /* x is large, use the asymptotic series */
+        y += detail::psi_asy(x);
+        return y;
+    }
+} // namespace cephes
+} // namespace special

.venv/Lib/site-packages/scipy/special/tests/test_dd.py

@@ -0,0 +1,46 @@
+# Tests for a few of the "double-double" C functions defined in cephes/dd_*.
+
+import pytest
+from numpy.testing import assert_allclose
+from scipy.special._test_internal import _dd_exp, _dd_log, _dd_expm1
+
+
+# Each tuple in test_data contains:
+#   (dd_func, xhi, xlo, expected_yhi, expected_ylo)
+# The expected values were computed with mpmath, e.g.
+#
+#   import mpmath
+#   mpmath.mp.dps = 100
+#   xhi = 10.0
+#   xlo = 0.0
+#   x = mpmath.mpf(xhi) + mpmath.mpf(xlo)
+#   y = mpmath.log(x)
+#   expected_yhi = float(y)
+#   expected_ylo = float(y - expected_yhi)
+#
+test_data = [
+    (_dd_exp, -0.3333333333333333, -1.850371707708594e-17,
+     0.7165313105737893, -2.0286948382455594e-17),
+    (_dd_exp, 0.0, 0.0, 1.0, 0.0),
+    (_dd_exp, 10.0, 0.0, 22026.465794806718, -1.3780134700517372e-12),
+    (_dd_log, 0.03125, 0.0, -3.4657359027997265, -4.930038229799327e-18),
+    (_dd_log, 10.0, 0.0, 2.302585092994046, -2.1707562233822494e-16),
+    (_dd_expm1, -1.25, 0.0, -0.7134952031398099, -4.7031321153650186e-17),
+    (_dd_expm1, -0.484375, 0.0, -0.3839178722093218, 7.609376052156984e-18),
+    (_dd_expm1, -0.25, 0.0, -0.22119921692859512, -1.0231869534531498e-17),
+    (_dd_expm1, -0.0625, 0.0, -0.06058693718652421, -7.077887227488846e-19),
+    (_dd_expm1, 0.0, 0.0, 0.0, 0.0),
+    (_dd_expm1, 0.0625, 3.5e-18, 0.06449445891785943, 1.4323095758164254e-18),
+    (_dd_expm1, 0.25, 0.0, 0.2840254166877415, -2.133257464457841e-17),
+    (_dd_expm1, 0.498046875, 0.0, 0.645504254608231, -9.198435524984236e-18),
+    (_dd_expm1, 1.25, 0.0, 2.4903429574618414, -4.604261945372796e-17)
+]
+
+
+@pytest.mark.parametrize('dd_func, xhi, xlo, expected_yhi, expected_ylo',
+                         test_data)
+def test_dd(dd_func, xhi, xlo, expected_yhi, expected_ylo):
+    yhi, ylo = dd_func(xhi, xlo)
+    assert yhi == expected_yhi, (f"high double ({yhi}) does not equal the "
+                                 f"expected value {expected_yhi}")
+    assert_allclose(ylo, expected_ylo, rtol=5e-15)

.venv/Lib/site-packages/scipy/special/tests/test_digamma.py

@@ -0,0 +1,45 @@
+import numpy as np
+from numpy import pi, log, sqrt
+from numpy.testing import assert_, assert_equal
+
+from scipy.special._testutils import FuncData
+import scipy.special as sc
+
+# Euler-Mascheroni constant
+euler = 0.57721566490153286
+
+
+def test_consistency():
+    # Make sure the implementation of digamma for real arguments
+    # agrees with the implementation of digamma for complex arguments.
+
+    # It's all poles after -1e16
+    x = np.r_[-np.logspace(15, -30, 200), np.logspace(-30, 300, 200)]
+    dataset = np.vstack((x + 0j, sc.digamma(x))).T
+    FuncData(sc.digamma, dataset, 0, 1, rtol=5e-14, nan_ok=True).check()
+
+
+def test_special_values():
+    # Test special values from Gauss's digamma theorem. See
+    #
+    # https://en.wikipedia.org/wiki/Digamma_function
+
+    dataset = [
+        (1, -euler),
+        (0.5, -2*log(2) - euler),
+        (1/3, -pi/(2*sqrt(3)) - 3*log(3)/2 - euler),
+        (1/4, -pi/2 - 3*log(2) - euler),
+        (1/6, -pi*sqrt(3)/2 - 2*log(2) - 3*log(3)/2 - euler),
+        (1/8,
+         -pi/2 - 4*log(2) - (pi + log(2 + sqrt(2)) - log(2 - sqrt(2)))/sqrt(2) - euler)
+    ]
+
+    dataset = np.asarray(dataset)
+    FuncData(sc.digamma, dataset, 0, 1, rtol=1e-14).check()
+
+
+def test_nonfinite():
+    pts = [0.0, -0.0, np.inf]
+    std = [-np.inf, np.inf, np.inf]
+    assert_equal(sc.digamma(pts), std)
+    assert_(all(np.isnan(sc.digamma([-np.inf, -1]))))

.venv/Lib/site-packages/scipy/special/tests/test_ellip_harm.py

@@ -0,0 +1,278 @@
+#
+# Tests for the Ellipsoidal Harmonic Function,
+# Distributed under the same license as SciPy itself.
+#
+
+import numpy as np
+from numpy.testing import (assert_equal, assert_almost_equal, assert_allclose,
+                           assert_, suppress_warnings)
+from scipy.special._testutils import assert_func_equal
+from scipy.special import ellip_harm, ellip_harm_2, ellip_normal
+from scipy.integrate import IntegrationWarning
+from numpy import sqrt, pi
+
+
+def test_ellip_potential():
+    def change_coefficient(lambda1, mu, nu, h2, k2):
+        x = sqrt(lambda1**2*mu**2*nu**2/(h2*k2))
+        y = sqrt((lambda1**2 - h2)*(mu**2 - h2)*(h2 - nu**2)/(h2*(k2 - h2)))
+        z = sqrt((lambda1**2 - k2)*(k2 - mu**2)*(k2 - nu**2)/(k2*(k2 - h2)))
+        return x, y, z
+
+    def solid_int_ellip(lambda1, mu, nu, n, p, h2, k2):
+        return (ellip_harm(h2, k2, n, p, lambda1)*ellip_harm(h2, k2, n, p, mu)
+               * ellip_harm(h2, k2, n, p, nu))
+
+    def solid_int_ellip2(lambda1, mu, nu, n, p, h2, k2):
+        return (ellip_harm_2(h2, k2, n, p, lambda1)
+                * ellip_harm(h2, k2, n, p, mu)*ellip_harm(h2, k2, n, p, nu))
+
+    def summation(lambda1, mu1, nu1, lambda2, mu2, nu2, h2, k2):
+        tol = 1e-8
+        sum1 = 0
+        for n in range(20):
+            xsum = 0
+            for p in range(1, 2*n+2):
+                xsum += (4*pi*(solid_int_ellip(lambda2, mu2, nu2, n, p, h2, k2)
+                    * solid_int_ellip2(lambda1, mu1, nu1, n, p, h2, k2)) /
+                    (ellip_normal(h2, k2, n, p)*(2*n + 1)))
+            if abs(xsum) < 0.1*tol*abs(sum1):
+                break
+            sum1 += xsum
+        return sum1, xsum
+
+    def potential(lambda1, mu1, nu1, lambda2, mu2, nu2, h2, k2):
+        x1, y1, z1 = change_coefficient(lambda1, mu1, nu1, h2, k2)
+        x2, y2, z2 = change_coefficient(lambda2, mu2, nu2, h2, k2)
+        res = sqrt((x2 - x1)**2 + (y2 - y1)**2 + (z2 - z1)**2)
+        return 1/res
+
+    pts = [
+        (120, sqrt(19), 2, 41, sqrt(17), 2, 15, 25),
+        (120, sqrt(16), 3.2, 21, sqrt(11), 2.9, 11, 20),
+       ]
+
+    with suppress_warnings() as sup:
+        sup.filter(IntegrationWarning, "The occurrence of roundoff error")
+        sup.filter(IntegrationWarning, "The maximum number of subdivisions")
+
+        for p in pts:
+            err_msg = repr(p)
+            exact = potential(*p)
+            result, last_term = summation(*p)
+            assert_allclose(exact, result, atol=0, rtol=1e-8, err_msg=err_msg)
+            assert_(abs(result - exact) < 10*abs(last_term), err_msg)
+
+
+def test_ellip_norm():
+
+    def G01(h2, k2):
+        return 4*pi
+
+    def G11(h2, k2):
+        return 4*pi*h2*k2/3
+
+    def G12(h2, k2):
+        return 4*pi*h2*(k2 - h2)/3
+
+    def G13(h2, k2):
+        return 4*pi*k2*(k2 - h2)/3
+
+    def G22(h2, k2):
+        res = (2*(h2**4 + k2**4) - 4*h2*k2*(h2**2 + k2**2) + 6*h2**2*k2**2 +
+        sqrt(h2**2 + k2**2 - h2*k2)*(-2*(h2**3 + k2**3) + 3*h2*k2*(h2 + k2)))
+        return 16*pi/405*res
+
+    def G21(h2, k2):
+        res = (2*(h2**4 + k2**4) - 4*h2*k2*(h2**2 + k2**2) + 6*h2**2*k2**2
+        + sqrt(h2**2 + k2**2 - h2*k2)*(2*(h2**3 + k2**3) - 3*h2*k2*(h2 + k2)))
+        return 16*pi/405*res
+
+    def G23(h2, k2):
+        return 4*pi*h2**2*k2*(k2 - h2)/15
+
+    def G24(h2, k2):
+        return 4*pi*h2*k2**2*(k2 - h2)/15
+
+    def G25(h2, k2):
+        return 4*pi*h2*k2*(k2 - h2)**2/15
+
+    def G32(h2, k2):
+        res = (16*(h2**4 + k2**4) - 36*h2*k2*(h2**2 + k2**2) + 46*h2**2*k2**2
+        + sqrt(4*(h2**2 + k2**2) - 7*h2*k2)*(-8*(h2**3 + k2**3) +
+        11*h2*k2*(h2 + k2)))
+        return 16*pi/13125*k2*h2*res
+
+    def G31(h2, k2):
+        res = (16*(h2**4 + k2**4) - 36*h2*k2*(h2**2 + k2**2) + 46*h2**2*k2**2
+        + sqrt(4*(h2**2 + k2**2) - 7*h2*k2)*(8*(h2**3 + k2**3) -
+        11*h2*k2*(h2 + k2)))
+        return 16*pi/13125*h2*k2*res
+
+    def G34(h2, k2):
+        res = (6*h2**4 + 16*k2**4 - 12*h2**3*k2 - 28*h2*k2**3 + 34*h2**2*k2**2
+        + sqrt(h2**2 + 4*k2**2 - h2*k2)*(-6*h2**3 - 8*k2**3 + 9*h2**2*k2 +
+                                            13*h2*k2**2))
+        return 16*pi/13125*h2*(k2 - h2)*res
+
+    def G33(h2, k2):
+        res = (6*h2**4 + 16*k2**4 - 12*h2**3*k2 - 28*h2*k2**3 + 34*h2**2*k2**2
+        + sqrt(h2**2 + 4*k2**2 - h2*k2)*(6*h2**3 + 8*k2**3 - 9*h2**2*k2 -
+        13*h2*k2**2))
+        return 16*pi/13125*h2*(k2 - h2)*res
+
+    def G36(h2, k2):
+        res = (16*h2**4 + 6*k2**4 - 28*h2**3*k2 - 12*h2*k2**3 + 34*h2**2*k2**2
+        + sqrt(4*h2**2 + k2**2 - h2*k2)*(-8*h2**3 - 6*k2**3 + 13*h2**2*k2 +
+        9*h2*k2**2))
+        return 16*pi/13125*k2*(k2 - h2)*res
+
+    def G35(h2, k2):
+        res = (16*h2**4 + 6*k2**4 - 28*h2**3*k2 - 12*h2*k2**3 + 34*h2**2*k2**2
+        + sqrt(4*h2**2 + k2**2 - h2*k2)*(8*h2**3 + 6*k2**3 - 13*h2**2*k2 -
+        9*h2*k2**2))
+        return 16*pi/13125*k2*(k2 - h2)*res
+
+    def G37(h2, k2):
+        return 4*pi*h2**2*k2**2*(k2 - h2)**2/105
+
+    known_funcs = {(0, 1): G01, (1, 1): G11, (1, 2): G12, (1, 3): G13,
+                   (2, 1): G21, (2, 2): G22, (2, 3): G23, (2, 4): G24,
+                   (2, 5): G25, (3, 1): G31, (3, 2): G32, (3, 3): G33,
+                   (3, 4): G34, (3, 5): G35, (3, 6): G36, (3, 7): G37}
+
+    def _ellip_norm(n, p, h2, k2):
+        func = known_funcs[n, p]
+        return func(h2, k2)
+    _ellip_norm = np.vectorize(_ellip_norm)
+
+    def ellip_normal_known(h2, k2, n, p):
+        return _ellip_norm(n, p, h2, k2)
+
+    # generate both large and small h2 < k2 pairs
+    np.random.seed(1234)
+    h2 = np.random.pareto(0.5, size=1)
+    k2 = h2 * (1 + np.random.pareto(0.5, size=h2.size))
+
+    points = []
+    for n in range(4):
+        for p in range(1, 2*n+2):
+            points.append((h2, k2, np.full(h2.size, n), np.full(h2.size, p)))
+    points = np.array(points)
+    with suppress_warnings() as sup:
+        sup.filter(IntegrationWarning, "The occurrence of roundoff error")
+        assert_func_equal(ellip_normal, ellip_normal_known, points, rtol=1e-12)
+
+
+def test_ellip_harm_2():
+
+    def I1(h2, k2, s):
+        res = (ellip_harm_2(h2, k2, 1, 1, s)/(3 * ellip_harm(h2, k2, 1, 1, s))
+        + ellip_harm_2(h2, k2, 1, 2, s)/(3 * ellip_harm(h2, k2, 1, 2, s)) +
+        ellip_harm_2(h2, k2, 1, 3, s)/(3 * ellip_harm(h2, k2, 1, 3, s)))
+        return res
+
+    with suppress_warnings() as sup:
+        sup.filter(IntegrationWarning, "The occurrence of roundoff error")
+        assert_almost_equal(I1(5, 8, 10), 1/(10*sqrt((100-5)*(100-8))))
+
+        # Values produced by code from arXiv:1204.0267
+        assert_almost_equal(ellip_harm_2(5, 8, 2, 1, 10), 0.00108056853382)
+        assert_almost_equal(ellip_harm_2(5, 8, 2, 2, 10), 0.00105820513809)
+        assert_almost_equal(ellip_harm_2(5, 8, 2, 3, 10), 0.00106058384743)
+        assert_almost_equal(ellip_harm_2(5, 8, 2, 4, 10), 0.00106774492306)
+        assert_almost_equal(ellip_harm_2(5, 8, 2, 5, 10), 0.00107976356454)
+
+
+def test_ellip_harm():
+
+    def E01(h2, k2, s):
+        return 1
+
+    def E11(h2, k2, s):
+        return s
+
+    def E12(h2, k2, s):
+        return sqrt(abs(s*s - h2))
+
+    def E13(h2, k2, s):
+        return sqrt(abs(s*s - k2))
+
+    def E21(h2, k2, s):
+        return s*s - 1/3*((h2 + k2) + sqrt(abs((h2 + k2)*(h2 + k2)-3*h2*k2)))
+
+    def E22(h2, k2, s):
+        return s*s - 1/3*((h2 + k2) - sqrt(abs((h2 + k2)*(h2 + k2)-3*h2*k2)))
+
+    def E23(h2, k2, s):
+        return s * sqrt(abs(s*s - h2))
+
+    def E24(h2, k2, s):
+        return s * sqrt(abs(s*s - k2))
+
+    def E25(h2, k2, s):
+        return sqrt(abs((s*s - h2)*(s*s - k2)))
+
+    def E31(h2, k2, s):
+        return s*s*s - (s/5)*(2*(h2 + k2) + sqrt(4*(h2 + k2)*(h2 + k2) -
+        15*h2*k2))
+
+    def E32(h2, k2, s):
+        return s*s*s - (s/5)*(2*(h2 + k2) - sqrt(4*(h2 + k2)*(h2 + k2) -
+        15*h2*k2))
+
+    def E33(h2, k2, s):
+        return sqrt(abs(s*s - h2))*(s*s - 1/5*((h2 + 2*k2) + sqrt(abs((h2 +
+        2*k2)*(h2 + 2*k2) - 5*h2*k2))))
+
+    def E34(h2, k2, s):
+        return sqrt(abs(s*s - h2))*(s*s - 1/5*((h2 + 2*k2) - sqrt(abs((h2 +
+        2*k2)*(h2 + 2*k2) - 5*h2*k2))))
+
+    def E35(h2, k2, s):
+        return sqrt(abs(s*s - k2))*(s*s - 1/5*((2*h2 + k2) + sqrt(abs((2*h2
+        + k2)*(2*h2 + k2) - 5*h2*k2))))
+
+    def E36(h2, k2, s):
+        return sqrt(abs(s*s - k2))*(s*s - 1/5*((2*h2 + k2) - sqrt(abs((2*h2
+        + k2)*(2*h2 + k2) - 5*h2*k2))))
+
+    def E37(h2, k2, s):
+        return s * sqrt(abs((s*s - h2)*(s*s - k2)))
+
+    assert_equal(ellip_harm(5, 8, 1, 2, 2.5, 1, 1),
+    ellip_harm(5, 8, 1, 2, 2.5))
+
+    known_funcs = {(0, 1): E01, (1, 1): E11, (1, 2): E12, (1, 3): E13,
+                   (2, 1): E21, (2, 2): E22, (2, 3): E23, (2, 4): E24,
+                   (2, 5): E25, (3, 1): E31, (3, 2): E32, (3, 3): E33,
+                   (3, 4): E34, (3, 5): E35, (3, 6): E36, (3, 7): E37}
+
+    point_ref = []
+
+    def ellip_harm_known(h2, k2, n, p, s):
+        for i in range(h2.size):
+            func = known_funcs[(int(n[i]), int(p[i]))]
+            point_ref.append(func(h2[i], k2[i], s[i]))
+        return point_ref
+
+    np.random.seed(1234)
+    h2 = np.random.pareto(0.5, size=30)
+    k2 = h2*(1 + np.random.pareto(0.5, size=h2.size))
+    s = np.random.pareto(0.5, size=h2.size)
+    points = []
+    for i in range(h2.size):
+        for n in range(4):
+            for p in range(1, 2*n+2):
+                points.append((h2[i], k2[i], n, p, s[i]))
+    points = np.array(points)
+    assert_func_equal(ellip_harm, ellip_harm_known, points, rtol=1e-12)
+
+
+def test_ellip_harm_invalid_p():
+    # Regression test. This should return nan.
+    n = 4
+    # Make p > 2*n + 1.
+    p = 2*n + 2
+    result = ellip_harm(0.5, 2.0, n, p, 0.2)
+    assert np.isnan(result)

.venv/Lib/site-packages/scipy/special/tests/test_erfinv.py

@@ -0,0 +1,89 @@
+import numpy as np
+from numpy.testing import assert_allclose, assert_equal
+import pytest
+
+import scipy.special as sc
+
+
+class TestInverseErrorFunction:
+    def test_compliment(self):
+        # Test erfcinv(1 - x) == erfinv(x)
+        x = np.linspace(-1, 1, 101)
+        assert_allclose(sc.erfcinv(1 - x), sc.erfinv(x), rtol=0, atol=1e-15)
+
+    def test_literal_values(self):
+        # The expected values were calculated with mpmath:
+        #
+        #   import mpmath
+        #   mpmath.mp.dps = 200
+        #   for y in [0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]:
+        #       x = mpmath.erfinv(y)
+        #       print(x)
+        #
+        y = np.array([0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9])
+        actual = sc.erfinv(y)
+        expected = [
+            0.0,
+            0.08885599049425769,
+            0.1791434546212917,
+            0.2724627147267543,
+            0.37080715859355795,
+            0.4769362762044699,
+            0.5951160814499948,
+            0.7328690779592167,
+            0.9061938024368233,
+            1.1630871536766743,
+        ]
+        assert_allclose(actual, expected, rtol=0, atol=1e-15)
+
+    @pytest.mark.parametrize(
+        'f, x, y',
+        [
+            (sc.erfinv, -1, -np.inf),
+            (sc.erfinv, 0, 0),
+            (sc.erfinv, 1, np.inf),
+            (sc.erfinv, -100, np.nan),
+            (sc.erfinv, 100, np.nan),
+            (sc.erfcinv, 0, np.inf),
+            (sc.erfcinv, 1, -0.0),
+            (sc.erfcinv, 2, -np.inf),
+            (sc.erfcinv, -100, np.nan),
+            (sc.erfcinv, 100, np.nan),
+        ],
+        ids=[
+            'erfinv at lower bound',
+            'erfinv at midpoint',
+            'erfinv at upper bound',
+            'erfinv below lower bound',
+            'erfinv above upper bound',
+            'erfcinv at lower bound',
+            'erfcinv at midpoint',
+            'erfcinv at upper bound',
+            'erfcinv below lower bound',
+            'erfcinv above upper bound',
+        ]
+    )
+    def test_domain_bounds(self, f, x, y):
+        assert_equal(f(x), y)
+
+    def test_erfinv_asympt(self):
+        # regression test for gh-12758: erfinv(x) loses precision at small x
+        # expected values precomputed with mpmath:
+        # >>> mpmath.mp.dps = 100
+        # >>> expected = [float(mpmath.erfinv(t)) for t in x]
+        x = np.array([1e-20, 1e-15, 1e-14, 1e-10, 1e-8, 0.9e-7, 1.1e-7, 1e-6])
+        expected = np.array([8.86226925452758e-21,
+                             8.862269254527581e-16,
+                             8.86226925452758e-15,
+                             8.862269254527581e-11,
+                             8.86226925452758e-09,
+                             7.97604232907484e-08,
+                             9.74849617998037e-08,
+                             8.8622692545299e-07])
+        assert_allclose(sc.erfinv(x), expected,
+                        rtol=1e-15)
+
+        # also test the roundtrip consistency
+        assert_allclose(sc.erf(sc.erfinv(x)),
+                        x,
+                        rtol=5e-15)

.venv/Lib/site-packages/scipy/special/tests/test_exponential_integrals.py

@@ -0,0 +1,118 @@
+import pytest
+
+import numpy as np
+from numpy.testing import assert_allclose
+import scipy.special as sc
+
+
+class TestExp1:
+
+    def test_branch_cut(self):
+        assert np.isnan(sc.exp1(-1))
+        assert sc.exp1(complex(-1, 0)).imag == (
+            -sc.exp1(complex(-1, -0.0)).imag
+        )
+
+        assert_allclose(
+            sc.exp1(complex(-1, 0)),
+            sc.exp1(-1 + 1e-20j),
+            atol=0,
+            rtol=1e-15
+        )
+        assert_allclose(
+            sc.exp1(complex(-1, -0.0)),
+            sc.exp1(-1 - 1e-20j),
+            atol=0,
+            rtol=1e-15
+        )
+
+    def test_834(self):
+        # Regression test for #834
+        a = sc.exp1(-complex(19.9999990))
+        b = sc.exp1(-complex(19.9999991))
+        assert_allclose(a.imag, b.imag, atol=0, rtol=1e-15)
+
+
+class TestScaledExp1:
+
+    @pytest.mark.parametrize('x, expected', [(0, 0), (np.inf, 1)])
+    def test_limits(self, x, expected):
+        y = sc._ufuncs._scaled_exp1(x)
+        assert y == expected
+
+    # The expected values were computed with mpmath, e.g.:
+    #
+    #   from mpmath import mp
+    #   mp.dps = 80
+    #   x = 1e-25
+    #   print(float(x*mp.exp(x)*np.expint(1, x)))
+    #
+    # prints 5.698741165994961e-24
+    #
+    # The method used to compute _scaled_exp1 changes at x=1
+    # and x=1250, so values at those inputs, and values just
+    # above and below them, are included in the test data.
+    @pytest.mark.parametrize('x, expected',
+                             [(1e-25, 5.698741165994961e-24),
+                              (0.1, 0.20146425447084518),
+                              (0.9995, 0.5962509885831002),
+                              (1.0, 0.5963473623231941),
+                              (1.0005, 0.5964436833238044),
+                              (2.5, 0.7588145912149602),
+                              (10.0, 0.9156333393978808),
+                              (100.0, 0.9901942286733019),
+                              (500.0, 0.9980079523802055),
+                              (1000.0, 0.9990019940238807),
+                              (1249.5, 0.9992009578306811),
+                              (1250.0, 0.9992012769377913),
+                              (1250.25, 0.9992014363957858),
+                              (2000.0, 0.9995004992514963),
+                              (1e4, 0.9999000199940024),
+                              (1e10, 0.9999999999),
+                              (1e15, 0.999999999999999),
+                              ])
+    def test_scaled_exp1(self, x, expected):
+        y = sc._ufuncs._scaled_exp1(x)
+        assert_allclose(y, expected, rtol=2e-15)
+
+
+class TestExpi:
+
+    @pytest.mark.parametrize('result', [
+        sc.expi(complex(-1, 0)),
+        sc.expi(complex(-1, -0.0)),
+        sc.expi(-1)
+    ])
+    def test_branch_cut(self, result):
+        desired = -0.21938393439552027368  # Computed using Mpmath
+        assert_allclose(result, desired, atol=0, rtol=1e-14)
+
+    def test_near_branch_cut(self):
+        lim_from_above = sc.expi(-1 + 1e-20j)
+        lim_from_below = sc.expi(-1 - 1e-20j)
+        assert_allclose(
+            lim_from_above.real,
+            lim_from_below.real,
+            atol=0,
+            rtol=1e-15
+        )
+        assert_allclose(
+            lim_from_above.imag,
+            -lim_from_below.imag,
+            atol=0,
+            rtol=1e-15
+        )
+
+    def test_continuity_on_positive_real_axis(self):
+        assert_allclose(
+            sc.expi(complex(1, 0)),
+            sc.expi(complex(1, -0.0)),
+            atol=0,
+            rtol=1e-15
+        )
+
+
+class TestExpn:
+
+    def test_out_of_domain(self):
+        assert all(np.isnan([sc.expn(-1, 1.0), sc.expn(1, -1.0)]))

.venv/Lib/site-packages/scipy/special/tests/test_faddeeva.py

@@ -0,0 +1,85 @@
+import pytest
+
+import numpy as np
+from numpy.testing import assert_allclose
+import scipy.special as sc
+from scipy.special._testutils import FuncData
+
+
+class TestVoigtProfile:
+
+    @pytest.mark.parametrize('x, sigma, gamma', [
+        (np.nan, 1, 1),
+        (0, np.nan, 1),
+        (0, 1, np.nan),
+        (1, np.nan, 0),
+        (np.nan, 1, 0),
+        (1, 0, np.nan),
+        (np.nan, 0, 1),
+        (np.nan, 0, 0)
+    ])
+    def test_nan(self, x, sigma, gamma):
+        assert np.isnan(sc.voigt_profile(x, sigma, gamma))
+
+    @pytest.mark.parametrize('x, desired', [
+        (-np.inf, 0),
+        (np.inf, 0)
+    ])
+    def test_inf(self, x, desired):
+        assert sc.voigt_profile(x, 1, 1) == desired
+
+    def test_against_mathematica(self):
+        # Results obtained from Mathematica by computing
+        #
+        # PDF[VoigtDistribution[gamma, sigma], x]
+        #
+        points = np.array([
+            [-7.89, 45.06, 6.66, 0.0077921073660388806401],
+            [-0.05, 7.98, 24.13, 0.012068223646769913478],
+            [-13.98, 16.83, 42.37, 0.0062442236362132357833],
+            [-12.66, 0.21, 6.32, 0.010052516161087379402],
+            [11.34, 4.25, 21.96, 0.0113698923627278917805],
+            [-11.56, 20.40, 30.53, 0.0076332760432097464987],
+            [-9.17, 25.61, 8.32, 0.011646345779083005429],
+            [16.59, 18.05, 2.50, 0.013637768837526809181],
+            [9.11, 2.12, 39.33, 0.0076644040807277677585],
+            [-43.33, 0.30, 45.68, 0.0036680463875330150996]
+        ])
+        FuncData(
+            sc.voigt_profile,
+            points,
+            (0, 1, 2),
+            3,
+            atol=0,
+            rtol=1e-15
+        ).check()
+
+    def test_symmetry(self):
+        x = np.linspace(0, 10, 20)
+        assert_allclose(
+            sc.voigt_profile(x, 1, 1),
+            sc.voigt_profile(-x, 1, 1),
+            rtol=1e-15,
+            atol=0
+        )
+
+    @pytest.mark.parametrize('x, sigma, gamma, desired', [
+        (0, 0, 0, np.inf),
+        (1, 0, 0, 0)
+    ])
+    def test_corner_cases(self, x, sigma, gamma, desired):
+        assert sc.voigt_profile(x, sigma, gamma) == desired
+
+    @pytest.mark.parametrize('sigma1, gamma1, sigma2, gamma2', [
+        (0, 1, 1e-16, 1),
+        (1, 0, 1, 1e-16),
+        (0, 0, 1e-16, 1e-16)
+    ])
+    def test_continuity(self, sigma1, gamma1, sigma2, gamma2):
+        x = np.linspace(1, 10, 20)
+        assert_allclose(
+            sc.voigt_profile(x, sigma1, gamma1),
+            sc.voigt_profile(x, sigma2, gamma2),
+            rtol=1e-16,
+            atol=1e-16
+        )

.venv/Lib/site-packages/scipy/special/tests/test_gamma.py

@@ -0,0 +1,12 @@
+import numpy as np
+import scipy.special as sc
+
+
+class TestRgamma:
+
+    def test_gh_11315(self):
+        assert sc.rgamma(-35) == 0
+
+    def test_rgamma_zeros(self):
+        x = np.array([0, -10, -100, -1000, -10000])
+        assert np.all(sc.rgamma(x) == 0)

.venv/Lib/site-packages/scipy/special/tests/test_gammainc.py

@@ -0,0 +1,136 @@
+import pytest
+
+import numpy as np
+from numpy.testing import assert_allclose, assert_array_equal
+
+import scipy.special as sc
+from scipy.special._testutils import FuncData
+
+
+INVALID_POINTS = [
+    (1, -1),
+    (0, 0),
+    (-1, 1),
+    (np.nan, 1),
+    (1, np.nan)
+]
+
+
+class TestGammainc:
+
+    @pytest.mark.parametrize('a, x', INVALID_POINTS)
+    def test_domain(self, a, x):
+        assert np.isnan(sc.gammainc(a, x))
+
+    def test_a_eq_0_x_gt_0(self):
+        assert sc.gammainc(0, 1) == 1
+
+    @pytest.mark.parametrize('a, x, desired', [
+        (np.inf, 1, 0),
+        (np.inf, 0, 0),
+        (np.inf, np.inf, np.nan),
+        (1, np.inf, 1)
+    ])
+    def test_infinite_arguments(self, a, x, desired):
+        result = sc.gammainc(a, x)
+        if np.isnan(desired):
+            assert np.isnan(result)
+        else:
+            assert result == desired
+
+    def test_infinite_limits(self):
+        # Test that large arguments converge to the hard-coded limits
+        # at infinity.
+        assert_allclose(
+            sc.gammainc(1000, 100),
+            sc.gammainc(np.inf, 100),
+            atol=1e-200,  # Use `atol` since the function converges to 0.
+            rtol=0
+        )
+        assert sc.gammainc(100, 1000) == sc.gammainc(100, np.inf)
+
+    def test_x_zero(self):
+        a = np.arange(1, 10)
+        assert_array_equal(sc.gammainc(a, 0), 0)
+
+    def test_limit_check(self):
+        result = sc.gammainc(1e-10, 1)
+        limit = sc.gammainc(0, 1)
+        assert np.isclose(result, limit)
+
+    def gammainc_line(self, x):
+        # The line a = x where a simpler asymptotic expansion (analog
+        # of DLMF 8.12.15) is available.
+        c = np.array([-1/3, -1/540, 25/6048, 101/155520,
+                      -3184811/3695155200, -2745493/8151736420])
+        res = 0
+        xfac = 1
+        for ck in c:
+            res -= ck*xfac
+            xfac /= x
+        res /= np.sqrt(2*np.pi*x)
+        res += 0.5
+        return res
+
+    def test_line(self):
+        x = np.logspace(np.log10(25), 300, 500)
+        a = x
+        dataset = np.vstack((a, x, self.gammainc_line(x))).T
+        FuncData(sc.gammainc, dataset, (0, 1), 2, rtol=1e-11).check()
+
+    def test_roundtrip(self):
+        a = np.logspace(-5, 10, 100)
+        x = np.logspace(-5, 10, 100)
+
+        y = sc.gammaincinv(a, sc.gammainc(a, x))
+        assert_allclose(x, y, rtol=1e-10)
+
+
+class TestGammaincc:
+
+    @pytest.mark.parametrize('a, x', INVALID_POINTS)
+    def test_domain(self, a, x):
+        assert np.isnan(sc.gammaincc(a, x))
+
+    def test_a_eq_0_x_gt_0(self):
+        assert sc.gammaincc(0, 1) == 0
+
+    @pytest.mark.parametrize('a, x, desired', [
+        (np.inf, 1, 1),
+        (np.inf, 0, 1),
+        (np.inf, np.inf, np.nan),
+        (1, np.inf, 0)
+    ])
+    def test_infinite_arguments(self, a, x, desired):
+        result = sc.gammaincc(a, x)
+        if np.isnan(desired):
+            assert np.isnan(result)
+        else:
+            assert result == desired
+
+    def test_infinite_limits(self):
+        # Test that large arguments converge to the hard-coded limits
+        # at infinity.
+        assert sc.gammaincc(1000, 100) == sc.gammaincc(np.inf, 100)
+        assert_allclose(
+            sc.gammaincc(100, 1000),
+            sc.gammaincc(100, np.inf),
+            atol=1e-200,  # Use `atol` since the function converges to 0.
+            rtol=0
+        )
+
+    def test_limit_check(self):
+        result = sc.gammaincc(1e-10,1)
+        limit = sc.gammaincc(0,1)
+        assert np.isclose(result, limit)
+
+    def test_x_zero(self):
+        a = np.arange(1, 10)
+        assert_array_equal(sc.gammaincc(a, 0), 1)
+
+    def test_roundtrip(self):
+        a = np.logspace(-5, 10, 100)
+        x = np.logspace(-5, 10, 100)
+
+        y = sc.gammainccinv(a, sc.gammaincc(a, x))
+        assert_allclose(x, y, rtol=1e-14)

.venv/Lib/site-packages/scipy/special/tests/test_hyp2f1.py

@@ -0,0 +1,2180 @@
+"""Tests for hyp2f1 for complex values.
+
+Author: Albert Steppi, with credit to Adam Kullberg (FormerPhycisist) for
+the implementation of mp_hyp2f1 below, which modifies mpmath's hyp2f1 to
+return the same branch as scipy's on the standard branch cut.
+"""
+
+import sys
+import pytest
+import numpy as np
+from typing import NamedTuple
+from numpy.testing import assert_allclose
+
+from scipy.special import hyp2f1
+from scipy.special._testutils import check_version, MissingModule
+
+
+try:
+    import mpmath
+except ImportError:
+    mpmath = MissingModule("mpmath")
+
+
+def mp_hyp2f1(a, b, c, z):
+    """Return mpmath hyp2f1 calculated on same branch as scipy hyp2f1.
+
+    For most values of a,b,c mpmath returns the x - 0j branch of hyp2f1 on the
+    branch cut x=(1,inf) whereas scipy's hyp2f1 calculates the x + 0j branch.
+    Thus, to generate the right comparison values on the branch cut, we
+    evaluate mpmath.hyp2f1 at x + 1e-15*j.
+
+    The exception to this occurs when c-a=-m in which case both mpmath and
+    scipy calculate the x + 0j branch on the branch cut. When this happens
+    mpmath.hyp2f1 will be evaluated at the original z point.
+    """
+    on_branch_cut = z.real > 1.0 and abs(z.imag) < 1.0e-15
+    cond1 = abs(c - a - round(c - a)) < 1.0e-15 and round(c - a) <= 0
+    cond2 = abs(c - b - round(c - b)) < 1.0e-15 and round(c - b) <= 0
+    # Make sure imaginary part is *exactly* zero
+    if on_branch_cut:
+        z = z.real + 0.0j
+    if on_branch_cut and not (cond1 or cond2):
+        z_mpmath = z.real + 1.0e-15j
+    else:
+        z_mpmath = z
+    return complex(mpmath.hyp2f1(a, b, c, z_mpmath))
+
+
+class Hyp2f1TestCase(NamedTuple):
+    a: float
+    b: float
+    c: float
+    z: complex
+    expected: complex
+    rtol: float
+
+
+class TestHyp2f1:
+    """Tests for hyp2f1 for complex values.
+
+    Expected values for test cases were computed using mpmath. See
+    `scipy.special._precompute.hyp2f1_data`. The verbose style of specifying
+    test cases is used for readability and to make it easier to mark individual
+    cases as expected to fail. Expected failures are used to highlight cases
+    where improvements are needed. See
+    `scipy.special._precompute.hyp2f1_data.make_hyp2f1_test_cases` for a
+    function to generate the boilerplate for the test cases.
+
+    Assertions have been added to each test to ensure that the test cases match
+    the situations that are intended. A final test `test_test_hyp2f1` checks
+    that the expected values in the test cases actually match what is computed
+    by mpmath. This test is marked slow even though it isn't particularly slow
+    so that it won't run by default on continuous integration builds.
+    """
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=0.2,
+                    c=-10,
+                    z=0.2 + 0.2j,
+                    expected=np.inf + 0j,
+                    rtol=0
+                )
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=0.2,
+                    c=-10,
+                    z=0 + 0j,
+                    expected=1 + 0j,
+                    rtol=0
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=0,
+                    c=-10,
+                    z=0.2 + 0.2j,
+                    expected=1 + 0j,
+                    rtol=0
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=0,
+                    c=0,
+                    z=0.2 + 0.2j,
+                    expected=1 + 0j,
+                    rtol=0,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=0.2,
+                    c=0,
+                    z=0.2 + 0.2j,
+                    expected=np.inf + 0j,
+                    rtol=0,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=0.2,
+                    c=0,
+                    z=0 + 0j,
+                    expected=np.nan + 0j,
+                    rtol=0,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=-5,
+                    c=-10,
+                    z=0.2 + 0.2j,
+                    expected=(1.0495404166666666+0.05708208333333334j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=-10,
+                    c=-10,
+                    z=0.2 + 0.2j,
+                    expected=(1.092966013125+0.13455014673750001j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-10,
+                    b=-20,
+                    c=-10,
+                    z=0.2 + 0.2j,
+                    expected=(-0.07712512000000005+0.12752814080000005j),
+                    rtol=1e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1,
+                    b=3.2,
+                    c=-1,
+                    z=0.2 + 0.2j,
+                    expected=(1.6400000000000001+0.6400000000000001j),
+                    rtol=1e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-2,
+                    b=1.2,
+                    c=-4,
+                    z=1 + 0j,
+                    expected=1.8200000000000001 + 0j,
+                    rtol=1e-15,
+                ),
+            ),
+        ]
+    )
+    def test_c_non_positive_int(self, hyp2f1_test_case):
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=0.2,
+                    c=1.5,
+                    z=1 + 0j,
+                    expected=1.1496439092239847 + 0j,
+                    rtol=1e-15
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=12.3,
+                    b=8.0,
+                    c=20.31,
+                    z=1 + 0j,
+                    expected=69280986.75273195 + 0j,
+                    rtol=1e-15
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=290.2,
+                    b=321.5,
+                    c=700.1,
+                    z=1 + 0j,
+                    expected=1.3396562400934e117 + 0j,
+                    rtol=1e-12,
+                ),
+            ),
+            # Note that here even mpmath produces different results for
+            # results that should be equivalent.
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=9.2,
+                    b=621.5,
+                    c=700.1,
+                    z=(1+0j),
+                    expected=(952726652.4158565+0j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=621.5,
+                    b=9.2,
+                    c=700.1,
+                    z=(1+0j),
+                    expected=(952726652.4160284+0j),
+                    rtol=5e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-101.2,
+                    b=-400.4,
+                    c=-172.1,
+                    z=(1+0j),
+                    expected=(2.2253618341394838e+37+0j),
+                    rtol=1e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-400.4,
+                    b=-101.2,
+                    c=-172.1,
+                    z=(1+0j),
+                    expected=(2.2253618341394838e+37+0j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=172.5,
+                    b=-201.3,
+                    c=151.2,
+                    z=(1+0j),
+                    expected=(7.072266653650905e-135+0j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-201.3,
+                    b=172.5,
+                    c=151.2,
+                    z=(1+0j),
+                    expected=(7.072266653650905e-135+0j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-102.1,
+                    b=-20.3,
+                    c=1.3,
+                    z=1 + 0j,
+                    expected=2.7899070752746906e22 + 0j,
+                    rtol=3e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-202.6,
+                    b=60.3,
+                    c=1.5,
+                    z=1 + 0j,
+                    expected=-1.3113641413099326e-56 + 0j,
+                    rtol=1e-12,
+                ),
+            ),
+        ],
+    )
+    def test_unital_argument(self, hyp2f1_test_case):
+        """Tests for case z = 1, c - a - b > 0.
+
+        Expected answers computed using mpmath.
+        """
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert z == 1 and c - a - b > 0  # Tests the test
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=0.5,
+                    b=0.2,
+                    c=1.3,
+                    z=-1 + 0j,
+                    expected=0.9428846409614143 + 0j,
+                    rtol=1e-15),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=12.3,
+                    b=8.0,
+                    c=5.300000000000001,
+                    z=-1 + 0j,
+                    expected=-4.845809986595704e-06 + 0j,
+                    rtol=1e-15
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=221.5,
+                    b=90.2,
+                    c=132.3,
+                    z=-1 + 0j,
+                    expected=2.0490488728377282e-42 + 0j,
+                    rtol=1e-7,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-102.1,
+                    b=-20.3,
+                    c=-80.8,
+                    z=-1 + 0j,
+                    expected=45143784.46783885 + 0j,
+                    rtol=1e-7,
+                ),
+                marks=pytest.mark.xfail(
+                    condition=sys.maxsize < 2**32,
+                    reason="Fails on 32 bit.",
+                )
+            ),
+        ],
+    )
+    def test_special_case_z_near_minus_1(self, hyp2f1_test_case):
+        """Tests for case z ~ -1, c ~ 1 + a - b
+
+        Expected answers computed using mpmath.
+        """
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert abs(1 + a - b - c) < 1e-15 and abs(z + 1) < 1e-15
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-4,
+                    b=2.02764642551431,
+                    c=1.0561196186065624,
+                    z=(0.9473684210526314-0.10526315789473695j),
+                    expected=(0.0031961077109535375-0.0011313924606557173j),
+                    rtol=1e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-8,
+                    b=-7.937789122896016,
+                    c=-15.964218273004214,
+                    z=(2-0.10526315789473695j),
+                    expected=(0.005543763196412503-0.0025948879065698306j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-8,
+                    b=8.095813935368371,
+                    c=4.0013768449590685,
+                    z=(0.9473684210526314-0.10526315789473695j),
+                    expected=(-0.0003054674127221263-9.261359291755414e-05j),
+                    rtol=1e-10,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-4,
+                    b=-3.956227226099288,
+                    c=-3.9316537064827854,
+                    z=(1.1578947368421053-0.3157894736842106j),
+                    expected=(-0.0020809502580892937-0.0041877333232365095j),
+                    rtol=5e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=-4,
+                    c=2.050308316530781,
+                    z=(0.9473684210526314-0.10526315789473695j),
+                    expected=(0.0011282435590058734+0.0002027062303465851j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=-8,
+                    c=-15.964218273004214,
+                    z=(1.3684210526315788+0.10526315789473673j),
+                    expected=(-9.134907719238265e-05-0.00040219233987390723j),
+                    rtol=5e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=-4,
+                    c=4.0013768449590685,
+                    z=(0.9473684210526314-0.10526315789473695j),
+                    expected=(-0.000519013062087489-0.0005855883076830948j),
+                    rtol=5e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-10000,
+                    b=2.2,
+                    c=93459345.3,
+                    z=(2+2j),
+                    expected=(0.9995292071559088-0.00047047067522659253j),
+                    rtol=1e-12,
+                ),
+            ),
+        ]
+    )
+    def test_a_b_negative_int(self, hyp2f1_test_case):
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert a == int(a) and a < 0 or b == int(b) and b < 0  # Tests the test
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.5,
+                    b=-0.9629749245209605,
+                    c=-15.5,
+                    z=(1.1578947368421053-1.1578947368421053j),
+                    expected=(0.9778506962676361+0.044083801141231616j),
+                    rtol=3e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.5,
+                    b=-3.9316537064827854,
+                    c=1.5,
+                    z=(0.9473684210526314-0.10526315789473695j),
+                    expected=(4.0793167523167675-10.11694246310966j),
+                    rtol=6e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.5,
+                    b=-0.9629749245209605,
+                    c=2.5,
+                    z=(1.1578947368421053-0.10526315789473695j),
+                    expected=(-2.9692999501916915+0.6394599899845594j),
+                    rtol=1e-11,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.5,
+                    b=-0.9629749245209605,
+                    c=-15.5,
+                    z=(1.5789473684210522-1.1578947368421053j),
+                    expected=(0.9493076367106102-0.04316852977183447j),
+                    rtol=1e-11,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=-0.5,
+                    c=-15.5,
+                    z=(0.5263157894736841+0.10526315789473673j),
+                    expected=(0.9844377175631795-0.003120587561483841j),
+                    rtol=1e-10,
+                ),
+            ),
+        ],
+    )
+    def test_a_b_neg_int_after_euler_hypergeometric_transformation(
+        self, hyp2f1_test_case
+    ):
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert (  # Tests the test
+            (abs(c - a - int(c - a)) < 1e-15 and c - a < 0) or
+            (abs(c - b - int(c - b)) < 1e-15 and c - b < 0)
+        )
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=-0.9629749245209605,
+                    c=-15.963511401609862,
+                    z=(0.10526315789473673-0.3157894736842106j),
+                    expected=(0.9941449585778349+0.01756335047931358j),
+                    rtol=1e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0272592605282642,
+                    b=-0.9629749245209605,
+                    c=-15.963511401609862,
+                    z=(0.5263157894736841+0.5263157894736841j),
+                    expected=(1.0388722293372104-0.09549450380041416j),
+                    rtol=5e-11,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=1.0561196186065624,
+                    c=-7.93846038215665,
+                    z=(0.10526315789473673+0.7368421052631575j),
+                    expected=(2.1948378809826434+24.934157235172222j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=16.088264119063613,
+                    c=8.031683612216888,
+                    z=(0.3157894736842106-0.736842105263158j),
+                    expected=(-0.4075277891264672-0.06819344579666956j),
+                    rtol=2e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=2.050308316530781,
+                    c=8.031683612216888,
+                    z=(0.7368421052631575-0.10526315789473695j),
+                    expected=(2.833535530740603-0.6925373701408158j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=2.050308316530781,
+                    c=4.078873014294075,
+                    z=(0.10526315789473673-0.3157894736842106j),
+                    expected=(1.005347176329683-0.3580736009337313j),
+                    rtol=5e-16,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=-0.9629749245209605,
+                    c=-15.963511401609862,
+                    z=(0.3157894736842106-0.5263157894736843j),
+                    expected=(0.9824353641135369+0.029271018868990268j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=-0.9629749245209605,
+                    c=-159.63511401609862,
+                    z=(0.3157894736842106-0.5263157894736843j),
+                    expected=(0.9982436200365834+0.002927268199671111j),
+                    rtol=1e-7,
+                ),
+                marks=pytest.mark.xfail(reason="Poor convergence.")
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=16.088264119063613,
+                    c=8.031683612216888,
+                    z=(0.5263157894736841-0.5263157894736843j),
+                    expected=(-0.6906825165778091+0.8176575137504892j),
+                    rtol=5e-13,
+                ),
+            ),
+        ]
+    )
+    def test_region1(self, hyp2f1_test_case):
+        """|z| < 0.9 and real(z) >= 0."""
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert abs(z) < 0.9 and z.real >= 0  # Tests the test
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=1.0561196186065624,
+                    c=4.078873014294075,
+                    z=(-0.3157894736842106+0.7368421052631575j),
+                    expected=(0.7751915029081136+0.24068493258607315j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.087593263474208,
+                    b=16.088264119063613,
+                    c=2.0397202577726152,
+                    z=(-0.9473684210526316-0.3157894736842106j),
+                    expected=(6.564549348474962e-07+1.6761570598334562e-06j),
+                    rtol=5e-09,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0272592605282642,
+                    b=2.050308316530781,
+                    c=16.056809865262608,
+                    z=(-0.10526315789473695-0.10526315789473695j),
+                    expected=(0.9862043298997204-0.013293151372712681j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=8.077282662161238,
+                    c=16.056809865262608,
+                    z=(-0.3157894736842106-0.736842105263158j),
+                    expected=(0.16163826638754716-0.41378530376373734j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=2.050308316530781,
+                    c=-0.906685989801748,
+                    z=(-0.5263157894736843+0.3157894736842106j),
+                    expected=(-6.256871535165936+0.13824973858225484j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=8.077282662161238,
+                    c=-3.9924618758357022,
+                    z=(-0.9473684210526316-0.3157894736842106j),
+                    expected=(75.54672526086316+50.56157041797548j),
+                    rtol=5e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.087593263474208,
+                    b=8.077282662161238,
+                    c=-1.9631175993998025,
+                    z=(-0.5263157894736843+0.5263157894736841j),
+                    expected=(282.0602536306534-82.31597306936214j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.095813935368371,
+                    b=-3.9316537064827854,
+                    c=8.031683612216888,
+                    z=(-0.5263157894736843-0.10526315789473695j),
+                    expected=(5.179603735575851+1.4445374002099813j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=-7.949900487447654,
+                    c=1.0651378143226575,
+                    z=(-0.3157894736842106-0.9473684210526316j),
+                    expected=(2317.623517606141-269.51476321010324j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.087593263474208,
+                    b=-1.92872979730171,
+                    c=2.0397202577726152,
+                    z=(-0.736842105263158-0.3157894736842106j),
+                    expected=(29.179154096175836+22.126690357535043j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.095813935368371,
+                    b=-3.9316537064827854,
+                    c=-15.963511401609862,
+                    z=(-0.736842105263158-0.10526315789473695j),
+                    expected=(0.20820247892032057-0.04763956711248794j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0272592605282642,
+                    b=-15.964218273004214,
+                    c=-1.9631175993998025,
+                    z=(-0.3157894736842106-0.5263157894736843j),
+                    expected=(-157471.63920142158+991294.0587828817j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.095813935368371,
+                    b=-7.949900487447654,
+                    c=-7.93846038215665,
+                    z=(-0.10526315789473695-0.10526315789473695j),
+                    expected=(0.30765349653210194-0.2979706363594157j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.956227226099288,
+                    b=1.0561196186065624,
+                    c=8.031683612216888,
+                    z=(-0.9473684210526316-0.10526315789473695j),
+                    expected=(1.6787607400597109+0.10056620134616838j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=16.088264119063613,
+                    c=4.078873014294075,
+                    z=(-0.5263157894736843-0.736842105263158j),
+                    expected=(7062.07842506049-12768.77955655703j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=16.088264119063613,
+                    c=2.0397202577726152,
+                    z=(-0.3157894736842106+0.7368421052631575j),
+                    expected=(54749.216391029935-23078.144720887536j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.956227226099288,
+                    b=1.0561196186065624,
+                    c=-0.906685989801748,
+                    z=(-0.10526315789473695-0.10526315789473695j),
+                    expected=(1.21521766411428-4.449385173946672j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.980848054962111,
+                    b=4.0013768449590685,
+                    c=-1.9631175993998025,
+                    z=(-0.736842105263158+0.5263157894736841j),
+                    expected=(19234693144.196907+1617913967.7294445j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=1.0561196186065624,
+                    c=-15.963511401609862,
+                    z=(-0.5263157894736843+0.3157894736842106j),
+                    expected=(0.9345201094534371+0.03745712558992195j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=-0.9629749245209605,
+                    c=2.0397202577726152,
+                    z=(-0.10526315789473695+0.10526315789473673j),
+                    expected=(0.605732446296829+0.398171533680972j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=-15.964218273004214,
+                    c=2.0397202577726152,
+                    z=(-0.10526315789473695-0.5263157894736843j),
+                    expected=(-9.753761888305416-4.590126012666959j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.956227226099288,
+                    b=-1.92872979730171,
+                    c=2.0397202577726152,
+                    z=(-0.10526315789473695+0.3157894736842106j),
+                    expected=(0.45587226291120714+1.0694545265819797j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=-7.949900487447654,
+                    c=-0.906685989801748,
+                    z=(-0.736842105263158+0.3157894736842106j),
+                    expected=(12.334808243233418-76.26089051819054j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=-7.949900487447654,
+                    c=-15.963511401609862,
+                    z=(-0.5263157894736843+0.10526315789473673j),
+                    expected=(1.2396019687632678-0.047507973161146286j),
+                    rtol=1e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.980848054962111,
+                    b=-0.9629749245209605,
+                    c=-0.906685989801748,
+                    z=(-0.3157894736842106-0.5263157894736843j),
+                    expected=(97.7889554372208-18.999754543400016j),
+                    rtol=5e-13,
+                ),
+            ),
+        ]
+    )
+    def test_region2(self, hyp2f1_test_case):
+        """|z| < 1 and real(z) < 0."""
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert abs(z) < 1 and z.real < 0  # Tests the test
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.25,
+                    b=4.25,
+                    c=2.5,
+                    z=(0.4931034482758623-0.7965517241379311j),
+                    expected=(38.41207903409937-30.510151276075792j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.0,
+                    b=16.087593263474208,
+                    c=16.088264119063613,
+                    z=(0.5689655172413794-0.7965517241379311j),
+                    expected=(-0.6667857912761286-1.0206224321443573j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.0,
+                    b=1.0272592605282642,
+                    c=-7.949900487447654,
+                    z=(0.4931034482758623-0.7965517241379311j),
+                    expected=(1679024.1647997478-2748129.775857212j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=16.0,
+                    c=-7.949900487447654,
+                    z=(0.4931034482758623-0.7965517241379311j),
+                    expected=(424747226301.16986-1245539049327.2856j),
+                    rtol=1e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=-15.964218273004214,
+                    c=4.0,
+                    z=(0.4931034482758623-0.7965517241379311j),
+                    expected=(-0.0057826199201757595+0.026359861999025885j),
+                    rtol=5e-06,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=-0.9629749245209605,
+                    c=2.0397202577726152,
+                    z=(0.5689655172413794-0.7965517241379311j),
+                    expected=(0.4671901063492606+0.7769632229834897j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.0,
+                    b=-3.956227226099288,
+                    c=-7.949900487447654,
+                    z=(0.4931034482758623+0.7965517241379312j),
+                    expected=(0.9422283708145973+1.3476905754773343j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0,
+                    b=-15.980848054962111,
+                    c=-15.964218273004214,
+                    z=(0.4931034482758623-0.7965517241379311j),
+                    expected=(0.4168719497319604-0.9770953555235625j),
+                    rtol=5e-10,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.5,
+                    b=16.088264119063613,
+                    c=2.5,
+                    z=(0.5689655172413794+0.7965517241379312j),
+                    expected=(1.279096377550619-2.173827694297929j),
+                    rtol=5e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=4.0013768449590685,
+                    c=2.0397202577726152,
+                    z=(0.4931034482758623+0.7965517241379312j),
+                    expected=(-2.071520656161738-0.7846098268395909j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=8.0,
+                    c=-0.9629749245209605,
+                    z=(0.5689655172413794-0.7965517241379311j),
+                    expected=(-7.740015495862889+3.386766435696699j),
+                    rtol=5e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=16.088264119063613,
+                    c=-7.93846038215665,
+                    z=(0.4931034482758623+0.7965517241379312j),
+                    expected=(-6318.553685853241-7133.416085202879j),
+                    rtol=1e-10,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.980848054962111,
+                    b=-3.9316537064827854,
+                    c=16.056809865262608,
+                    z=(0.5689655172413794+0.7965517241379312j),
+                    expected=(-0.8854577905547399+8.135089099967278j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=-0.9629749245209605,
+                    c=4.078873014294075,
+                    z=(0.4931034482758623+0.7965517241379312j),
+                    expected=(1.224291301521487+0.36014711766402485j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.75,
+                    b=-0.75,
+                    c=-1.5,
+                    z=(0.4931034482758623+0.7965517241379312j),
+                    expected=(-1.5765685855028473-3.9399766961046323j),
+                    rtol=1e-3,
+                ),
+                marks=pytest.mark.xfail(
+                    reason="Unhandled parameters."
+                )
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.980848054962111,
+                    b=-1.92872979730171,
+                    c=-7.93846038215665,
+                    z=(0.5689655172413794-0.7965517241379311j),
+                    expected=(56.794588688231194+4.556286783533971j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.5,
+                    b=4.5,
+                    c=2.050308316530781,
+                    z=(0.5689655172413794+0.7965517241379312j),
+                    expected=(-4.251456563455306+6.737837111569671j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.5,
+                    b=8.5,
+                    c=-1.92872979730171,
+                    z=(0.4931034482758623-0.7965517241379311j),
+                    expected=(2177143.9156599627-3313617.2748088865j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.5,
+                    b=-1.5,
+                    c=4.0013768449590685,
+                    z=(0.4931034482758623-0.7965517241379311j),
+                    expected=(0.45563554481603946+0.6212000158060831j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.5,
+                    b=-7.5,
+                    c=-15.964218273004214,
+                    z=(0.4931034482758623+0.7965517241379312j),
+                    expected=(61.03201617828073-37.185626416756214j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.5,
+                    b=16.5,
+                    c=4.0013768449590685,
+                    z=(0.4931034482758623+0.7965517241379312j),
+                    expected=(-33143.425963520735+20790.608514722644j),
+                    rtol=1e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.5,
+                    b=4.5,
+                    c=-0.9629749245209605,
+                    z=(0.5689655172413794+0.7965517241379312j),
+                    expected=(30.778600270824423-26.65160354466787j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.5,
+                    b=-3.5,
+                    c=16.088264119063613,
+                    z=(0.5689655172413794-0.7965517241379311j),
+                    expected=(1.0629792615560487-0.08308454486044772j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.5,
+                    b=-7.5,
+                    c=-0.9629749245209605,
+                    z=(0.4931034482758623-0.7965517241379311j),
+                    expected=(17431.571802591767+3553.7129767034507j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.25,
+                    b=8.25,
+                    c=16.5,
+                    z=(0.11379310344827598+0.9482758620689657j),
+                    expected=(0.4468600750211926+0.7313214934036885j),
+                    rtol=1e-3,
+                ),
+                marks=pytest.mark.xfail(
+                    reason="Unhandled parameters."
+                )
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.25,
+                    b=16.25,
+                    c=4.5,
+                    z=(0.3413793103448277+0.8724137931034486j),
+                    expected=(-3.905704438293991+3.693347860329299j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.25,
+                    b=4.25,
+                    c=-0.5,
+                    z=(0.11379310344827598-0.9482758620689655j),
+                    expected=(-40.31777941834244-89.89852492432011j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0272592605282642,
+                    b=8.0,
+                    c=-15.964218273004214,
+                    z=(0.11379310344827598-0.9482758620689655j),
+                    expected=(52584.347773055284-109197.86244309516j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.095813935368371,
+                    b=-15.964218273004214,
+                    c=16.056809865262608,
+                    z=(0.03793103448275881+0.9482758620689657j),
+                    expected=(-1.187733570412592-1.5147865053584582j),
+                    rtol=5e-10,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=-3.9316537064827854,
+                    c=1.0651378143226575,
+                    z=(0.26551724137931054+0.9482758620689657j),
+                    expected=(13.077494677898947+35.071599628224966j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=-3.5,
+                    c=-3.5,
+                    z=(0.26551724137931054+0.8724137931034486j),
+                    expected=(-0.5359656237994614-0.2344483936591811j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.25,
+                    b=-3.75,
+                    c=-1.5,
+                    z=(0.26551724137931054+0.9482758620689657j),
+                    expected=(1204.8114871663133+64.41022826840198j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=16.0,
+                    c=4.0013768449590685,
+                    z=(0.03793103448275881-0.9482758620689655j),
+                    expected=(-9.85268872413994+7.011107558429154j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=16.0,
+                    c=4.0013768449590685,
+                    z=(0.3413793103448277-0.8724137931034484j),
+                    expected=(528.5522951158454-1412.21630264791j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.5,
+                    b=1.0561196186065624,
+                    c=-7.5,
+                    z=(0.4172413793103451+0.8724137931034486j),
+                    expected=(133306.45260685298+256510.7045225382j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=8.077282662161238,
+                    c=-15.963511401609862,
+                    z=(0.3413793103448277-0.8724137931034484j),
+                    expected=(-0.998555715276967+2.774198742229889j),
+                    rtol=5e-11,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.75,
+                    b=-0.75,
+                    c=1.5,
+                    z=(0.11379310344827598-0.9482758620689655j),
+                    expected=(2.072445019723025-2.9793504811373515j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.5,
+                    b=-1.92872979730171,
+                    c=1.5,
+                    z=(0.11379310344827598-0.9482758620689655j),
+                    expected=(-41.87581944176649-32.52980303527139j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.75,
+                    b=-15.75,
+                    c=-0.5,
+                    z=(0.11379310344827598-0.9482758620689655j),
+                    expected=(-3729.6214864209774-30627.510509112635j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.956227226099288,
+                    b=-15.964218273004214,
+                    c=-0.906685989801748,
+                    z=(0.03793103448275881+0.9482758620689657j),
+                    expected=(-131615.07820609974+145596.13384245415j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.5,
+                    b=16.5,
+                    c=16.088264119063613,
+                    z=(0.26551724137931054+0.8724137931034486j),
+                    expected=(0.18981844071070744+0.7855036242583742j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.5,
+                    b=8.5,
+                    c=-3.9316537064827854,
+                    z=(0.11379310344827598-0.9482758620689655j),
+                    expected=(110224529.2376068+128287212.04290268j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.5,
+                    b=-7.5,
+                    c=4.0013768449590685,
+                    z=(0.3413793103448277-0.8724137931034484j),
+                    expected=(0.2722302180888523-0.21790187837266162j),
+                    rtol=1e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.5,
+                    b=-7.5,
+                    c=-15.964218273004214,
+                    z=(0.11379310344827598-0.9482758620689655j),
+                    expected=(-2.8252338010989035+2.430661949756161j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.5,
+                    b=16.5,
+                    c=4.0013768449590685,
+                    z=(0.03793103448275881+0.9482758620689657j),
+                    expected=(-20.604894257647945+74.5109432558078j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.5,
+                    b=8.5,
+                    c=-0.9629749245209605,
+                    z=(0.3413793103448277+0.8724137931034486j),
+                    expected=(-2764422.521269463-3965966.9965808876j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.5,
+                    b=-0.5,
+                    c=1.0561196186065624,
+                    z=(0.26551724137931054+0.9482758620689657j),
+                    expected=(1.2262338560994905+0.6545051266925549j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.5,
+                    b=-15.5,
+                    c=-7.949900487447654,
+                    z=(0.4172413793103451-0.8724137931034484j),
+                    expected=(-2258.1590330318213+8860.193389158803j),
+                    rtol=1.4e-10,
+                ),
+            ),
+        ]
+    )
+    def test_region4(self, hyp2f1_test_case):
+        """0.9 <= |z| <= 1 and |1 - z| >= 1.
+
+        This region is unhandled by of the standard transformations and
+        needs special care.
+        """
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert 0.9 <= abs(z) <= 1 and abs(1 - z) >= 0.9  # Tests the test
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.5,
+                    b=16.088264119063613,
+                    c=8.5,
+                    z=(0.6448275862068968+0.8724137931034486j),
+                    expected=(0.018601324701770394-0.07618420586062377j),
+                    rtol=5e-08,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.25,
+                    b=4.25,
+                    c=4.5,
+                    z=(0.6448275862068968-0.8724137931034484j),
+                    expected=(-1.391549471425551-0.118036604903893j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=2.050308316530781,
+                    c=-1.9631175993998025,
+                    z=(0.6448275862068968+0.8724137931034486j),
+                    expected=(-2309.178768155151-1932.7247727595172j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.087593263474208,
+                    b=1.0,
+                    c=-15.964218273004214,
+                    z=(0.6448275862068968+0.8724137931034486j),
+                    expected=(85592537010.05054-8061416766688.324j),
+                    rtol=2e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.095813935368371,
+                    b=-0.5,
+                    c=1.5,
+                    z=(0.6448275862068968+0.8724137931034486j),
+                    expected=(1.2334498208515172-2.1639498536219732j),
+                    rtol=5e-11,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.087593263474208,
+                    b=-15.964218273004214,
+                    c=4.0,
+                    z=(0.6448275862068968+0.8724137931034486j),
+                    expected=(102266.35398605966-44976.97828737755j),
+                    rtol=1e-3,
+                ),
+                marks=pytest.mark.xfail(
+                    reason="Unhandled parameters."
+                )
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.0,
+                    b=-3.956227226099288,
+                    c=-15.964218273004214,
+                    z=(0.6448275862068968-0.8724137931034484j),
+                    expected=(-2.9590030930007236-4.190770764773225j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=-15.5,
+                    c=-7.5,
+                    z=(0.5689655172413794-0.8724137931034484j),
+                    expected=(-112554838.92074208+174941462.9202412j),
+                    rtol=5e-05,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.980848054962111,
+                    b=2.050308316530781,
+                    c=1.0,
+                    z=(0.6448275862068968-0.8724137931034484j),
+                    expected=(3.7519882374080145+7.360753798667486j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=2.050308316530781,
+                    c=4.0,
+                    z=(0.6448275862068968-0.8724137931034484j),
+                    expected=(0.000181132943964693+0.07742903103815582j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=4.0013768449590685,
+                    c=-1.9631175993998025,
+                    z=(0.5689655172413794+0.8724137931034486j),
+                    expected=(386338.760913596-386166.51762171905j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.980848054962111,
+                    b=8.0,
+                    c=-1.92872979730171,
+                    z=(0.6448275862068968+0.8724137931034486j),
+                    expected=(1348667126.3444858-2375132427.158893j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.5,
+                    b=-0.9629749245209605,
+                    c=4.5,
+                    z=(0.5689655172413794+0.8724137931034486j),
+                    expected=(1.428353429538678+0.6472718120804372j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=-0.9629749245209605,
+                    c=2.0397202577726152,
+                    z=(0.5689655172413794-0.8724137931034484j),
+                    expected=(3.1439267526119643-3.145305240375117j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=-15.964218273004214,
+                    c=-7.93846038215665,
+                    z=(0.6448275862068968-0.8724137931034484j),
+                    expected=(75.27467675681773+144.0946946292215j),
+                    rtol=1e-07,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.75,
+                    b=-7.75,
+                    c=-7.5,
+                    z=(0.5689655172413794+0.8724137931034486j),
+                    expected=(-0.3699450626264222+0.8732812475910993j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.5,
+                    b=16.5,
+                    c=1.0561196186065624,
+                    z=(0.5689655172413794-0.8724137931034484j),
+                    expected=(5.5361025821300665-2.4709693474656285j),
+                    rtol=5e-09,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.5,
+                    b=8.5,
+                    c=-3.9316537064827854,
+                    z=(0.6448275862068968-0.8724137931034484j),
+                    expected=(-782805.6699207705-537192.581278909j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.5,
+                    b=-15.5,
+                    c=1.0561196186065624,
+                    z=(0.6448275862068968+0.8724137931034486j),
+                    expected=(12.345113400639693-14.993248992902007j),
+                    rtol=0.0005,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.5,
+                    b=-0.5,
+                    c=-15.964218273004214,
+                    z=(0.6448275862068968+0.8724137931034486j),
+                    expected=(23.698109392667842+97.15002033534108j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.5,
+                    b=16.5,
+                    c=4.0013768449590685,
+                    z=(0.6448275862068968-0.8724137931034484j),
+                    expected=(1115.2978631811834+915.9212658718577j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.5,
+                    b=16.5,
+                    c=-0.9629749245209605,
+                    z=(0.6448275862068968+0.8724137931034486j),
+                    expected=(642077722221.6489+535274495398.21027j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.5,
+                    b=-3.5,
+                    c=4.0013768449590685,
+                    z=(0.5689655172413794+0.8724137931034486j),
+                    expected=(-5.689219222945697+16.877463062787143j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.5,
+                    b=-1.5,
+                    c=-0.9629749245209605,
+                    z=(0.5689655172413794-0.8724137931034484j),
+                    expected=(-44.32070290703576+1026.9127058617403j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.25,
+                    b=2.25,
+                    c=4.5,
+                    z=(0.11379310344827598-1.024137931034483j),
+                    expected=(-0.021965227124574663+0.009908300237809064j),
+                    rtol=1e-3,
+                ),
+                marks=pytest.mark.xfail(
+                    reason="Unhandled parameters."
+                )
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.02764642551431,
+                    b=1.5,
+                    c=16.5,
+                    z=(0.26551724137931054+1.024137931034483j),
+                    expected=(1.0046072901244183+0.19945500134119992j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.087593263474208,
+                    b=1.0,
+                    c=-3.9316537064827854,
+                    z=(0.3413793103448277+0.9482758620689657j),
+                    expected=(21022.30133421465+49175.98317370489j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=16.088264119063613,
+                    c=-1.9631175993998025,
+                    z=(0.4172413793103451-0.9482758620689655j),
+                    expected=(-7024239.358547302+2481375.02681063j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.25,
+                    b=-15.75,
+                    c=1.5,
+                    z=(0.18965517241379315+1.024137931034483j),
+                    expected=(92371704.94848-403546832.548352j),
+                    rtol=5e-06,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.5,
+                    b=-7.949900487447654,
+                    c=8.5,
+                    z=(0.26551724137931054-1.024137931034483j),
+                    expected=(1.9335109845308265+5.986542524829654j),
+                    rtol=5e-10,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.095813935368371,
+                    b=-1.92872979730171,
+                    c=-7.93846038215665,
+                    z=(0.4931034482758623+0.8724137931034486j),
+                    expected=(-122.52639696039328-59.72428067512221j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.25,
+                    b=-1.75,
+                    c=-1.5,
+                    z=(0.4931034482758623+0.9482758620689657j),
+                    expected=(-90.40642053579428+50.50649180047921j),
+                    rtol=5e-08,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.5,
+                    b=8.077282662161238,
+                    c=16.5,
+                    z=(0.4931034482758623+0.9482758620689657j),
+                    expected=(-0.2155745818150323-0.564628986876639j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=1.0561196186065624,
+                    c=8.031683612216888,
+                    z=(0.4172413793103451-0.9482758620689655j),
+                    expected=(0.9503140488280465+0.11574960074292677j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.75,
+                    b=2.25,
+                    c=-15.5,
+                    z=(0.4172413793103451+0.9482758620689657j),
+                    expected=(0.9285862488442175+0.8203699266719692j),
+                    rtol=5e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.75,
+                    b=4.25,
+                    c=-15.5,
+                    z=(0.3413793103448277-0.9482758620689655j),
+                    expected=(-1.0509834850116921-1.1145522325486075j),
+                    rtol=1e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=-0.9629749245209605,
+                    c=2.0397202577726152,
+                    z=(0.4931034482758623-0.9482758620689655j),
+                    expected=(2.88119116536769-3.4249933450696806j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.5,
+                    b=-15.964218273004214,
+                    c=16.5,
+                    z=(0.18965517241379315+1.024137931034483j),
+                    expected=(199.65868451496038+347.79384207302877j),
+                    rtol=1e-13,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.75,
+                    b=-15.75,
+                    c=-3.5,
+                    z=(0.4931034482758623-0.8724137931034484j),
+                    expected=(-208138312553.07013+58631611809.026955j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=-15.5,
+                    c=-7.5,
+                    z=(0.3413793103448277+0.9482758620689657j),
+                    expected=(-23032.90519856288-18256.94050457296j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.5,
+                    b=1.5,
+                    c=1.0561196186065624,
+                    z=(0.4931034482758623-0.8724137931034484j),
+                    expected=(1.507342459587056+1.2332023580148403j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=2.5,
+                    b=4.5,
+                    c=-3.9316537064827854,
+                    z=(0.4172413793103451+0.9482758620689657j),
+                    expected=(7044.766127108853-40210.365567285575j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.5,
+                    b=-1.5,
+                    c=1.0561196186065624,
+                    z=(0.03793103448275881+1.024137931034483j),
+                    expected=(0.2725347741628333-2.247314875514784j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.5,
+                    b=-1.5,
+                    c=-7.949900487447654,
+                    z=(0.26551724137931054+1.024137931034483j),
+                    expected=(-11.250200011017546+12.597393659160472j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.5,
+                    b=8.5,
+                    c=16.088264119063613,
+                    z=(0.26551724137931054+1.024137931034483j),
+                    expected=(-0.18515160890991517+0.7959014164484782j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.5,
+                    b=16.5,
+                    c=-3.9316537064827854,
+                    z=(0.3413793103448277-1.024137931034483j),
+                    expected=(998246378.8556538+1112032928.103645j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.5,
+                    b=-3.5,
+                    c=2.050308316530781,
+                    z=(0.03793103448275881+1.024137931034483j),
+                    expected=(0.5527670397711952+2.697662715303637j),
+                    rtol=1.2e-15,       # rtol bumped from 1e-15 in gh18414
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-15.5,
+                    b=-1.5,
+                    c=-0.9629749245209605,
+                    z=(0.4931034482758623-0.8724137931034484j),
+                    expected=(55.396931662136886+968.467463806326j),
+                    rtol=5e-14,
+                ),
+            ),
+        ]
+    )
+    def test_region5(self, hyp2f1_test_case):
+        """1 < |z| < 1.1 and |1 - z| >= 0.9 and real(z) >= 0"""
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert 1 < abs(z) < 1.1 and abs(1 - z) >= 0.9 and z.real >= 0
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.parametrize(
+        "hyp2f1_test_case",
+        [
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.095813935368371,
+                    b=4.0013768449590685,
+                    c=4.078873014294075,
+                    z=(-0.9473684210526316+0.5263157894736841j),
+                    expected=(-0.0018093573941378783+0.003481887377423739j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.087593263474208,
+                    b=2.050308316530781,
+                    c=1.0651378143226575,
+                    z=(-0.736842105263158-0.736842105263158j),
+                    expected=(-0.00023401243818780545-1.7983496305603562e-05j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0272592605282642,
+                    b=8.077282662161238,
+                    c=4.078873014294075,
+                    z=(-0.5263157894736843-0.9473684210526316j),
+                    expected=(0.22359773002226846-0.24092487123993353j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0272592605282642,
+                    b=2.050308316530781,
+                    c=-15.963511401609862,
+                    z=(-0.9473684210526316-0.5263157894736843j),
+                    expected=(1.191573745740011+0.14347394589721466j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=4.0013768449590685,
+                    c=-15.963511401609862,
+                    z=(-0.9473684210526316-0.5263157894736843j),
+                    expected=(31.822620756901784-66.09094396747611j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=8.077282662161238,
+                    c=-7.93846038215665,
+                    z=(-0.9473684210526316+0.5263157894736841j),
+                    expected=(207.16750179245952+34.80478274924269j),
+                    rtol=5e-12,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=8.095813935368371,
+                    b=-7.949900487447654,
+                    c=8.031683612216888,
+                    z=(-0.736842105263158+0.7368421052631575j),
+                    expected=(-159.62429364277145+9.154224290644898j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0272592605282642,
+                    b=-1.92872979730171,
+                    c=16.056809865262608,
+                    z=(-0.9473684210526316+0.5263157894736841j),
+                    expected=(1.121122351247184-0.07170260470126685j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=16.087593263474208,
+                    b=-0.9629749245209605,
+                    c=16.056809865262608,
+                    z=(-0.9473684210526316+0.5263157894736841j),
+                    expected=(1.9040596681316053-0.4951799449960107j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0272592605282642,
+                    b=-1.92872979730171,
+                    c=-0.906685989801748,
+                    z=(-0.9473684210526316-0.5263157894736843j),
+                    expected=(-14.496623497780739-21.897524523299875j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=4.080187217753502,
+                    b=-3.9316537064827854,
+                    c=-3.9924618758357022,
+                    z=(-0.5263157894736843-0.9473684210526316j),
+                    expected=(36.33473466026878+253.88728442029577j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=1.0272592605282642,
+                    b=-15.964218273004214,
+                    c=-0.906685989801748,
+                    z=(-0.9473684210526316+0.5263157894736841j),
+                    expected=(1505052.5653144997-50820766.81043443j),
+                    rtol=1e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.956227226099288,
+                    b=4.0013768449590685,
+                    c=1.0651378143226575,
+                    z=(-0.5263157894736843+0.9473684210526314j),
+                    expected=(-127.79407519260877-28.69899444941112j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=8.077282662161238,
+                    c=16.056809865262608,
+                    z=(-0.9473684210526316-0.5263157894736843j),
+                    expected=(2.0623331933754976+0.741234463565458j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.956227226099288,
+                    b=8.077282662161238,
+                    c=2.0397202577726152,
+                    z=(-0.9473684210526316+0.5263157894736841j),
+                    expected=(30.729193458862525-292.5700835046965j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=1.0561196186065624,
+                    c=-1.9631175993998025,
+                    z=(-0.5263157894736843-0.9473684210526316j),
+                    expected=(1.1285917906203495-0.735264575450189j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=1.0561196186065624,
+                    c=-3.9924618758357022,
+                    z=(-0.736842105263158+0.7368421052631575j),
+                    expected=(0.6356474446678052-0.02429663008952248j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-1.9214641416286231,
+                    b=16.088264119063613,
+                    c=-7.93846038215665,
+                    z=(-0.736842105263158+0.7368421052631575j),
+                    expected=(0.4718880510273174+0.655083067736377j),
+                    rtol=1e-11,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-7.937789122896016,
+                    b=-3.9316537064827854,
+                    c=16.056809865262608,
+                    z=(-0.9473684210526316+0.5263157894736841j),
+                    expected=(-0.14681550942352714+0.16092206364265146j),
+                    rtol=5e-11,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=-15.964218273004214,
+                    c=1.0651378143226575,
+                    z=(-0.5263157894736843+0.9473684210526314j),
+                    expected=(-6.436835190526225+22.883156700606182j),
+                    rtol=5e-14,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-0.9220024191881196,
+                    b=-7.949900487447654,
+                    c=4.078873014294075,
+                    z=(-0.9473684210526316-0.5263157894736843j),
+                    expected=(-0.7505682955068583-1.1026583264249945j),
+                    rtol=1e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.956227226099288,
+                    b=-3.9316537064827854,
+                    c=-7.93846038215665,
+                    z=(-0.9473684210526316-0.5263157894736843j),
+                    expected=(3.6247814989198166+2.596041360148318j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.956227226099288,
+                    b=-15.964218273004214,
+                    c=-1.9631175993998025,
+                    z=(-0.5263157894736843-0.9473684210526316j),
+                    expected=(-59537.65287927933-669074.4342539902j),
+                    rtol=5e-15,
+                ),
+            ),
+            pytest.param(
+                Hyp2f1TestCase(
+                    a=-3.956227226099288,
+                    b=-15.964218273004214,
+                    c=-1.9631175993998025,
+                    z=(-0.9473684210526316-0.5263157894736843j),
+                    expected=(-433084.9970266166+431088.393918521j),
+                    rtol=5e-14,
+                ),
+            ),
+        ]
+    )
+    def test_region6(self, hyp2f1_test_case):
+        """|z| > 1 but not in region 5."""
+        a, b, c, z, expected, rtol = hyp2f1_test_case
+        assert (
+            abs(z) > 1 and
+            not (1 < abs(z) < 1.1 and abs(1 - z) >= 0.9 and z.real >= 0)
+        )
+        assert_allclose(hyp2f1(a, b, c, z), expected, rtol=rtol)
+
+    @pytest.mark.slow
+    @check_version(mpmath, "1.0.0")
+    def test_test_hyp2f1(self):
+        """Test that expected values match what is computed by mpmath.
+
+        This gathers the parameters for the test cases out of the pytest marks.
+        The parameters are a, b, c, z, expected, rtol, where expected should
+        be the value of hyp2f1(a, b, c, z) computed with mpmath. The test
+        recomputes hyp2f1(a, b, c, z) using mpmath and verifies that expected
+        actually is the correct value. This allows the data for the tests to
+        live within the test code instead of an external datafile, while
+        avoiding having to compute the results with mpmath during the test,
+        except for when slow tests are being run.
+        """
+        test_methods = [
+            test_method for test_method in dir(self)
+            if test_method.startswith('test') and
+            # Filter properties and attributes (futureproofing).
+            callable(getattr(self, test_method)) and
+            # Filter out this test
+            test_method != 'test_test_hyp2f1'
+        ]
+        for test_method in test_methods:
+            params = self._get_test_parameters(getattr(self, test_method))
+            for a, b, c, z, expected, _ in params:
+                assert_allclose(mp_hyp2f1(a, b, c, z), expected, rtol=2.25e-16)
+
+    def _get_test_parameters(self, test_method):
+        """Get pytest.mark parameters for a test in this class."""
+        return [
+            case.values[0] for mark in test_method.pytestmark
+            if mark.name == 'parametrize'
+            for case in mark.args[1]
+        ]

.venv/Lib/site-packages/scipy/special/tests/test_hypergeometric.py

@@ -0,0 +1,140 @@
+import pytest
+import numpy as np
+from numpy.testing import assert_allclose, assert_equal
+import scipy.special as sc
+
+
+class TestHyperu:
+
+    def test_negative_x(self):
+        a, b, x = np.meshgrid(
+            [-1, -0.5, 0, 0.5, 1],
+            [-1, -0.5, 0, 0.5, 1],
+            np.linspace(-100, -1, 10),
+        )
+        assert np.all(np.isnan(sc.hyperu(a, b, x)))
+
+    def test_special_cases(self):
+        assert sc.hyperu(0, 1, 1) == 1.0
+
+    @pytest.mark.parametrize('a', [0.5, 1, np.nan])
+    @pytest.mark.parametrize('b', [1, 2, np.nan])
+    @pytest.mark.parametrize('x', [0.25, 3, np.nan])
+    def test_nan_inputs(self, a, b, x):
+        assert np.isnan(sc.hyperu(a, b, x)) == np.any(np.isnan([a, b, x]))
+
+
+class TestHyp1f1:
+
+    @pytest.mark.parametrize('a, b, x', [
+        (np.nan, 1, 1),
+        (1, np.nan, 1),
+        (1, 1, np.nan)
+    ])
+    def test_nan_inputs(self, a, b, x):
+        assert np.isnan(sc.hyp1f1(a, b, x))
+
+    def test_poles(self):
+        assert_equal(sc.hyp1f1(1, [0, -1, -2, -3, -4], 0.5), np.inf)
+
+    @pytest.mark.parametrize('a, b, x, result', [
+        (-1, 1, 0.5, 0.5),
+        (1, 1, 0.5, 1.6487212707001281468),
+        (2, 1, 0.5, 2.4730819060501922203),
+        (1, 2, 0.5, 1.2974425414002562937),
+        (-10, 1, 0.5, -0.38937441413785204475)
+    ])
+    def test_special_cases(self, a, b, x, result):
+        # Hit all the special case branches at the beginning of the
+        # function. Desired answers computed using Mpmath.
+        assert_allclose(sc.hyp1f1(a, b, x), result, atol=0, rtol=1e-15)
+
+    @pytest.mark.parametrize('a, b, x, result', [
+        (1, 1, 0.44, 1.5527072185113360455),
+        (-1, 1, 0.44, 0.55999999999999999778),
+        (100, 100, 0.89, 2.4351296512898745592),
+        (-100, 100, 0.89, 0.40739062490768104667),
+        (1.5, 100, 59.99, 3.8073513625965598107),
+        (-1.5, 100, 59.99, 0.25099240047125826943)
+    ])
+    def test_geometric_convergence(self, a, b, x, result):
+        # Test the region where we are relying on the ratio of
+        #
+        # (|a| + 1) * |x| / |b|
+        #
+        # being small. Desired answers computed using Mpmath
+        assert_allclose(sc.hyp1f1(a, b, x), result, atol=0, rtol=1e-15)
+
+    @pytest.mark.parametrize('a, b, x, result', [
+        (-1, 1, 1.5, -0.5),
+        (-10, 1, 1.5, 0.41801777430943080357),
+        (-25, 1, 1.5, 0.25114491646037839809),
+        (-50, 1, 1.5, -0.25683643975194756115),
+        (-80, 1, 1.5, -0.24554329325751503601),
+        (-150, 1, 1.5, -0.173364795515420454496),
+    ])
+    def test_a_negative_integer(self, a, b, x, result):
+        # Desired answers computed using Mpmath.
+        assert_allclose(sc.hyp1f1(a, b, x), result, atol=0, rtol=2e-14)
+
+    @pytest.mark.parametrize('a, b, x, expected', [
+        (0.01, 150, -4, 0.99973683897677527773),        # gh-3492
+        (1, 5, 0.01, 1.0020033381011970966),            # gh-3593
+        (50, 100, 0.01, 1.0050126452421463411),         # gh-3593
+        (1, 0.3, -1e3, -7.011932249442947651455e-04),   # gh-14149
+        (1, 0.3, -1e4, -7.001190321418937164734e-05),   # gh-14149
+        (9, 8.5, -350, -5.224090831922378361082e-20),   # gh-17120
+        (9, 8.5, -355, -4.595407159813368193322e-20),   # gh-17120
+        (75, -123.5, 15, 3.425753920814889017493e+06),
+    ])
+    def test_assorted_cases(self, a, b, x, expected):
+        # Expected values were computed with mpmath.hyp1f1(a, b, x).
+        assert_allclose(sc.hyp1f1(a, b, x), expected, atol=0, rtol=1e-14)
+
+    def test_a_neg_int_and_b_equal_x(self):
+        # This is a case where the Boost wrapper will call hypergeometric_pFq
+        # instead of hypergeometric_1F1.  When we use a version of Boost in
+        # which https://github.com/boostorg/math/issues/833 is fixed, this
+        # test case can probably be moved into test_assorted_cases.
+        # The expected value was computed with mpmath.hyp1f1(a, b, x).
+        a = -10.0
+        b = 2.5
+        x = 2.5
+        expected = 0.0365323664364104338721
+        computed = sc.hyp1f1(a, b, x)
+        assert_allclose(computed, expected, atol=0, rtol=1e-13)
+
+    @pytest.mark.parametrize('a, b, x, desired', [
+        (-1, -2, 2, 2),
+        (-1, -4, 10, 3.5),
+        (-2, -2, 1, 2.5)
+    ])
+    def test_gh_11099(self, a, b, x, desired):
+        # All desired results computed using Mpmath
+        assert sc.hyp1f1(a, b, x) == desired
+
+    @pytest.mark.parametrize('a', [-3, -2])
+    def test_x_zero_a_and_b_neg_ints_and_a_ge_b(self, a):
+        assert sc.hyp1f1(a, -3, 0) == 1
+
+    # The "legacy edge cases" mentioned in the comments in the following
+    # tests refers to the behavior of hyp1f1(a, b, x) when b is a nonpositive
+    # integer.  In some subcases, the behavior of SciPy does not match that
+    # of Boost (1.81+), mpmath and Mathematica (via Wolfram Alpha online).
+    # If the handling of these edges cases is changed to agree with those
+    # libraries, these test will have to be updated.
+
+    @pytest.mark.parametrize('b', [0, -1, -5])
+    def test_legacy_case1(self, b):
+        # Test results of hyp1f1(0, n, x) for n <= 0.
+        # This is a legacy edge case.
+        # Boost (versions greater than 1.80), Mathematica (via Wolfram Alpha
+        # online) and mpmath all return 1 in this case, but SciPy's hyp1f1
+        # returns inf.
+        assert_equal(sc.hyp1f1(0, b, [-1.5, 0, 1.5]), [np.inf, np.inf, np.inf])
+
+    def test_legacy_case2(self):
+        # This is a legacy edge case.
+        # In software such as boost (1.81+), mpmath and Mathematica,
+        # the value is 1.
+        assert sc.hyp1f1(-4, -3, 0) == np.inf

.venv/Lib/site-packages/scipy/special/tests/test_kolmogorov.py

@@ -0,0 +1,495 @@
+import itertools
+import sys
+import pytest
+
+import numpy as np
+from numpy.testing import assert_
+from scipy.special._testutils import FuncData
+
+from scipy.special import kolmogorov, kolmogi, smirnov, smirnovi
+from scipy.special._ufuncs import (_kolmogc, _kolmogci, _kolmogp,
+                                   _smirnovc, _smirnovci, _smirnovp)
+
+_rtol = 1e-10
+
+class TestSmirnov:
+    def test_nan(self):
+        assert_(np.isnan(smirnov(1, np.nan)))
+
+    def test_basic(self):
+        dataset = [(1, 0.1, 0.9),
+                   (1, 0.875, 0.125),
+                   (2, 0.875, 0.125 * 0.125),
+                   (3, 0.875, 0.125 * 0.125 * 0.125)]
+
+        dataset = np.asarray(dataset)
+        FuncData(
+            smirnov, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, -1] = 1 - dataset[:, -1]
+        FuncData(
+            _smirnovc, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_x_equals_0(self):
+        dataset = [(n, 0, 1) for n in itertools.chain(range(2, 20), range(1010, 1020))]
+        dataset = np.asarray(dataset)
+        FuncData(
+            smirnov, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, -1] = 1 - dataset[:, -1]
+        FuncData(
+            _smirnovc, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_x_equals_1(self):
+        dataset = [(n, 1, 0) for n in itertools.chain(range(2, 20), range(1010, 1020))]
+        dataset = np.asarray(dataset)
+        FuncData(
+            smirnov, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, -1] = 1 - dataset[:, -1]
+        FuncData(
+            _smirnovc, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_x_equals_0point5(self):
+        dataset = [(1, 0.5, 0.5),
+                   (2, 0.5, 0.25),
+                   (3, 0.5, 0.166666666667),
+                   (4, 0.5, 0.09375),
+                   (5, 0.5, 0.056),
+                   (6, 0.5, 0.0327932098765),
+                   (7, 0.5, 0.0191958707681),
+                   (8, 0.5, 0.0112953186035),
+                   (9, 0.5, 0.00661933257355),
+                   (10, 0.5, 0.003888705)]
+
+        dataset = np.asarray(dataset)
+        FuncData(
+            smirnov, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, -1] = 1 - dataset[:, -1]
+        FuncData(
+            _smirnovc, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_n_equals_1(self):
+        x = np.linspace(0, 1, 101, endpoint=True)
+        dataset = np.column_stack([[1]*len(x), x, 1-x])
+        FuncData(
+            smirnov, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, -1] = 1 - dataset[:, -1]
+        FuncData(
+            _smirnovc, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_n_equals_2(self):
+        x = np.linspace(0.5, 1, 101, endpoint=True)
+        p = np.power(1-x, 2)
+        n = np.array([2] * len(x))
+        dataset = np.column_stack([n, x, p])
+        FuncData(
+            smirnov, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, -1] = 1 - dataset[:, -1]
+        FuncData(
+            _smirnovc, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_n_equals_3(self):
+        x = np.linspace(0.7, 1, 31, endpoint=True)
+        p = np.power(1-x, 3)
+        n = np.array([3] * len(x))
+        dataset = np.column_stack([n, x, p])
+        FuncData(
+            smirnov, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, -1] = 1 - dataset[:, -1]
+        FuncData(
+            _smirnovc, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_n_large(self):
+        # test for large values of n
+        # Probabilities should go down as n goes up
+        x = 0.4
+        pvals = np.array([smirnov(n, x) for n in range(400, 1100, 20)])
+        dfs = np.diff(pvals)
+        assert_(np.all(dfs <= 0), msg='Not all diffs negative %s' % dfs)
+
+
+class TestSmirnovi:
+    def test_nan(self):
+        assert_(np.isnan(smirnovi(1, np.nan)))
+
+    def test_basic(self):
+        dataset = [(1, 0.4, 0.6),
+                   (1, 0.6, 0.4),
+                   (1, 0.99, 0.01),
+                   (1, 0.01, 0.99),
+                   (2, 0.125 * 0.125, 0.875),
+                   (3, 0.125 * 0.125 * 0.125, 0.875),
+                   (10, 1.0 / 16 ** 10, 1 - 1.0 / 16)]
+
+        dataset = np.asarray(dataset)
+        FuncData(
+            smirnovi, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, 1] = 1 - dataset[:, 1]
+        FuncData(
+            _smirnovci, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_x_equals_0(self):
+        dataset = [(n, 0, 1) for n in itertools.chain(range(2, 20), range(1010, 1020))]
+        dataset = np.asarray(dataset)
+        FuncData(
+            smirnovi, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, 1] = 1 - dataset[:, 1]
+        FuncData(
+            _smirnovci, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_x_equals_1(self):
+        dataset = [(n, 1, 0) for n in itertools.chain(range(2, 20), range(1010, 1020))]
+        dataset = np.asarray(dataset)
+        FuncData(
+            smirnovi, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, 1] = 1 - dataset[:, 1]
+        FuncData(
+            _smirnovci, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_n_equals_1(self):
+        pp = np.linspace(0, 1, 101, endpoint=True)
+        # dataset = np.array([(1, p, 1-p) for p in pp])
+        dataset = np.column_stack([[1]*len(pp), pp, 1-pp])
+        FuncData(
+            smirnovi, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, 1] = 1 - dataset[:, 1]
+        FuncData(
+            _smirnovci, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_n_equals_2(self):
+        x = np.linspace(0.5, 1, 101, endpoint=True)
+        p = np.power(1-x, 2)
+        n = np.array([2] * len(x))
+        dataset = np.column_stack([n, p, x])
+        FuncData(
+            smirnovi, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, 1] = 1 - dataset[:, 1]
+        FuncData(
+            _smirnovci, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_n_equals_3(self):
+        x = np.linspace(0.7, 1, 31, endpoint=True)
+        p = np.power(1-x, 3)
+        n = np.array([3] * len(x))
+        dataset = np.column_stack([n, p, x])
+        FuncData(
+            smirnovi, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, 1] = 1 - dataset[:, 1]
+        FuncData(
+            _smirnovci, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_round_trip(self):
+        def _sm_smi(n, p):
+            return smirnov(n, smirnovi(n, p))
+
+        def _smc_smci(n, p):
+            return _smirnovc(n, _smirnovci(n, p))
+
+        dataset = [(1, 0.4, 0.4),
+                   (1, 0.6, 0.6),
+                   (2, 0.875, 0.875),
+                   (3, 0.875, 0.875),
+                   (3, 0.125, 0.125),
+                   (10, 0.999, 0.999),
+                   (10, 0.0001, 0.0001)]
+
+        dataset = np.asarray(dataset)
+        FuncData(
+            _sm_smi, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        FuncData(
+            _smc_smci, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_x_equals_0point5(self):
+        dataset = [(1, 0.5, 0.5),
+                   (2, 0.5, 0.366025403784),
+                   (2, 0.25, 0.5),
+                   (3, 0.5, 0.297156508177),
+                   (4, 0.5, 0.255520481121),
+                   (5, 0.5, 0.234559536069),
+                   (6, 0.5, 0.21715965898),
+                   (7, 0.5, 0.202722580034),
+                   (8, 0.5, 0.190621765256),
+                   (9, 0.5, 0.180363501362),
+                   (10, 0.5, 0.17157867006)]
+
+        dataset = np.asarray(dataset)
+        FuncData(
+            smirnovi, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+        dataset[:, 1] = 1 - dataset[:, 1]
+        FuncData(
+            _smirnovci, dataset, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+
+class TestSmirnovp:
+    def test_nan(self):
+        assert_(np.isnan(_smirnovp(1, np.nan)))
+
+    def test_basic(self):
+        # Check derivative at endpoints
+        n1_10 = np.arange(1, 10)
+        dataset0 = np.column_stack([n1_10,
+                                    np.full_like(n1_10, 0),
+                                    np.full_like(n1_10, -1)])
+        FuncData(
+            _smirnovp, dataset0, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+        n2_10 = np.arange(2, 10)
+        dataset1 = np.column_stack([n2_10,
+                                    np.full_like(n2_10, 1.0),
+                                    np.full_like(n2_10, 0)])
+        FuncData(
+            _smirnovp, dataset1, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_oneminusoneovern(self):
+        # Check derivative at x=1-1/n
+        n = np.arange(1, 20)
+        x = 1.0/n
+        xm1 = 1-1.0/n
+        pp1 = -n * x**(n-1)
+        pp1 -= (1-np.sign(n-2)**2) * 0.5  # n=2, x=0.5, 1-1/n = 0.5, need to adjust
+        dataset1 = np.column_stack([n, xm1, pp1])
+        FuncData(
+            _smirnovp, dataset1, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_oneovertwon(self):
+        # Check derivative at x=1/2n  (Discontinuous at x=1/n, so check at x=1/2n)
+        n = np.arange(1, 20)
+        x = 1.0/2/n
+        pp = -(n*x+1) * (1+x)**(n-2)
+        dataset0 = np.column_stack([n, x, pp])
+        FuncData(
+            _smirnovp, dataset0, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    def test_oneovern(self):
+        # Check derivative at x=1/n
+        # (Discontinuous at x=1/n, hard to tell if x==1/n, only use n=power of 2)
+        n = 2**np.arange(1, 10)
+        x = 1.0/n
+        pp = -(n*x+1) * (1+x)**(n-2) + 0.5
+        dataset0 = np.column_stack([n, x, pp])
+        FuncData(
+            _smirnovp, dataset0, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+    @pytest.mark.xfail(sys.maxsize <= 2**32,
+                       reason="requires 64-bit platform")
+    def test_oneovernclose(self):
+        # Check derivative at x=1/n
+        # (Discontinuous at x=1/n, test on either side: x=1/n +/- 2epsilon)
+        n = np.arange(3, 20)
+
+        x = 1.0/n - 2*np.finfo(float).eps
+        pp = -(n*x+1) * (1+x)**(n-2)
+        dataset0 = np.column_stack([n, x, pp])
+        FuncData(
+            _smirnovp, dataset0, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+        x = 1.0/n + 2*np.finfo(float).eps
+        pp = -(n*x+1) * (1+x)**(n-2) + 1
+        dataset1 = np.column_stack([n, x, pp])
+        FuncData(
+            _smirnovp, dataset1, (0, 1), 2, rtol=_rtol
+        ).check(dtypes=[int, float, float])
+
+
+class TestKolmogorov:
+    def test_nan(self):
+        assert_(np.isnan(kolmogorov(np.nan)))
+
+    def test_basic(self):
+        dataset = [(0, 1.0),
+                   (0.5, 0.96394524366487511),
+                   (0.8275735551899077, 0.5000000000000000),
+                   (1, 0.26999967167735456),
+                   (2, 0.00067092525577969533)]
+
+        dataset = np.asarray(dataset)
+        FuncData(kolmogorov, dataset, (0,), 1, rtol=_rtol).check()
+
+    def test_linspace(self):
+        x = np.linspace(0, 2.0, 21)
+        dataset = [1.0000000000000000, 1.0000000000000000, 0.9999999999994950,
+                   0.9999906941986655, 0.9971923267772983, 0.9639452436648751,
+                   0.8642827790506042, 0.7112351950296890, 0.5441424115741981,
+                   0.3927307079406543, 0.2699996716773546, 0.1777181926064012,
+                   0.1122496666707249, 0.0680922218447664, 0.0396818795381144,
+                   0.0222179626165251, 0.0119520432391966, 0.0061774306344441,
+                   0.0030676213475797, 0.0014636048371873, 0.0006709252557797]
+
+        dataset_c = [0.0000000000000000, 6.609305242245699e-53, 5.050407338670114e-13,
+                     9.305801334566668e-06, 0.0028076732227017, 0.0360547563351249,
+                     0.1357172209493958, 0.2887648049703110, 0.4558575884258019,
+                     0.6072692920593457, 0.7300003283226455, 0.8222818073935988,
+                     0.8877503333292751, 0.9319077781552336, 0.9603181204618857,
+                     0.9777820373834749, 0.9880479567608034, 0.9938225693655559,
+                     0.9969323786524203, 0.9985363951628127, 0.9993290747442203]
+
+        dataset = np.column_stack([x, dataset])
+        FuncData(kolmogorov, dataset, (0,), 1, rtol=_rtol).check()
+        dataset_c = np.column_stack([x, dataset_c])
+        FuncData(_kolmogc, dataset_c, (0,), 1, rtol=_rtol).check()
+
+    def test_linspacei(self):
+        p = np.linspace(0, 1.0, 21, endpoint=True)
+        dataset = [np.inf, 1.3580986393225507, 1.2238478702170823,
+                   1.1379465424937751, 1.0727491749396481, 1.0191847202536859,
+                   0.9730633753323726, 0.9320695842357622, 0.8947644549851197,
+                   0.8601710725555463, 0.8275735551899077, 0.7964065373291559,
+                   0.7661855555617682, 0.7364542888171910, 0.7067326523068980,
+                   0.6764476915028201, 0.6448126061663567, 0.6105590999244391,
+                   0.5711732651063401, 0.5196103791686224, 0.0000000000000000]
+
+        dataset_c = [0.0000000000000000, 0.5196103791686225, 0.5711732651063401,
+                     0.6105590999244391, 0.6448126061663567, 0.6764476915028201,
+                     0.7067326523068980, 0.7364542888171910, 0.7661855555617682,
+                     0.7964065373291559, 0.8275735551899077, 0.8601710725555463,
+                     0.8947644549851196, 0.9320695842357622, 0.9730633753323727,
+                     1.0191847202536859, 1.0727491749396481, 1.1379465424937754,
+                     1.2238478702170825, 1.3580986393225509, np.inf]
+
+        dataset = np.column_stack([p[1:], dataset[1:]])
+        FuncData(kolmogi, dataset, (0,), 1, rtol=_rtol).check()
+        dataset_c = np.column_stack([p[:-1], dataset_c[:-1]])
+        FuncData(_kolmogci, dataset_c, (0,), 1, rtol=_rtol).check()
+
+    def test_smallx(self):
+        epsilon = 0.1 ** np.arange(1, 14)
+        x = np.array([0.571173265106, 0.441027698518, 0.374219690278, 0.331392659217,
+                      0.300820537459, 0.277539353999, 0.259023494805, 0.243829561254,
+                      0.231063086389, 0.220135543236, 0.210641372041, 0.202290283658,
+                      0.19487060742])
+
+        dataset = np.column_stack([x, 1-epsilon])
+        FuncData(kolmogorov, dataset, (0,), 1, rtol=_rtol).check()
+
+    def test_round_trip(self):
+        def _ki_k(_x):
+            return kolmogi(kolmogorov(_x))
+
+        def _kci_kc(_x):
+            return _kolmogci(_kolmogc(_x))
+
+        x = np.linspace(0.0, 2.0, 21, endpoint=True)
+        # Exclude 0.1, 0.2.  0.2 almost makes succeeds, but 0.1 has no chance.
+        x02 = x[(x == 0) | (x > 0.21)]
+        dataset02 = np.column_stack([x02, x02])
+        FuncData(_ki_k, dataset02, (0,), 1, rtol=_rtol).check()
+
+        dataset = np.column_stack([x, x])
+        FuncData(_kci_kc, dataset, (0,), 1, rtol=_rtol).check()
+
+
+class TestKolmogi:
+    def test_nan(self):
+        assert_(np.isnan(kolmogi(np.nan)))
+
+    def test_basic(self):
+        dataset = [(1.0, 0),
+                   (0.96394524366487511, 0.5),
+                   (0.9, 0.571173265106),
+                   (0.5000000000000000, 0.8275735551899077),
+                   (0.26999967167735456, 1),
+                   (0.00067092525577969533, 2)]
+
+        dataset = np.asarray(dataset)
+        FuncData(kolmogi, dataset, (0,), 1, rtol=_rtol).check()
+
+    def test_smallpcdf(self):
+        epsilon = 0.5 ** np.arange(1, 55, 3)
+        # kolmogi(1-p) == _kolmogci(p) if  1-(1-p) == p, but not necessarily otherwise
+        # Use epsilon s.t. 1-(1-epsilon)) == epsilon,
+        # so can use same x-array for both results
+
+        x = np.array([0.8275735551899077, 0.5345255069097583, 0.4320114038786941,
+                      0.3736868442620478, 0.3345161714909591, 0.3057833329315859,
+                      0.2835052890528936, 0.2655578150208676, 0.2506869966107999,
+                      0.2380971058736669, 0.2272549289962079, 0.2177876361600040,
+                      0.2094254686862041, 0.2019676748836232, 0.1952612948137504,
+                      0.1891874239646641, 0.1836520225050326, 0.1785795904846466])
+
+        dataset = np.column_stack([1-epsilon, x])
+        FuncData(kolmogi, dataset, (0,), 1, rtol=_rtol).check()
+
+        dataset = np.column_stack([epsilon, x])
+        FuncData(_kolmogci, dataset, (0,), 1, rtol=_rtol).check()
+
+    def test_smallpsf(self):
+        epsilon = 0.5 ** np.arange(1, 55, 3)
+        # kolmogi(p) == _kolmogci(1-p) if  1-(1-p) == p, but not necessarily otherwise
+        # Use epsilon s.t. 1-(1-epsilon)) == epsilon,
+        # so can use same x-array for both results
+
+        x = np.array([0.8275735551899077, 1.3163786275161036, 1.6651092133663343,
+                      1.9525136345289607, 2.2027324540033235, 2.4272929437460848,
+                      2.6327688477341593, 2.8233300509220260, 3.0018183401530627,
+                      3.1702735084088891, 3.3302184446307912, 3.4828258153113318,
+                      3.6290214150152051, 3.7695513262825959, 3.9050272690877326,
+                      4.0359582187082550, 4.1627730557884890, 4.2858371743264527])
+
+        dataset = np.column_stack([epsilon, x])
+        FuncData(kolmogi, dataset, (0,), 1, rtol=_rtol).check()
+
+        dataset = np.column_stack([1-epsilon, x])
+        FuncData(_kolmogci, dataset, (0,), 1, rtol=_rtol).check()
+
+    def test_round_trip(self):
+        def _k_ki(_p):
+            return kolmogorov(kolmogi(_p))
+
+        p = np.linspace(0.1, 1.0, 10, endpoint=True)
+        dataset = np.column_stack([p, p])
+        FuncData(_k_ki, dataset, (0,), 1, rtol=_rtol).check()
+
+
+class TestKolmogp:
+    def test_nan(self):
+        assert_(np.isnan(_kolmogp(np.nan)))
+
+    def test_basic(self):
+        dataset = [(0.000000, -0.0),
+                   (0.200000, -1.532420541338916e-10),
+                   (0.400000, -0.1012254419260496),
+                   (0.600000, -1.324123244249925),
+                   (0.800000, -1.627024345636592),
+                   (1.000000, -1.071948558356941),
+                   (1.200000, -0.538512430720529),
+                   (1.400000, -0.2222133182429472),
+                   (1.600000, -0.07649302775520538),
+                   (1.800000, -0.02208687346347873),
+                   (2.000000, -0.005367402045629683)]
+
+        dataset = np.asarray(dataset)
+        FuncData(_kolmogp, dataset, (0,), 1, rtol=_rtol).check()

.venv/Lib/site-packages/scipy/special/tests/test_lambertw.py

@@ -0,0 +1,109 @@
+#
+# Tests for the lambertw function,
+# Adapted from the MPMath tests [1] by Yosef Meller, [email protected]
+# Distributed under the same license as SciPy itself.
+#
+# [1] mpmath source code, Subversion revision 992
+#     http://code.google.com/p/mpmath/source/browse/trunk/mpmath/tests/test_functions2.py?spec=svn994&r=992
+
+import pytest
+import numpy as np
+from numpy.testing import assert_, assert_equal, assert_array_almost_equal
+from scipy.special import lambertw
+from numpy import nan, inf, pi, e, isnan, log, r_, array, complex128
+
+from scipy.special._testutils import FuncData
+
+
+def test_values():
+    assert_(isnan(lambertw(nan)))
+    assert_equal(lambertw(inf,1).real, inf)
+    assert_equal(lambertw(inf,1).imag, 2*pi)
+    assert_equal(lambertw(-inf,1).real, inf)
+    assert_equal(lambertw(-inf,1).imag, 3*pi)
+
+    assert_equal(lambertw(1.), lambertw(1., 0))
+
+    data = [
+        (0,0, 0),
+        (0+0j,0, 0),
+        (inf,0, inf),
+        (0,-1, -inf),
+        (0,1, -inf),
+        (0,3, -inf),
+        (e,0, 1),
+        (1,0, 0.567143290409783873),
+        (-pi/2,0, 1j*pi/2),
+        (-log(2)/2,0, -log(2)),
+        (0.25,0, 0.203888354702240164),
+        (-0.25,0, -0.357402956181388903),
+        (-1./10000,0, -0.000100010001500266719),
+        (-0.25,-1, -2.15329236411034965),
+        (0.25,-1, -3.00899800997004620-4.07652978899159763j),
+        (-0.25,-1, -2.15329236411034965),
+        (0.25,1, -3.00899800997004620+4.07652978899159763j),
+        (-0.25,1, -3.48973228422959210+7.41405453009603664j),
+        (-4,0, 0.67881197132094523+1.91195078174339937j),
+        (-4,1, -0.66743107129800988+7.76827456802783084j),
+        (-4,-1, 0.67881197132094523-1.91195078174339937j),
+        (1000,0, 5.24960285240159623),
+        (1000,1, 4.91492239981054535+5.44652615979447070j),
+        (1000,-1, 4.91492239981054535-5.44652615979447070j),
+        (1000,5, 3.5010625305312892+29.9614548941181328j),
+        (3+4j,0, 1.281561806123775878+0.533095222020971071j),
+        (-0.4+0.4j,0, -0.10396515323290657+0.61899273315171632j),
+        (3+4j,1, -0.11691092896595324+5.61888039871282334j),
+        (3+4j,-1, 0.25856740686699742-3.85211668616143559j),
+        (-0.5,-1, -0.794023632344689368-0.770111750510379110j),
+        (-1./10000,1, -11.82350837248724344+6.80546081842002101j),
+        (-1./10000,-1, -11.6671145325663544),
+        (-1./10000,-2, -11.82350837248724344-6.80546081842002101j),
+        (-1./100000,4, -14.9186890769540539+26.1856750178782046j),
+        (-1./100000,5, -15.0931437726379218666+32.5525721210262290086j),
+        ((2+1j)/10,0, 0.173704503762911669+0.071781336752835511j),
+        ((2+1j)/10,1, -3.21746028349820063+4.56175438896292539j),
+        ((2+1j)/10,-1, -3.03781405002993088-3.53946629633505737j),
+        ((2+1j)/10,4, -4.6878509692773249+23.8313630697683291j),
+        (-(2+1j)/10,0, -0.226933772515757933-0.164986470020154580j),
+        (-(2+1j)/10,1, -2.43569517046110001+0.76974067544756289j),
+        (-(2+1j)/10,-1, -3.54858738151989450-6.91627921869943589j),
+        (-(2+1j)/10,4, -4.5500846928118151+20.6672982215434637j),
+        (pi,0, 1.073658194796149172092178407024821347547745350410314531),
+
+        # Former bug in generated branch,
+        (-0.5+0.002j,0, -0.78917138132659918344 + 0.76743539379990327749j),
+        (-0.5-0.002j,0, -0.78917138132659918344 - 0.76743539379990327749j),
+        (-0.448+0.4j,0, -0.11855133765652382241 + 0.66570534313583423116j),
+        (-0.448-0.4j,0, -0.11855133765652382241 - 0.66570534313583423116j),
+    ]
+    data = array(data, dtype=complex128)
+
+    def w(x, y):
+        return lambertw(x, y.real.astype(int))
+    with np.errstate(all='ignore'):
+        FuncData(w, data, (0,1), 2, rtol=1e-10, atol=1e-13).check()
+
+
+def test_ufunc():
+    assert_array_almost_equal(
+        lambertw(r_[0., e, 1.]), r_[0., 1., 0.567143290409783873])
+
+
+def test_lambertw_ufunc_loop_selection():
+    # see https://github.com/scipy/scipy/issues/4895
+    dt = np.dtype(np.complex128)
+    assert_equal(lambertw(0, 0, 0).dtype, dt)
+    assert_equal(lambertw([0], 0, 0).dtype, dt)
+    assert_equal(lambertw(0, [0], 0).dtype, dt)
+    assert_equal(lambertw(0, 0, [0]).dtype, dt)
+    assert_equal(lambertw([0], [0], [0]).dtype, dt)
+
+
+@pytest.mark.parametrize('z', [1e-316, -2e-320j, -5e-318+1e-320j])
+def test_lambertw_subnormal_k0(z):
+    # Verify that subnormal inputs are handled correctly on
+    # the branch k=0 (regression test for gh-16291).
+    w = lambertw(z)
+    # For values this small, we can be sure that numerically,
+    # lambertw(z) is z.
+    assert w == z

.venv/Lib/site-packages/scipy/special/tests/test_log_softmax.py

@@ -0,0 +1,109 @@
+import numpy as np
+from numpy.testing import assert_allclose
+
+import pytest
+
+import scipy.special as sc
+
+
+@pytest.mark.parametrize('x, expected', [
+    (np.array([1000, 1]), np.array([0, -999])),
+
+    # Expected value computed using mpmath (with mpmath.mp.dps = 200) and then
+    # converted to float.
+    (np.arange(4), np.array([-3.4401896985611953,
+                             -2.4401896985611953,
+                             -1.4401896985611953,
+                             -0.44018969856119533]))
+])
+def test_log_softmax(x, expected):
+    assert_allclose(sc.log_softmax(x), expected, rtol=1e-13)
+
+
+@pytest.fixture
+def log_softmax_x():
+    x = np.arange(4)
+    return x
+
+
+@pytest.fixture
+def log_softmax_expected():
+    # Expected value computed using mpmath (with mpmath.mp.dps = 200) and then
+    # converted to float.
+    expected = np.array([-3.4401896985611953,
+                         -2.4401896985611953,
+                         -1.4401896985611953,
+                         -0.44018969856119533])
+    return expected
+
+
+def test_log_softmax_translation(log_softmax_x, log_softmax_expected):
+    # Translation property.  If all the values are changed by the same amount,
+    # the softmax result does not change.
+    x = log_softmax_x + 100
+    expected = log_softmax_expected
+    assert_allclose(sc.log_softmax(x), expected, rtol=1e-13)
+
+
+def test_log_softmax_noneaxis(log_softmax_x, log_softmax_expected):
+    # When axis=None, softmax operates on the entire array, and preserves
+    # the shape.
+    x = log_softmax_x.reshape(2, 2)
+    expected = log_softmax_expected.reshape(2, 2)
+    assert_allclose(sc.log_softmax(x), expected, rtol=1e-13)
+
+
+@pytest.mark.parametrize('axis_2d, expected_2d', [
+    (0, np.log(0.5) * np.ones((2, 2))),
+    (1, np.array([[0, -999], [0, -999]]))
+])
+def test_axes(axis_2d, expected_2d):
+    assert_allclose(
+        sc.log_softmax([[1000, 1], [1000, 1]], axis=axis_2d),
+        expected_2d,
+        rtol=1e-13,
+    )
+
+
+@pytest.fixture
+def log_softmax_2d_x():
+    x = np.arange(8).reshape(2, 4)
+    return x
+
+
+@pytest.fixture
+def log_softmax_2d_expected():
+    # Expected value computed using mpmath (with mpmath.mp.dps = 200) and then
+    # converted to float.
+    expected = np.array([[-3.4401896985611953,
+                         -2.4401896985611953,
+                         -1.4401896985611953,
+                         -0.44018969856119533],
+                        [-3.4401896985611953,
+                         -2.4401896985611953,
+                         -1.4401896985611953,
+                         -0.44018969856119533]])
+    return expected
+
+
+def test_log_softmax_2d_axis1(log_softmax_2d_x, log_softmax_2d_expected):
+    x = log_softmax_2d_x
+    expected = log_softmax_2d_expected
+    assert_allclose(sc.log_softmax(x, axis=1), expected, rtol=1e-13)
+
+
+def test_log_softmax_2d_axis0(log_softmax_2d_x, log_softmax_2d_expected):
+    x = log_softmax_2d_x.T
+    expected = log_softmax_2d_expected.T
+    assert_allclose(sc.log_softmax(x, axis=0), expected, rtol=1e-13)
+
+
+def test_log_softmax_3d(log_softmax_2d_x, log_softmax_2d_expected):
+    # 3-d input, with a tuple for the axis.
+    x_3d = log_softmax_2d_x.reshape(2, 2, 2)
+    expected_3d = log_softmax_2d_expected.reshape(2, 2, 2)
+    assert_allclose(sc.log_softmax(x_3d, axis=(1, 2)), expected_3d, rtol=1e-13)
+
+
+def test_log_softmax_scalar():
+    assert_allclose(sc.log_softmax(1.0), 0.0, rtol=1e-13)

.venv/Lib/site-packages/scipy/special/tests/test_loggamma.py

@@ -0,0 +1,70 @@
+import numpy as np
+from numpy.testing import assert_allclose, assert_
+
+from scipy.special._testutils import FuncData
+from scipy.special import gamma, gammaln, loggamma
+
+
+def test_identities1():
+    # test the identity exp(loggamma(z)) = gamma(z)
+    x = np.array([-99.5, -9.5, -0.5, 0.5, 9.5, 99.5])
+    y = x.copy()
+    x, y = np.meshgrid(x, y)
+    z = (x + 1J*y).flatten()
+    dataset = np.vstack((z, gamma(z))).T
+
+    def f(z):
+        return np.exp(loggamma(z))
+
+    FuncData(f, dataset, 0, 1, rtol=1e-14, atol=1e-14).check()
+
+
+def test_identities2():
+    # test the identity loggamma(z + 1) = log(z) + loggamma(z)
+    x = np.array([-99.5, -9.5, -0.5, 0.5, 9.5, 99.5])
+    y = x.copy()
+    x, y = np.meshgrid(x, y)
+    z = (x + 1J*y).flatten()
+    dataset = np.vstack((z, np.log(z) + loggamma(z))).T
+
+    def f(z):
+        return loggamma(z + 1)
+
+    FuncData(f, dataset, 0, 1, rtol=1e-14, atol=1e-14).check()
+
+
+def test_complex_dispatch_realpart():
+    # Test that the real parts of loggamma and gammaln agree on the
+    # real axis.
+    x = np.r_[-np.logspace(10, -10), np.logspace(-10, 10)] + 0.5
+
+    dataset = np.vstack((x, gammaln(x))).T
+
+    def f(z):
+        z = np.array(z, dtype='complex128')
+        return loggamma(z).real
+
+    FuncData(f, dataset, 0, 1, rtol=1e-14, atol=1e-14).check()
+
+
+def test_real_dispatch():
+    x = np.logspace(-10, 10) + 0.5
+    dataset = np.vstack((x, gammaln(x))).T
+
+    FuncData(loggamma, dataset, 0, 1, rtol=1e-14, atol=1e-14).check()
+    assert_(loggamma(0) == np.inf)
+    assert_(np.isnan(loggamma(-1)))
+
+
+def test_gh_6536():
+    z = loggamma(complex(-3.4, +0.0))
+    zbar = loggamma(complex(-3.4, -0.0))
+    assert_allclose(z, zbar.conjugate(), rtol=1e-15, atol=0)
+
+
+def test_branch_cut():
+    # Make sure negative zero is treated correctly
+    x = -np.logspace(300, -30, 100)
+    z = np.asarray([complex(x0, 0.0) for x0 in x])
+    zbar = np.asarray([complex(x0, -0.0) for x0 in x])
+    assert_allclose(z, zbar.conjugate(), rtol=1e-15, atol=0)

.venv/Lib/site-packages/scipy/special/tests/test_logit.py

@@ -0,0 +1,145 @@
+import numpy as np
+from numpy.testing import (assert_equal, assert_almost_equal,
+                           assert_allclose)
+from scipy.special import logit, expit, log_expit
+
+
+class TestLogit:
+    def check_logit_out(self, dtype, expected):
+        a = np.linspace(0, 1, 10)
+        a = np.array(a, dtype=dtype)
+        with np.errstate(divide='ignore'):
+            actual = logit(a)
+
+        assert_almost_equal(actual, expected)
+
+        assert_equal(actual.dtype, np.dtype(dtype))
+
+    def test_float32(self):
+        expected = np.array([-np.inf, -2.07944155,
+                            -1.25276291, -0.69314718,
+                            -0.22314353, 0.22314365,
+                            0.6931473, 1.25276303,
+                            2.07944155, np.inf], dtype=np.float32)
+        self.check_logit_out('f4', expected)
+
+    def test_float64(self):
+        expected = np.array([-np.inf, -2.07944154,
+                            -1.25276297, -0.69314718,
+                            -0.22314355, 0.22314355,
+                            0.69314718, 1.25276297,
+                            2.07944154, np.inf])
+        self.check_logit_out('f8', expected)
+
+    def test_nan(self):
+        expected = np.array([np.nan]*4)
+        with np.errstate(invalid='ignore'):
+            actual = logit(np.array([-3., -2., 2., 3.]))
+
+        assert_equal(expected, actual)
+
+
+class TestExpit:
+    def check_expit_out(self, dtype, expected):
+        a = np.linspace(-4, 4, 10)
+        a = np.array(a, dtype=dtype)
+        actual = expit(a)
+        assert_almost_equal(actual, expected)
+        assert_equal(actual.dtype, np.dtype(dtype))
+
+    def test_float32(self):
+        expected = np.array([0.01798621, 0.04265125,
+                            0.09777259, 0.20860852,
+                            0.39068246, 0.60931754,
+                            0.79139149, 0.9022274,
+                            0.95734876, 0.98201376], dtype=np.float32)
+        self.check_expit_out('f4', expected)
+
+    def test_float64(self):
+        expected = np.array([0.01798621, 0.04265125,
+                            0.0977726, 0.20860853,
+                            0.39068246, 0.60931754,
+                            0.79139147, 0.9022274,
+                            0.95734875, 0.98201379])
+        self.check_expit_out('f8', expected)
+
+    def test_large(self):
+        for dtype in (np.float32, np.float64, np.longdouble):
+            for n in (88, 89, 709, 710, 11356, 11357):
+                n = np.array(n, dtype=dtype)
+                assert_allclose(expit(n), 1.0, atol=1e-20)
+                assert_allclose(expit(-n), 0.0, atol=1e-20)
+                assert_equal(expit(n).dtype, dtype)
+                assert_equal(expit(-n).dtype, dtype)
+
+
+class TestLogExpit:
+
+    def test_large_negative(self):
+        x = np.array([-10000.0, -750.0, -500.0, -35.0])
+        y = log_expit(x)
+        assert_equal(y, x)
+
+    def test_large_positive(self):
+        x = np.array([750.0, 1000.0, 10000.0])
+        y = log_expit(x)
+        # y will contain -0.0, and -0.0 is used in the expected value,
+        # but assert_equal does not check the sign of zeros, and I don't
+        # think the sign is an essential part of the test (i.e. it would
+        # probably be OK if log_expit(1000) returned 0.0 instead of -0.0).
+        assert_equal(y, np.array([-0.0, -0.0, -0.0]))
+
+    def test_basic_float64(self):
+        x = np.array([-32, -20, -10, -3, -1, -0.1, -1e-9,
+                      0, 1e-9, 0.1, 1, 10, 100, 500, 710, 725, 735])
+        y = log_expit(x)
+        #
+        # Expected values were computed with mpmath:
+        #
+        #   import mpmath
+        #
+        #   mpmath.mp.dps = 100
+        #
+        #   def mp_log_expit(x):
+        #       return -mpmath.log1p(mpmath.exp(-x))
+        #
+        #   expected = [float(mp_log_expit(t)) for t in x]
+        #
+        expected = [-32.000000000000014, -20.000000002061153,
+                    -10.000045398899218, -3.048587351573742,
+                    -1.3132616875182228, -0.7443966600735709,
+                    -0.6931471810599453, -0.6931471805599453,
+                    -0.6931471800599454, -0.6443966600735709,
+                    -0.3132616875182228, -4.539889921686465e-05,
+                    -3.720075976020836e-44, -7.124576406741286e-218,
+                    -4.47628622567513e-309, -1.36930634e-315,
+                    -6.217e-320]
+
+        # When tested locally, only one value in y was not exactly equal to
+        # expected.  That was for x=1, and the y value differed from the
+        # expected by 1 ULP.  For this test, however, I'll use rtol=1e-15.
+        assert_allclose(y, expected, rtol=1e-15)
+
+    def test_basic_float32(self):
+        x = np.array([-32, -20, -10, -3, -1, -0.1, -1e-9,
+                      0, 1e-9, 0.1, 1, 10, 100], dtype=np.float32)
+        y = log_expit(x)
+        #
+        # Expected values were computed with mpmath:
+        #
+        #   import mpmath
+        #
+        #   mpmath.mp.dps = 100
+        #
+        #   def mp_log_expit(x):
+        #       return -mpmath.log1p(mpmath.exp(-x))
+        #
+        #   expected = [np.float32(mp_log_expit(t)) for t in x]
+        #
+        expected = np.array([-32.0, -20.0, -10.000046, -3.0485873,
+                             -1.3132616, -0.7443967, -0.6931472,
+                             -0.6931472, -0.6931472, -0.64439666,
+                             -0.3132617, -4.5398898e-05, -3.8e-44],
+                            dtype=np.float32)
+
+        assert_allclose(y, expected, rtol=5e-7)

.venv/Lib/site-packages/scipy/special/tests/test_logsumexp.py

@@ -0,0 +1,207 @@
+import numpy as np
+from numpy.testing import (assert_almost_equal, assert_equal, assert_allclose,
+                           assert_array_almost_equal, assert_)
+
+from scipy.special import logsumexp, softmax
+
+
+def test_logsumexp():
+    # Test whether logsumexp() function correctly handles large inputs.
+    a = np.arange(200)
+    desired = np.log(np.sum(np.exp(a)))
+    assert_almost_equal(logsumexp(a), desired)
+
+    # Now test with large numbers
+    b = [1000, 1000]
+    desired = 1000.0 + np.log(2.0)
+    assert_almost_equal(logsumexp(b), desired)
+
+    n = 1000
+    b = np.full(n, 10000, dtype='float64')
+    desired = 10000.0 + np.log(n)
+    assert_almost_equal(logsumexp(b), desired)
+
+    x = np.array([1e-40] * 1000000)
+    logx = np.log(x)
+
+    X = np.vstack([x, x])
+    logX = np.vstack([logx, logx])
+    assert_array_almost_equal(np.exp(logsumexp(logX)), X.sum())
+    assert_array_almost_equal(np.exp(logsumexp(logX, axis=0)), X.sum(axis=0))
+    assert_array_almost_equal(np.exp(logsumexp(logX, axis=1)), X.sum(axis=1))
+
+    # Handling special values properly
+    assert_equal(logsumexp(np.inf), np.inf)
+    assert_equal(logsumexp(-np.inf), -np.inf)
+    assert_equal(logsumexp(np.nan), np.nan)
+    assert_equal(logsumexp([-np.inf, -np.inf]), -np.inf)
+
+    # Handling an array with different magnitudes on the axes
+    assert_array_almost_equal(logsumexp([[1e10, 1e-10],
+                                         [-1e10, -np.inf]], axis=-1),
+                              [1e10, -1e10])
+
+    # Test keeping dimensions
+    assert_array_almost_equal(logsumexp([[1e10, 1e-10],
+                                         [-1e10, -np.inf]],
+                                        axis=-1,
+                                        keepdims=True),
+                              [[1e10], [-1e10]])
+
+    # Test multiple axes
+    assert_array_almost_equal(logsumexp([[1e10, 1e-10],
+                                         [-1e10, -np.inf]],
+                                        axis=(-1,-2)),
+                              1e10)
+
+
+def test_logsumexp_b():
+    a = np.arange(200)
+    b = np.arange(200, 0, -1)
+    desired = np.log(np.sum(b*np.exp(a)))
+    assert_almost_equal(logsumexp(a, b=b), desired)
+
+    a = [1000, 1000]
+    b = [1.2, 1.2]
+    desired = 1000 + np.log(2 * 1.2)
+    assert_almost_equal(logsumexp(a, b=b), desired)
+
+    x = np.array([1e-40] * 100000)
+    b = np.linspace(1, 1000, 100000)
+    logx = np.log(x)
+
+    X = np.vstack((x, x))
+    logX = np.vstack((logx, logx))
+    B = np.vstack((b, b))
+    assert_array_almost_equal(np.exp(logsumexp(logX, b=B)), (B * X).sum())
+    assert_array_almost_equal(np.exp(logsumexp(logX, b=B, axis=0)),
+                                (B * X).sum(axis=0))
+    assert_array_almost_equal(np.exp(logsumexp(logX, b=B, axis=1)),
+                                (B * X).sum(axis=1))
+
+
+def test_logsumexp_sign():
+    a = [1,1,1]
+    b = [1,-1,-1]
+
+    r, s = logsumexp(a, b=b, return_sign=True)
+    assert_almost_equal(r,1)
+    assert_equal(s,-1)
+
+
+def test_logsumexp_sign_zero():
+    a = [1,1]
+    b = [1,-1]
+
+    r, s = logsumexp(a, b=b, return_sign=True)
+    assert_(not np.isfinite(r))
+    assert_(not np.isnan(r))
+    assert_(r < 0)
+    assert_equal(s,0)
+
+
+def test_logsumexp_sign_shape():
+    a = np.ones((1,2,3,4))
+    b = np.ones_like(a)
+
+    r, s = logsumexp(a, axis=2, b=b, return_sign=True)
+
+    assert_equal(r.shape, s.shape)
+    assert_equal(r.shape, (1,2,4))
+
+    r, s = logsumexp(a, axis=(1,3), b=b, return_sign=True)
+
+    assert_equal(r.shape, s.shape)
+    assert_equal(r.shape, (1,3))
+
+
+def test_logsumexp_complex_sign():
+    a = np.array([1 + 1j, 2 - 1j, -2 + 3j])
+
+    r, s = logsumexp(a, return_sign=True)
+
+    expected_sumexp = np.exp(a).sum()
+    # This is the numpy>=2.0 convention for np.sign
+    expected_sign = expected_sumexp / abs(expected_sumexp)
+
+    assert_allclose(s, expected_sign)
+    assert_allclose(s * np.exp(r), expected_sumexp)
+
+
+def test_logsumexp_shape():
+    a = np.ones((1, 2, 3, 4))
+    b = np.ones_like(a)
+
+    r = logsumexp(a, axis=2, b=b)
+    assert_equal(r.shape, (1, 2, 4))
+
+    r = logsumexp(a, axis=(1, 3), b=b)
+    assert_equal(r.shape, (1, 3))
+
+
+def test_logsumexp_b_zero():
+    a = [1,10000]
+    b = [1,0]
+
+    assert_almost_equal(logsumexp(a, b=b), 1)
+
+
+def test_logsumexp_b_shape():
+    a = np.zeros((4,1,2,1))
+    b = np.ones((3,1,5))
+
+    logsumexp(a, b=b)
+
+
+def test_softmax_fixtures():
+    assert_allclose(softmax([1000, 0, 0, 0]), np.array([1, 0, 0, 0]),
+                    rtol=1e-13)
+    assert_allclose(softmax([1, 1]), np.array([.5, .5]), rtol=1e-13)
+    assert_allclose(softmax([0, 1]), np.array([1, np.e])/(1 + np.e),
+                    rtol=1e-13)
+
+    # Expected value computed using mpmath (with mpmath.mp.dps = 200) and then
+    # converted to float.
+    x = np.arange(4)
+    expected = np.array([0.03205860328008499,
+                         0.08714431874203256,
+                         0.23688281808991013,
+                         0.6439142598879722])
+
+    assert_allclose(softmax(x), expected, rtol=1e-13)
+
+    # Translation property.  If all the values are changed by the same amount,
+    # the softmax result does not change.
+    assert_allclose(softmax(x + 100), expected, rtol=1e-13)
+
+    # When axis=None, softmax operates on the entire array, and preserves
+    # the shape.
+    assert_allclose(softmax(x.reshape(2, 2)), expected.reshape(2, 2),
+                    rtol=1e-13)
+
+
+def test_softmax_multi_axes():
+    assert_allclose(softmax([[1000, 0], [1000, 0]], axis=0),
+                    np.array([[.5, .5], [.5, .5]]), rtol=1e-13)
+    assert_allclose(softmax([[1000, 0], [1000, 0]], axis=1),
+                    np.array([[1, 0], [1, 0]]), rtol=1e-13)
+
+    # Expected value computed using mpmath (with mpmath.mp.dps = 200) and then
+    # converted to float.
+    x = np.array([[-25, 0, 25, 50],
+                  [1, 325, 749, 750]])
+    expected = np.array([[2.678636961770877e-33,
+                          1.9287498479371314e-22,
+                          1.3887943864771144e-11,
+                          0.999999999986112],
+                         [0.0,
+                          1.9444526359919372e-185,
+                          0.2689414213699951,
+                          0.7310585786300048]])
+    assert_allclose(softmax(x, axis=1), expected, rtol=1e-13)
+    assert_allclose(softmax(x.T, axis=0), expected.T, rtol=1e-13)
+
+    # 3-d input, with a tuple for the axis.
+    x3d = x.reshape(2, 2, 2)
+    assert_allclose(softmax(x3d, axis=(1, 2)), expected.reshape(2, 2, 2),
+                    rtol=1e-13)

.venv/Lib/site-packages/scipy/special/tests/test_mpmath.py

@@ -0,0 +1,2272 @@
+"""
+Test SciPy functions versus mpmath, if available.
+
+"""
+import numpy as np
+from numpy.testing import assert_, assert_allclose
+from numpy import pi
+import pytest
+import itertools
+
+from scipy._lib import _pep440
+
+import scipy.special as sc
+from scipy.special._testutils import (
+    MissingModule, check_version, FuncData,
+    assert_func_equal)
+from scipy.special._mptestutils import (
+    Arg, FixedArg, ComplexArg, IntArg, assert_mpmath_equal,
+    nonfunctional_tooslow, trace_args, time_limited, exception_to_nan,
+    inf_to_nan)
+from scipy.special._ufuncs import (
+    _sinpi, _cospi, _lgam1p, _lanczos_sum_expg_scaled, _log1pmx,
+    _igam_fac)
+
+try:
+    import mpmath
+except ImportError:
+    mpmath = MissingModule('mpmath')
+
+
+# ------------------------------------------------------------------------------
+# expi
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.10')
+def test_expi_complex():
+    dataset = []
+    for r in np.logspace(-99, 2, 10):
+        for p in np.linspace(0, 2*np.pi, 30):
+            z = r*np.exp(1j*p)
+            dataset.append((z, complex(mpmath.ei(z))))
+    dataset = np.array(dataset, dtype=np.cdouble)
+
+    FuncData(sc.expi, dataset, 0, 1).check()
+
+
+# ------------------------------------------------------------------------------
+# expn
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.19')
+def test_expn_large_n():
+    # Test the transition to the asymptotic regime of n.
+    dataset = []
+    for n in [50, 51]:
+        for x in np.logspace(0, 4, 200):
+            with mpmath.workdps(100):
+                dataset.append((n, x, float(mpmath.expint(n, x))))
+    dataset = np.asarray(dataset)
+
+    FuncData(sc.expn, dataset, (0, 1), 2, rtol=1e-13).check()
+
+# ------------------------------------------------------------------------------
+# hyp0f1
+# ------------------------------------------------------------------------------
+
+
+@check_version(mpmath, '0.19')
+def test_hyp0f1_gh5764():
+    # Do a small and somewhat systematic test that runs quickly
+    dataset = []
+    axis = [-99.5, -9.5, -0.5, 0.5, 9.5, 99.5]
+    for v in axis:
+        for x in axis:
+            for y in axis:
+                z = x + 1j*y
+                # mpmath computes the answer correctly at dps ~ 17 but
+                # fails for 20 < dps < 120 (uses a different method);
+                # set the dps high enough that this isn't an issue
+                with mpmath.workdps(120):
+                    res = complex(mpmath.hyp0f1(v, z))
+                dataset.append((v, z, res))
+    dataset = np.array(dataset)
+
+    FuncData(lambda v, z: sc.hyp0f1(v.real, z), dataset, (0, 1), 2,
+             rtol=1e-13).check()
+
+
+@check_version(mpmath, '0.19')
+def test_hyp0f1_gh_1609():
+    # this is a regression test for gh-1609
+    vv = np.linspace(150, 180, 21)
+    af = sc.hyp0f1(vv, 0.5)
+    mf = np.array([mpmath.hyp0f1(v, 0.5) for v in vv])
+    assert_allclose(af, mf.astype(float), rtol=1e-12)
+
+
+# ------------------------------------------------------------------------------
+# hyperu
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '1.1.0')
+def test_hyperu_around_0():
+    dataset = []
+    # DLMF 13.2.14-15 test points.
+    for n in np.arange(-5, 5):
+        for b in np.linspace(-5, 5, 20):
+            a = -n
+            dataset.append((a, b, 0, float(mpmath.hyperu(a, b, 0))))
+            a = -n + b - 1
+            dataset.append((a, b, 0, float(mpmath.hyperu(a, b, 0))))
+    # DLMF 13.2.16-22 test points.
+    for a in [-10.5, -1.5, -0.5, 0, 0.5, 1, 10]:
+        for b in [-1.0, -0.5, 0, 0.5, 1, 1.5, 2, 2.5]:
+            dataset.append((a, b, 0, float(mpmath.hyperu(a, b, 0))))
+    dataset = np.array(dataset)
+
+    FuncData(sc.hyperu, dataset, (0, 1, 2), 3, rtol=1e-15, atol=5e-13).check()
+
+
+# ------------------------------------------------------------------------------
+# hyp2f1
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '1.0.0')
+def test_hyp2f1_strange_points():
+    pts = [
+        (2, -1, -1, 0.7),  # expected: 2.4
+        (2, -2, -2, 0.7),  # expected: 3.87
+    ]
+    pts += list(itertools.product([2, 1, -0.7, -1000], repeat=4))
+    pts = [
+        (a, b, c, x) for a, b, c, x in pts
+        if b == c and round(b) == b and b < 0 and b != -1000
+    ]
+    kw = dict(eliminate=True)
+    dataset = [p + (float(mpmath.hyp2f1(*p, **kw)),) for p in pts]
+    dataset = np.array(dataset, dtype=np.float64)
+
+    FuncData(sc.hyp2f1, dataset, (0,1,2,3), 4, rtol=1e-10).check()
+
+
+@check_version(mpmath, '0.13')
+def test_hyp2f1_real_some_points():
+    pts = [
+        (1, 2, 3, 0),
+        (1./3, 2./3, 5./6, 27./32),
+        (1./4, 1./2, 3./4, 80./81),
+        (2,-2, -3, 3),
+        (2, -3, -2, 3),
+        (2, -1.5, -1.5, 3),
+        (1, 2, 3, 0),
+        (0.7235, -1, -5, 0.3),
+        (0.25, 1./3, 2, 0.999),
+        (0.25, 1./3, 2, -1),
+        (2, 3, 5, 0.99),
+        (3./2, -0.5, 3, 0.99),
+        (2, 2.5, -3.25, 0.999),
+        (-8, 18.016500331508873, 10.805295997850628, 0.90875647507000001),
+        (-10, 900, -10.5, 0.99),
+        (-10, 900, 10.5, 0.99),
+        (-1, 2, 1, 1.0),
+        (-1, 2, 1, -1.0),
+        (-3, 13, 5, 1.0),
+        (-3, 13, 5, -1.0),
+        (0.5, 1 - 270.5, 1.5, 0.999**2),  # from issue 1561
+    ]
+    dataset = [p + (float(mpmath.hyp2f1(*p)),) for p in pts]
+    dataset = np.array(dataset, dtype=np.float64)
+
+    with np.errstate(invalid='ignore'):
+        FuncData(sc.hyp2f1, dataset, (0,1,2,3), 4, rtol=1e-10).check()
+
+
+@check_version(mpmath, '0.14')
+def test_hyp2f1_some_points_2():
+    # Taken from mpmath unit tests -- this point failed for mpmath 0.13 but
+    # was fixed in their SVN since then
+    pts = [
+        (112, (51,10), (-9,10), -0.99999),
+        (10,-900,10.5,0.99),
+        (10,-900,-10.5,0.99),
+    ]
+
+    def fev(x):
+        if isinstance(x, tuple):
+            return float(x[0]) / x[1]
+        else:
+            return x
+
+    dataset = [tuple(map(fev, p)) + (float(mpmath.hyp2f1(*p)),) for p in pts]
+    dataset = np.array(dataset, dtype=np.float64)
+
+    FuncData(sc.hyp2f1, dataset, (0,1,2,3), 4, rtol=1e-10).check()
+
+
+@check_version(mpmath, '0.13')
+def test_hyp2f1_real_some():
+    dataset = []
+    for a in [-10, -5, -1.8, 1.8, 5, 10]:
+        for b in [-2.5, -1, 1, 7.4]:
+            for c in [-9, -1.8, 5, 20.4]:
+                for z in [-10, -1.01, -0.99, 0, 0.6, 0.95, 1.5, 10]:
+                    try:
+                        v = float(mpmath.hyp2f1(a, b, c, z))
+                    except Exception:
+                        continue
+                    dataset.append((a, b, c, z, v))
+    dataset = np.array(dataset, dtype=np.float64)
+
+    with np.errstate(invalid='ignore'):
+        FuncData(sc.hyp2f1, dataset, (0,1,2,3), 4, rtol=1e-9,
+                 ignore_inf_sign=True).check()
+
+
+@check_version(mpmath, '0.12')
+@pytest.mark.slow
+def test_hyp2f1_real_random():
+    npoints = 500
+    dataset = np.zeros((npoints, 5), np.float64)
+
+    np.random.seed(1234)
+    dataset[:, 0] = np.random.pareto(1.5, npoints)
+    dataset[:, 1] = np.random.pareto(1.5, npoints)
+    dataset[:, 2] = np.random.pareto(1.5, npoints)
+    dataset[:, 3] = 2*np.random.rand(npoints) - 1
+
+    dataset[:, 0] *= (-1)**np.random.randint(2, npoints)
+    dataset[:, 1] *= (-1)**np.random.randint(2, npoints)
+    dataset[:, 2] *= (-1)**np.random.randint(2, npoints)
+
+    for ds in dataset:
+        if mpmath.__version__ < '0.14':
+            # mpmath < 0.14 fails for c too much smaller than a, b
+            if abs(ds[:2]).max() > abs(ds[2]):
+                ds[2] = abs(ds[:2]).max()
+        ds[4] = float(mpmath.hyp2f1(*tuple(ds[:4])))
+
+    FuncData(sc.hyp2f1, dataset, (0, 1, 2, 3), 4, rtol=1e-9).check()
+
+
+# ------------------------------------------------------------------------------
+# erf (complex)
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.14')
+def test_erf_complex():
+    # need to increase mpmath precision for this test
+    old_dps, old_prec = mpmath.mp.dps, mpmath.mp.prec
+    try:
+        mpmath.mp.dps = 70
+        x1, y1 = np.meshgrid(np.linspace(-10, 1, 31), np.linspace(-10, 1, 11))
+        x2, y2 = np.meshgrid(np.logspace(-80, .8, 31), np.logspace(-80, .8, 11))
+        points = np.r_[x1.ravel(),x2.ravel()] + 1j*np.r_[y1.ravel(), y2.ravel()]
+
+        assert_func_equal(sc.erf, lambda x: complex(mpmath.erf(x)), points,
+                          vectorized=False, rtol=1e-13)
+        assert_func_equal(sc.erfc, lambda x: complex(mpmath.erfc(x)), points,
+                          vectorized=False, rtol=1e-13)
+    finally:
+        mpmath.mp.dps, mpmath.mp.prec = old_dps, old_prec
+
+
+# ------------------------------------------------------------------------------
+# lpmv
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.15')
+def test_lpmv():
+    pts = []
+    for x in [-0.99, -0.557, 1e-6, 0.132, 1]:
+        pts.extend([
+            (1, 1, x),
+            (1, -1, x),
+            (-1, 1, x),
+            (-1, -2, x),
+            (1, 1.7, x),
+            (1, -1.7, x),
+            (-1, 1.7, x),
+            (-1, -2.7, x),
+            (1, 10, x),
+            (1, 11, x),
+            (3, 8, x),
+            (5, 11, x),
+            (-3, 8, x),
+            (-5, 11, x),
+            (3, -8, x),
+            (5, -11, x),
+            (-3, -8, x),
+            (-5, -11, x),
+            (3, 8.3, x),
+            (5, 11.3, x),
+            (-3, 8.3, x),
+            (-5, 11.3, x),
+            (3, -8.3, x),
+            (5, -11.3, x),
+            (-3, -8.3, x),
+            (-5, -11.3, x),
+        ])
+
+    def mplegenp(nu, mu, x):
+        if mu == int(mu) and x == 1:
+            # mpmath 0.17 gets this wrong
+            if mu == 0:
+                return 1
+            else:
+                return 0
+        return mpmath.legenp(nu, mu, x)
+
+    dataset = [p + (mplegenp(p[1], p[0], p[2]),) for p in pts]
+    dataset = np.array(dataset, dtype=np.float64)
+
+    def evf(mu, nu, x):
+        return sc.lpmv(mu.astype(int), nu, x)
+
+    with np.errstate(invalid='ignore'):
+        FuncData(evf, dataset, (0,1,2), 3, rtol=1e-10, atol=1e-14).check()
+
+
+# ------------------------------------------------------------------------------
+# beta
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.15')
+def test_beta():
+    np.random.seed(1234)
+
+    b = np.r_[np.logspace(-200, 200, 4),
+              np.logspace(-10, 10, 4),
+              np.logspace(-1, 1, 4),
+              np.arange(-10, 11, 1),
+              np.arange(-10, 11, 1) + 0.5,
+              -1, -2.3, -3, -100.3, -10003.4]
+    a = b
+
+    ab = np.array(np.broadcast_arrays(a[:,None], b[None,:])).reshape(2, -1).T
+
+    old_dps, old_prec = mpmath.mp.dps, mpmath.mp.prec
+    try:
+        mpmath.mp.dps = 400
+
+        assert_func_equal(sc.beta,
+                          lambda a, b: float(mpmath.beta(a, b)),
+                          ab,
+                          vectorized=False,
+                          rtol=1e-10,
+                          ignore_inf_sign=True)
+
+        assert_func_equal(
+            sc.betaln,
+            lambda a, b: float(mpmath.log(abs(mpmath.beta(a, b)))),
+            ab,
+            vectorized=False,
+            rtol=1e-10)
+    finally:
+        mpmath.mp.dps, mpmath.mp.prec = old_dps, old_prec
+
+
+# ------------------------------------------------------------------------------
+# loggamma
+# ------------------------------------------------------------------------------
+
+LOGGAMMA_TAYLOR_RADIUS = 0.2
+
+
+@check_version(mpmath, '0.19')
+def test_loggamma_taylor_transition():
+    # Make sure there isn't a big jump in accuracy when we move from
+    # using the Taylor series to using the recurrence relation.
+
+    r = LOGGAMMA_TAYLOR_RADIUS + np.array([-0.1, -0.01, 0, 0.01, 0.1])
+    theta = np.linspace(0, 2*np.pi, 20)
+    r, theta = np.meshgrid(r, theta)
+    dz = r*np.exp(1j*theta)
+    z = np.r_[1 + dz, 2 + dz].flatten()
+
+    dataset = [(z0, complex(mpmath.loggamma(z0))) for z0 in z]
+    dataset = np.array(dataset)
+
+    FuncData(sc.loggamma, dataset, 0, 1, rtol=5e-14).check()
+
+
+@check_version(mpmath, '0.19')
+def test_loggamma_taylor():
+    # Test around the zeros at z = 1, 2.
+
+    r = np.logspace(-16, np.log10(LOGGAMMA_TAYLOR_RADIUS), 10)
+    theta = np.linspace(0, 2*np.pi, 20)
+    r, theta = np.meshgrid(r, theta)
+    dz = r*np.exp(1j*theta)
+    z = np.r_[1 + dz, 2 + dz].flatten()
+
+    dataset = [(z0, complex(mpmath.loggamma(z0))) for z0 in z]
+    dataset = np.array(dataset)
+
+    FuncData(sc.loggamma, dataset, 0, 1, rtol=5e-14).check()
+
+
+# ------------------------------------------------------------------------------
+# rgamma
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.19')
+@pytest.mark.slow
+def test_rgamma_zeros():
+    # Test around the zeros at z = 0, -1, -2, ...,  -169. (After -169 we
+    # get values that are out of floating point range even when we're
+    # within 0.1 of the zero.)
+
+    # Can't use too many points here or the test takes forever.
+    dx = np.r_[-np.logspace(-1, -13, 3), 0, np.logspace(-13, -1, 3)]
+    dy = dx.copy()
+    dx, dy = np.meshgrid(dx, dy)
+    dz = dx + 1j*dy
+    zeros = np.arange(0, -170, -1).reshape(1, 1, -1)
+    z = (zeros + np.dstack((dz,)*zeros.size)).flatten()
+    with mpmath.workdps(100):
+        dataset = [(z0, complex(mpmath.rgamma(z0))) for z0 in z]
+
+    dataset = np.array(dataset)
+    FuncData(sc.rgamma, dataset, 0, 1, rtol=1e-12).check()
+
+
+# ------------------------------------------------------------------------------
+# digamma
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.19')
+@pytest.mark.slow
+def test_digamma_roots():
+    # Test the special-cased roots for digamma.
+    root = mpmath.findroot(mpmath.digamma, 1.5)
+    roots = [float(root)]
+    root = mpmath.findroot(mpmath.digamma, -0.5)
+    roots.append(float(root))
+    roots = np.array(roots)
+
+    # If we test beyond a radius of 0.24 mpmath will take forever.
+    dx = np.r_[-0.24, -np.logspace(-1, -15, 10), 0, np.logspace(-15, -1, 10), 0.24]
+    dy = dx.copy()
+    dx, dy = np.meshgrid(dx, dy)
+    dz = dx + 1j*dy
+    z = (roots + np.dstack((dz,)*roots.size)).flatten()
+    with mpmath.workdps(30):
+        dataset = [(z0, complex(mpmath.digamma(z0))) for z0 in z]
+
+    dataset = np.array(dataset)
+    FuncData(sc.digamma, dataset, 0, 1, rtol=1e-14).check()
+
+
+@check_version(mpmath, '0.19')
+def test_digamma_negreal():
+    # Test digamma around the negative real axis. Don't do this in
+    # TestSystematic because the points need some jiggering so that
+    # mpmath doesn't take forever.
+
+    digamma = exception_to_nan(mpmath.digamma)
+
+    x = -np.logspace(300, -30, 100)
+    y = np.r_[-np.logspace(0, -3, 5), 0, np.logspace(-3, 0, 5)]
+    x, y = np.meshgrid(x, y)
+    z = (x + 1j*y).flatten()
+
+    with mpmath.workdps(40):
+        dataset = [(z0, complex(digamma(z0))) for z0 in z]
+    dataset = np.asarray(dataset)
+
+    FuncData(sc.digamma, dataset, 0, 1, rtol=1e-13).check()
+
+
+@check_version(mpmath, '0.19')
+def test_digamma_boundary():
+    # Check that there isn't a jump in accuracy when we switch from
+    # using the asymptotic series to the reflection formula.
+
+    x = -np.logspace(300, -30, 100)
+    y = np.array([-6.1, -5.9, 5.9, 6.1])
+    x, y = np.meshgrid(x, y)
+    z = (x + 1j*y).flatten()
+
+    with mpmath.workdps(30):
+        dataset = [(z0, complex(mpmath.digamma(z0))) for z0 in z]
+    dataset = np.asarray(dataset)
+
+    FuncData(sc.digamma, dataset, 0, 1, rtol=1e-13).check()
+
+
+# ------------------------------------------------------------------------------
+# gammainc
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.19')
+@pytest.mark.slow
+def test_gammainc_boundary():
+    # Test the transition to the asymptotic series.
+    small = 20
+    a = np.linspace(0.5*small, 2*small, 50)
+    x = a.copy()
+    a, x = np.meshgrid(a, x)
+    a, x = a.flatten(), x.flatten()
+    with mpmath.workdps(100):
+        dataset = [(a0, x0, float(mpmath.gammainc(a0, b=x0, regularized=True)))
+                   for a0, x0 in zip(a, x)]
+    dataset = np.array(dataset)
+
+    FuncData(sc.gammainc, dataset, (0, 1), 2, rtol=1e-12).check()
+
+
+# ------------------------------------------------------------------------------
+# spence
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.19')
+@pytest.mark.slow
+def test_spence_circle():
+    # The trickiest region for spence is around the circle |z - 1| = 1,
+    # so test that region carefully.
+
+    def spence(z):
+        return complex(mpmath.polylog(2, 1 - z))
+
+    r = np.linspace(0.5, 1.5)
+    theta = np.linspace(0, 2*pi)
+    z = (1 + np.outer(r, np.exp(1j*theta))).flatten()
+    dataset = np.asarray([(z0, spence(z0)) for z0 in z])
+
+    FuncData(sc.spence, dataset, 0, 1, rtol=1e-14).check()
+
+
+# ------------------------------------------------------------------------------
+# sinpi and cospi
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.19')
+def test_sinpi_zeros():
+    eps = np.finfo(float).eps
+    dx = np.r_[-np.logspace(0, -13, 3), 0, np.logspace(-13, 0, 3)]
+    dy = dx.copy()
+    dx, dy = np.meshgrid(dx, dy)
+    dz = dx + 1j*dy
+    zeros = np.arange(-100, 100, 1).reshape(1, 1, -1)
+    z = (zeros + np.dstack((dz,)*zeros.size)).flatten()
+    dataset = np.asarray([(z0, complex(mpmath.sinpi(z0)))
+                          for z0 in z])
+    FuncData(_sinpi, dataset, 0, 1, rtol=2*eps).check()
+
+
+@check_version(mpmath, '0.19')
+def test_cospi_zeros():
+    eps = np.finfo(float).eps
+    dx = np.r_[-np.logspace(0, -13, 3), 0, np.logspace(-13, 0, 3)]
+    dy = dx.copy()
+    dx, dy = np.meshgrid(dx, dy)
+    dz = dx + 1j*dy
+    zeros = (np.arange(-100, 100, 1) + 0.5).reshape(1, 1, -1)
+    z = (zeros + np.dstack((dz,)*zeros.size)).flatten()
+    dataset = np.asarray([(z0, complex(mpmath.cospi(z0)))
+                          for z0 in z])
+
+    FuncData(_cospi, dataset, 0, 1, rtol=2*eps).check()
+
+
+# ------------------------------------------------------------------------------
+# ellipj
+# ------------------------------------------------------------------------------
+
+@check_version(mpmath, '0.19')
+def test_dn_quarter_period():
+    def dn(u, m):
+        return sc.ellipj(u, m)[2]
+
+    def mpmath_dn(u, m):
+        return float(mpmath.ellipfun("dn", u=u, m=m))
+
+    m = np.linspace(0, 1, 20)
+    du = np.r_[-np.logspace(-1, -15, 10), 0, np.logspace(-15, -1, 10)]
+    dataset = []
+    for m0 in m:
+        u0 = float(mpmath.ellipk(m0))
+        for du0 in du:
+            p = u0 + du0
+            dataset.append((p, m0, mpmath_dn(p, m0)))
+    dataset = np.asarray(dataset)
+
+    FuncData(dn, dataset, (0, 1), 2, rtol=1e-10).check()
+
+
+# ------------------------------------------------------------------------------
+# Wright Omega
+# ------------------------------------------------------------------------------
+
+def _mpmath_wrightomega(z, dps):
+    with mpmath.workdps(dps):
+        z = mpmath.mpc(z)
+        unwind = mpmath.ceil((z.imag - mpmath.pi)/(2*mpmath.pi))
+        res = mpmath.lambertw(mpmath.exp(z), unwind)
+    return res
+
+
+@pytest.mark.slow
+@check_version(mpmath, '0.19')
+def test_wrightomega_branch():
+    x = -np.logspace(10, 0, 25)
+    picut_above = [np.nextafter(np.pi, np.inf)]
+    picut_below = [np.nextafter(np.pi, -np.inf)]
+    npicut_above = [np.nextafter(-np.pi, np.inf)]
+    npicut_below = [np.nextafter(-np.pi, -np.inf)]
+    for i in range(50):
+        picut_above.append(np.nextafter(picut_above[-1], np.inf))
+        picut_below.append(np.nextafter(picut_below[-1], -np.inf))
+        npicut_above.append(np.nextafter(npicut_above[-1], np.inf))
+        npicut_below.append(np.nextafter(npicut_below[-1], -np.inf))
+    y = np.hstack((picut_above, picut_below, npicut_above, npicut_below))
+    x, y = np.meshgrid(x, y)
+    z = (x + 1j*y).flatten()
+
+    dataset = np.asarray([(z0, complex(_mpmath_wrightomega(z0, 25)))
+                          for z0 in z])
+
+    FuncData(sc.wrightomega, dataset, 0, 1, rtol=1e-8).check()
+
+
+@pytest.mark.slow
+@check_version(mpmath, '0.19')
+def test_wrightomega_region1():
+    # This region gets less coverage in the TestSystematic test
+    x = np.linspace(-2, 1)
+    y = np.linspace(1, 2*np.pi)
+    x, y = np.meshgrid(x, y)
+    z = (x + 1j*y).flatten()
+
+    dataset = np.asarray([(z0, complex(_mpmath_wrightomega(z0, 25)))
+                          for z0 in z])
+
+    FuncData(sc.wrightomega, dataset, 0, 1, rtol=1e-15).check()
+
+
+@pytest.mark.slow
+@check_version(mpmath, '0.19')
+def test_wrightomega_region2():
+    # This region gets less coverage in the TestSystematic test
+    x = np.linspace(-2, 1)
+    y = np.linspace(-2*np.pi, -1)
+    x, y = np.meshgrid(x, y)
+    z = (x + 1j*y).flatten()
+
+    dataset = np.asarray([(z0, complex(_mpmath_wrightomega(z0, 25)))
+                          for z0 in z])
+
+    FuncData(sc.wrightomega, dataset, 0, 1, rtol=1e-15).check()
+
+
+# ------------------------------------------------------------------------------
+# lambertw
+# ------------------------------------------------------------------------------
+
+@pytest.mark.slow
+@check_version(mpmath, '0.19')
+def test_lambertw_smallz():
+    x, y = np.linspace(-1, 1, 25), np.linspace(-1, 1, 25)
+    x, y = np.meshgrid(x, y)
+    z = (x + 1j*y).flatten()
+
+    dataset = np.asarray([(z0, complex(mpmath.lambertw(z0)))
+                          for z0 in z])
+
+    FuncData(sc.lambertw, dataset, 0, 1, rtol=1e-13).check()
+
+
+# ------------------------------------------------------------------------------
+# Systematic tests
+# ------------------------------------------------------------------------------
+
+HYPERKW = dict(maxprec=200, maxterms=200)
+
+
+@pytest.mark.slow
+@check_version(mpmath, '0.17')
+class TestSystematic:
+
+    def test_airyai(self):
+        # oscillating function, limit range
+        assert_mpmath_equal(lambda z: sc.airy(z)[0],
+                            mpmath.airyai,
+                            [Arg(-1e8, 1e8)],
+                            rtol=1e-5)
+        assert_mpmath_equal(lambda z: sc.airy(z)[0],
+                            mpmath.airyai,
+                            [Arg(-1e3, 1e3)])
+
+    def test_airyai_complex(self):
+        assert_mpmath_equal(lambda z: sc.airy(z)[0],
+                            mpmath.airyai,
+                            [ComplexArg()])
+
+    def test_airyai_prime(self):
+        # oscillating function, limit range
+        assert_mpmath_equal(lambda z: sc.airy(z)[1], lambda z:
+                            mpmath.airyai(z, derivative=1),
+                            [Arg(-1e8, 1e8)],
+                            rtol=1e-5)
+        assert_mpmath_equal(lambda z: sc.airy(z)[1], lambda z:
+                            mpmath.airyai(z, derivative=1),
+                            [Arg(-1e3, 1e3)])
+
+    def test_airyai_prime_complex(self):
+        assert_mpmath_equal(lambda z: sc.airy(z)[1], lambda z:
+                            mpmath.airyai(z, derivative=1),
+                            [ComplexArg()])
+
+    def test_airybi(self):
+        # oscillating function, limit range
+        assert_mpmath_equal(lambda z: sc.airy(z)[2], lambda z:
+                            mpmath.airybi(z),
+                            [Arg(-1e8, 1e8)],
+                            rtol=1e-5)
+        assert_mpmath_equal(lambda z: sc.airy(z)[2], lambda z:
+                            mpmath.airybi(z),
+                            [Arg(-1e3, 1e3)])
+
+    def test_airybi_complex(self):
+        assert_mpmath_equal(lambda z: sc.airy(z)[2], lambda z:
+                            mpmath.airybi(z),
+                            [ComplexArg()])
+
+    def test_airybi_prime(self):
+        # oscillating function, limit range
+        assert_mpmath_equal(lambda z: sc.airy(z)[3], lambda z:
+                            mpmath.airybi(z, derivative=1),
+                            [Arg(-1e8, 1e8)],
+                            rtol=1e-5)
+        assert_mpmath_equal(lambda z: sc.airy(z)[3], lambda z:
+                            mpmath.airybi(z, derivative=1),
+                            [Arg(-1e3, 1e3)])
+
+    def test_airybi_prime_complex(self):
+        assert_mpmath_equal(lambda z: sc.airy(z)[3], lambda z:
+                            mpmath.airybi(z, derivative=1),
+                            [ComplexArg()])
+
+    def test_bei(self):
+        assert_mpmath_equal(sc.bei,
+                            exception_to_nan(lambda z: mpmath.bei(0, z, **HYPERKW)),
+                            [Arg(-1e3, 1e3)])
+
+    def test_ber(self):
+        assert_mpmath_equal(sc.ber,
+                            exception_to_nan(lambda z: mpmath.ber(0, z, **HYPERKW)),
+                            [Arg(-1e3, 1e3)])
+
+    def test_bernoulli(self):
+        assert_mpmath_equal(lambda n: sc.bernoulli(int(n))[int(n)],
+                            lambda n: float(mpmath.bernoulli(int(n))),
+                            [IntArg(0, 13000)],
+                            rtol=1e-9, n=13000)
+
+    def test_besseli(self):
+        assert_mpmath_equal(
+            sc.iv,
+            exception_to_nan(lambda v, z: mpmath.besseli(v, z, **HYPERKW)),
+            [Arg(-1e100, 1e100), Arg()],
+            atol=1e-270,
+        )
+
+    def test_besseli_complex(self):
+        assert_mpmath_equal(
+            lambda v, z: sc.iv(v.real, z),
+            exception_to_nan(lambda v, z: mpmath.besseli(v, z, **HYPERKW)),
+            [Arg(-1e100, 1e100), ComplexArg()],
+        )
+
+    def test_besselj(self):
+        assert_mpmath_equal(
+            sc.jv,
+            exception_to_nan(lambda v, z: mpmath.besselj(v, z, **HYPERKW)),
+            [Arg(-1e100, 1e100), Arg(-1e3, 1e3)],
+            ignore_inf_sign=True,
+        )
+
+        # loss of precision at large arguments due to oscillation
+        assert_mpmath_equal(
+            sc.jv,
+            exception_to_nan(lambda v, z: mpmath.besselj(v, z, **HYPERKW)),
+            [Arg(-1e100, 1e100), Arg(-1e8, 1e8)],
+            ignore_inf_sign=True,
+            rtol=1e-5,
+        )
+
+    def test_besselj_complex(self):
+        assert_mpmath_equal(
+            lambda v, z: sc.jv(v.real, z),
+            exception_to_nan(lambda v, z: mpmath.besselj(v, z, **HYPERKW)),
+            [Arg(), ComplexArg()]
+        )
+
+    def test_besselk(self):
+        assert_mpmath_equal(
+            sc.kv,
+            mpmath.besselk,
+            [Arg(-200, 200), Arg(0, np.inf)],
+            nan_ok=False,
+            rtol=1e-12,
+        )
+
+    def test_besselk_int(self):
+        assert_mpmath_equal(
+            sc.kn,
+            mpmath.besselk,
+            [IntArg(-200, 200), Arg(0, np.inf)],
+            nan_ok=False,
+            rtol=1e-12,
+        )
+
+    def test_besselk_complex(self):
+        assert_mpmath_equal(
+            lambda v, z: sc.kv(v.real, z),
+            exception_to_nan(lambda v, z: mpmath.besselk(v, z, **HYPERKW)),
+            [Arg(-1e100, 1e100), ComplexArg()],
+        )
+
+    def test_bessely(self):
+        def mpbessely(v, x):
+            r = float(mpmath.bessely(v, x, **HYPERKW))
+            if abs(r) > 1e305:
+                # overflowing to inf a bit earlier is OK
+                r = np.inf * np.sign(r)
+            if abs(r) == 0 and x == 0:
+                # invalid result from mpmath, point x=0 is a divergence
+                return np.nan
+            return r
+        assert_mpmath_equal(
+            sc.yv,
+            exception_to_nan(mpbessely),
+            [Arg(-1e100, 1e100), Arg(-1e8, 1e8)],
+            n=5000,
+        )
+
+    def test_bessely_complex(self):
+        def mpbessely(v, x):
+            r = complex(mpmath.bessely(v, x, **HYPERKW))
+            if abs(r) > 1e305:
+                # overflowing to inf a bit earlier is OK
+                with np.errstate(invalid='ignore'):
+                    r = np.inf * np.sign(r)
+            return r
+        assert_mpmath_equal(
+            lambda v, z: sc.yv(v.real, z),
+            exception_to_nan(mpbessely),
+            [Arg(), ComplexArg()],
+            n=15000,
+        )
+
+    def test_bessely_int(self):
+        def mpbessely(v, x):
+            r = float(mpmath.bessely(v, x))
+            if abs(r) == 0 and x == 0:
+                # invalid result from mpmath, point x=0 is a divergence
+                return np.nan
+            return r
+        assert_mpmath_equal(
+            lambda v, z: sc.yn(int(v), z),
+            exception_to_nan(mpbessely),
+            [IntArg(-1000, 1000), Arg(-1e8, 1e8)],
+        )
+
+    def test_beta(self):
+        bad_points = []
+
+        def beta(a, b, nonzero=False):
+            if a < -1e12 or b < -1e12:
+                # Function is defined here only at integers, but due
+                # to loss of precision this is numerically
+                # ill-defined. Don't compare values here.
+                return np.nan
+            if (a < 0 or b < 0) and (abs(float(a + b)) % 1) == 0:
+                # close to a zero of the function: mpmath and scipy
+                # will not round here the same, so the test needs to be
+                # run with an absolute tolerance
+                if nonzero:
+                    bad_points.append((float(a), float(b)))
+                    return np.nan
+            return mpmath.beta(a, b)
+
+        assert_mpmath_equal(
+            sc.beta,
+            lambda a, b: beta(a, b, nonzero=True),
+            [Arg(), Arg()],
+            dps=400,
+            ignore_inf_sign=True,
+        )
+
+        assert_mpmath_equal(
+            sc.beta,
+            beta,
+            np.array(bad_points),
+            dps=400,
+            ignore_inf_sign=True,
+            atol=1e-11,
+        )
+
+    def test_betainc(self):
+        assert_mpmath_equal(
+            sc.betainc,
+            time_limited()(
+                exception_to_nan(
+                    lambda a, b, x: mpmath.betainc(a, b, 0, x, regularized=True)
+                )
+            ),
+            [Arg(), Arg(), Arg()],
+        )
+
+    def test_betaincc(self):
+        assert_mpmath_equal(
+            sc.betaincc,
+            time_limited()(
+                exception_to_nan(
+                    lambda a, b, x: mpmath.betainc(a, b, x, 1, regularized=True)
+                )
+            ),
+            [Arg(), Arg(), Arg()],
+            dps=400,
+        )
+
+    def test_binom(self):
+        bad_points = []
+
+        def binomial(n, k, nonzero=False):
+            if abs(k) > 1e8*(abs(n) + 1):
+                # The binomial is rapidly oscillating in this region,
+                # and the function is numerically ill-defined. Don't
+                # compare values here.
+                return np.nan
+            if n < k and abs(float(n-k) - np.round(float(n-k))) < 1e-15:
+                # close to a zero of the function: mpmath and scipy
+                # will not round here the same, so the test needs to be
+                # run with an absolute tolerance
+                if nonzero:
+                    bad_points.append((float(n), float(k)))
+                    return np.nan
+            return mpmath.binomial(n, k)
+
+        assert_mpmath_equal(
+            sc.binom,
+            lambda n, k: binomial(n, k, nonzero=True),
+            [Arg(), Arg()],
+            dps=400,
+        )
+
+        assert_mpmath_equal(
+            sc.binom,
+            binomial,
+            np.array(bad_points),
+            dps=400,
+            atol=1e-14,
+        )
+
+    def test_chebyt_int(self):
+        assert_mpmath_equal(
+            lambda n, x: sc.eval_chebyt(int(n), x),
+            exception_to_nan(lambda n, x: mpmath.chebyt(n, x, **HYPERKW)),
+            [IntArg(), Arg()],
+            dps=50,
+        )
+
+    @pytest.mark.xfail(run=False, reason="some cases in hyp2f1 not fully accurate")
+    def test_chebyt(self):
+        assert_mpmath_equal(
+            sc.eval_chebyt,
+            lambda n, x: time_limited()(
+                exception_to_nan(mpmath.chebyt)
+            )(n, x, **HYPERKW),
+            [Arg(-101, 101), Arg()],
+            n=10000,
+        )
+
+    def test_chebyu_int(self):
+        assert_mpmath_equal(
+            lambda n, x: sc.eval_chebyu(int(n), x),
+            exception_to_nan(lambda n, x: mpmath.chebyu(n, x, **HYPERKW)),
+            [IntArg(), Arg()],
+            dps=50,
+        )
+
+    @pytest.mark.xfail(run=False, reason="some cases in hyp2f1 not fully accurate")
+    def test_chebyu(self):
+        assert_mpmath_equal(
+            sc.eval_chebyu,
+            lambda n, x: time_limited()(
+                exception_to_nan(mpmath.chebyu)
+            )(n, x, **HYPERKW),
+            [Arg(-101, 101), Arg()],
+        )
+
+    def test_chi(self):
+        def chi(x):
+            return sc.shichi(x)[1]
+        assert_mpmath_equal(chi, mpmath.chi, [Arg()])
+        # check asymptotic series cross-over
+        assert_mpmath_equal(chi, mpmath.chi, [FixedArg([88 - 1e-9, 88, 88 + 1e-9])])
+
+    def test_chi_complex(self):
+        def chi(z):
+            return sc.shichi(z)[1]
+        # chi oscillates as Im[z] -> +- inf, so limit range
+        assert_mpmath_equal(
+            chi,
+            mpmath.chi,
+            [ComplexArg(complex(-np.inf, -1e8), complex(np.inf, 1e8))],
+            rtol=1e-12,
+        )
+
+    def test_ci(self):
+        def ci(x):
+            return sc.sici(x)[1]
+        # oscillating function: limit range
+        assert_mpmath_equal(ci, mpmath.ci, [Arg(-1e8, 1e8)])
+
+    def test_ci_complex(self):
+        def ci(z):
+            return sc.sici(z)[1]
+        # ci oscillates as Re[z] -> +- inf, so limit range
+        assert_mpmath_equal(
+            ci,
+            mpmath.ci,
+            [ComplexArg(complex(-1e8, -np.inf), complex(1e8, np.inf))],
+            rtol=1e-8,
+        )
+
+    def test_cospi(self):
+        eps = np.finfo(float).eps
+        assert_mpmath_equal(_cospi, mpmath.cospi, [Arg()], nan_ok=False, rtol=2*eps)
+
+    def test_cospi_complex(self):
+        assert_mpmath_equal(
+            _cospi,
+            mpmath.cospi,
+            [ComplexArg()],
+            nan_ok=False,
+            rtol=1e-13,
+        )
+
+    def test_digamma(self):
+        assert_mpmath_equal(
+            sc.digamma,
+            exception_to_nan(mpmath.digamma),
+            [Arg()],
+            rtol=1e-12,
+            dps=50,
+        )
+
+    def test_digamma_complex(self):
+        # Test on a cut plane because mpmath will hang. See
+        # test_digamma_negreal for tests on the negative real axis.
+        def param_filter(z):
+            return np.where((z.real < 0) & (np.abs(z.imag) < 1.12), False, True)
+
+        assert_mpmath_equal(
+            sc.digamma,
+            exception_to_nan(mpmath.digamma),
+            [ComplexArg()],
+            rtol=1e-13,
+            dps=40,
+            param_filter=param_filter
+        )
+
+    def test_e1(self):
+        assert_mpmath_equal(
+            sc.exp1,
+            mpmath.e1,
+            [Arg()],
+            rtol=1e-14,
+        )
+
+    def test_e1_complex(self):
+        # E_1 oscillates as Im[z] -> +- inf, so limit range
+        assert_mpmath_equal(
+            sc.exp1,
+            mpmath.e1,
+            [ComplexArg(complex(-np.inf, -1e8), complex(np.inf, 1e8))],
+            rtol=1e-11,
+        )
+
+        # Check cross-over region
+        assert_mpmath_equal(
+            sc.exp1,
+            mpmath.e1,
+            (np.linspace(-50, 50, 171)[:, None]
+             + np.r_[0, np.logspace(-3, 2, 61), -np.logspace(-3, 2, 11)]*1j).ravel(),
+            rtol=1e-11,
+        )
+        assert_mpmath_equal(
+            sc.exp1,
+            mpmath.e1,
+            (np.linspace(-50, -35, 10000) + 0j),
+            rtol=1e-11,
+        )
+
+    def test_exprel(self):
+        assert_mpmath_equal(
+            sc.exprel,
+            lambda x: mpmath.expm1(x)/x if x != 0 else mpmath.mpf('1.0'),
+            [Arg(a=-np.log(np.finfo(np.float64).max),
+                 b=np.log(np.finfo(np.float64).max))],
+        )
+        assert_mpmath_equal(
+            sc.exprel,
+            lambda x: mpmath.expm1(x)/x if x != 0 else mpmath.mpf('1.0'),
+            np.array([1e-12, 1e-24, 0, 1e12, 1e24, np.inf]),
+            rtol=1e-11,
+        )
+        assert_(np.isinf(sc.exprel(np.inf)))
+        assert_(sc.exprel(-np.inf) == 0)
+
+    def test_expm1_complex(self):
+        # Oscillates as a function of Im[z], so limit range to avoid loss of precision
+        assert_mpmath_equal(
+            sc.expm1,
+            mpmath.expm1,
+            [ComplexArg(complex(-np.inf, -1e7), complex(np.inf, 1e7))],
+        )
+
+    def test_log1p_complex(self):
+        assert_mpmath_equal(
+            sc.log1p,
+            lambda x: mpmath.log(x+1),
+            [ComplexArg()],
+            dps=60,
+        )
+
+    def test_log1pmx(self):
+        assert_mpmath_equal(
+            _log1pmx,
+            lambda x: mpmath.log(x + 1) - x,
+            [Arg()],
+            dps=60,
+            rtol=1e-14,
+        )
+
+    def test_ei(self):
+        assert_mpmath_equal(sc.expi, mpmath.ei, [Arg()], rtol=1e-11)
+
+    def test_ei_complex(self):
+        # Ei oscillates as Im[z] -> +- inf, so limit range
+        assert_mpmath_equal(
+            sc.expi,
+            mpmath.ei,
+            [ComplexArg(complex(-np.inf, -1e8), complex(np.inf, 1e8))],
+            rtol=1e-9,
+        )
+
+    def test_ellipe(self):
+        assert_mpmath_equal(sc.ellipe, mpmath.ellipe, [Arg(b=1.0)])
+
+    def test_ellipeinc(self):
+        assert_mpmath_equal(sc.ellipeinc, mpmath.ellipe, [Arg(-1e3, 1e3), Arg(b=1.0)])
+
+    def test_ellipeinc_largephi(self):
+        assert_mpmath_equal(sc.ellipeinc, mpmath.ellipe, [Arg(), Arg()])
+
+    def test_ellipf(self):
+        assert_mpmath_equal(sc.ellipkinc, mpmath.ellipf, [Arg(-1e3, 1e3), Arg()])
+
+    def test_ellipf_largephi(self):
+        assert_mpmath_equal(sc.ellipkinc, mpmath.ellipf, [Arg(), Arg()])
+
+    def test_ellipk(self):
+        assert_mpmath_equal(sc.ellipk, mpmath.ellipk, [Arg(b=1.0)])
+        assert_mpmath_equal(
+            sc.ellipkm1,
+            lambda m: mpmath.ellipk(1 - m),
+            [Arg(a=0.0)],
+            dps=400,
+        )
+
+    def test_ellipkinc(self):
+        def ellipkinc(phi, m):
+            return mpmath.ellippi(0, phi, m)
+        assert_mpmath_equal(
+            sc.ellipkinc,
+            ellipkinc,
+            [Arg(-1e3, 1e3), Arg(b=1.0)],
+            ignore_inf_sign=True,
+        )
+
+    def test_ellipkinc_largephi(self):
+        def ellipkinc(phi, m):
+            return mpmath.ellippi(0, phi, m)
+        assert_mpmath_equal(
+            sc.ellipkinc,
+            ellipkinc,
+            [Arg(), Arg(b=1.0)],
+            ignore_inf_sign=True,
+        )
+
+    def test_ellipfun_sn(self):
+        def sn(u, m):
+            # mpmath doesn't get the zero at u = 0--fix that
+            if u == 0:
+                return 0
+            else:
+                return mpmath.ellipfun("sn", u=u, m=m)
+
+        # Oscillating function --- limit range of first argument; the
+        # loss of precision there is an expected numerical feature
+        # rather than an actual bug
+        assert_mpmath_equal(
+            lambda u, m: sc.ellipj(u, m)[0],
+            sn,
+            [Arg(-1e6, 1e6), Arg(a=0, b=1)],
+            rtol=1e-8,
+        )
+
+    def test_ellipfun_cn(self):
+        # see comment in ellipfun_sn
+        assert_mpmath_equal(
+            lambda u, m: sc.ellipj(u, m)[1],
+            lambda u, m: mpmath.ellipfun("cn", u=u, m=m),
+            [Arg(-1e6, 1e6), Arg(a=0, b=1)],
+            rtol=1e-8,
+        )
+
+    def test_ellipfun_dn(self):
+        # see comment in ellipfun_sn
+        assert_mpmath_equal(
+            lambda u, m: sc.ellipj(u, m)[2],
+            lambda u, m: mpmath.ellipfun("dn", u=u, m=m),
+            [Arg(-1e6, 1e6), Arg(a=0, b=1)],
+            rtol=1e-8,
+        )
+
+    def test_erf(self):
+        assert_mpmath_equal(sc.erf, lambda z: mpmath.erf(z), [Arg()])
+
+    def test_erf_complex(self):
+        assert_mpmath_equal(sc.erf, lambda z: mpmath.erf(z), [ComplexArg()], n=200)
+
+    def test_erfc(self):
+        assert_mpmath_equal(
+            sc.erfc,
+            exception_to_nan(lambda z: mpmath.erfc(z)),
+            [Arg()],
+            rtol=1e-13,
+        )
+
+    def test_erfc_complex(self):
+        assert_mpmath_equal(
+            sc.erfc,
+            exception_to_nan(lambda z: mpmath.erfc(z)),
+            [ComplexArg()],
+            n=200,
+        )
+
+    def test_erfi(self):
+        assert_mpmath_equal(sc.erfi, mpmath.erfi, [Arg()], n=200)
+
+    def test_erfi_complex(self):
+        assert_mpmath_equal(sc.erfi, mpmath.erfi, [ComplexArg()], n=200)
+
+    def test_ndtr(self):
+        assert_mpmath_equal(
+            sc.ndtr,
+            exception_to_nan(lambda z: mpmath.ncdf(z)),
+            [Arg()],
+            n=200,
+        )
+
+    def test_ndtr_complex(self):
+        assert_mpmath_equal(
+            sc.ndtr,
+            lambda z: mpmath.erfc(-z/np.sqrt(2.))/2.,
+            [ComplexArg(a=complex(-10000, -10000), b=complex(10000, 10000))],
+            n=400,
+        )
+
+    def test_log_ndtr(self):
+        assert_mpmath_equal(
+            sc.log_ndtr,
+            exception_to_nan(lambda z: mpmath.log(mpmath.ncdf(z))),
+            [Arg()], n=600, dps=300, rtol=1e-13,
+        )
+
+    def test_log_ndtr_complex(self):
+        assert_mpmath_equal(
+            sc.log_ndtr,
+            exception_to_nan(lambda z: mpmath.log(mpmath.erfc(-z/np.sqrt(2.))/2.)),
+            [ComplexArg(a=complex(-10000, -100), b=complex(10000, 100))],
+            n=200, dps=300,
+        )
+
+    def test_eulernum(self):
+        assert_mpmath_equal(
+            lambda n: sc.euler(n)[-1],
+            mpmath.eulernum,
+            [IntArg(1, 10000)],
+            n=10000,
+        )
+
+    def test_expint(self):
+        assert_mpmath_equal(
+            sc.expn,
+            mpmath.expint,
+            [IntArg(0, 200), Arg(0, np.inf)],
+            rtol=1e-13,
+            dps=160,
+        )
+
+    def test_fresnels(self):
+        def fresnels(x):
+            return sc.fresnel(x)[0]
+        assert_mpmath_equal(fresnels, mpmath.fresnels, [Arg()])
+
+    def test_fresnelc(self):
+        def fresnelc(x):
+            return sc.fresnel(x)[1]
+        assert_mpmath_equal(fresnelc, mpmath.fresnelc, [Arg()])
+
+    def test_gamma(self):
+        assert_mpmath_equal(sc.gamma, exception_to_nan(mpmath.gamma), [Arg()])
+
+    def test_gamma_complex(self):
+        assert_mpmath_equal(
+            sc.gamma,
+            exception_to_nan(mpmath.gamma),
+            [ComplexArg()], 
+            rtol=5e-13,
+        )
+
+    def test_gammainc(self):
+        # Larger arguments are tested in test_data.py:test_local
+        assert_mpmath_equal(
+            sc.gammainc,
+            lambda z, b: mpmath.gammainc(z, b=b, regularized=True),
+            [Arg(0, 1e4, inclusive_a=False), Arg(0, 1e4)],
+            nan_ok=False,
+            rtol=1e-11,
+        )
+
+    def test_gammaincc(self):
+        # Larger arguments are tested in test_data.py:test_local
+        assert_mpmath_equal(
+            sc.gammaincc,
+            lambda z, a: mpmath.gammainc(z, a=a, regularized=True),
+            [Arg(0, 1e4, inclusive_a=False), Arg(0, 1e4)],
+            nan_ok=False,
+            rtol=1e-11,
+        )
+
+    def test_gammaln(self):
+        # The real part of loggamma is log(|gamma(z)|).
+        def f(z):
+            return mpmath.loggamma(z).real
+
+        assert_mpmath_equal(sc.gammaln, exception_to_nan(f), [Arg()])
+
+    @pytest.mark.xfail(run=False)
+    def test_gegenbauer(self):
+        assert_mpmath_equal(
+            sc.eval_gegenbauer,
+            exception_to_nan(mpmath.gegenbauer),
+            [Arg(-1e3, 1e3), Arg(), Arg()],
+        )
+
+    def test_gegenbauer_int(self):
+        # Redefine functions to deal with numerical + mpmath issues
+        def gegenbauer(n, a, x):
+            # Avoid overflow at large `a` (mpmath would need an even larger
+            # dps to handle this correctly, so just skip this region)
+            if abs(a) > 1e100:
+                return np.nan
+
+            # Deal with n=0, n=1 correctly; mpmath 0.17 doesn't do these
+            # always correctly
+            if n == 0:
+                r = 1.0
+            elif n == 1:
+                r = 2*a*x
+            else:
+                r = mpmath.gegenbauer(n, a, x)
+
+            # Mpmath 0.17 gives wrong results (spurious zero) in some cases, so
+            # compute the value by perturbing the result
+            if float(r) == 0 and a < -1 and float(a) == int(float(a)):
+                r = mpmath.gegenbauer(n, a + mpmath.mpf('1e-50'), x)
+                if abs(r) < mpmath.mpf('1e-50'):
+                    r = mpmath.mpf('0.0')
+
+            # Differing overflow thresholds in scipy vs. mpmath
+            if abs(r) > 1e270:
+                return np.inf
+            return r
+
+        def sc_gegenbauer(n, a, x):
+            r = sc.eval_gegenbauer(int(n), a, x)
+            # Differing overflow thresholds in scipy vs. mpmath
+            if abs(r) > 1e270:
+                return np.inf
+            return r
+        assert_mpmath_equal(
+            sc_gegenbauer,
+            exception_to_nan(gegenbauer),
+            [IntArg(0, 100), Arg(-1e9, 1e9), Arg()],
+            n=40000, dps=100, ignore_inf_sign=True, rtol=1e-6,
+        )
+
+        # Check the small-x expansion
+        assert_mpmath_equal(
+            sc_gegenbauer,
+            exception_to_nan(gegenbauer),
+            [IntArg(0, 100), Arg(), FixedArg(np.logspace(-30, -4, 30))],
+            dps=100, ignore_inf_sign=True,
+        )
+
+    @pytest.mark.xfail(run=False)
+    def test_gegenbauer_complex(self):
+        assert_mpmath_equal(
+            lambda n, a, x: sc.eval_gegenbauer(int(n), a.real, x),
+            exception_to_nan(mpmath.gegenbauer),
+            [IntArg(0, 100), Arg(), ComplexArg()],
+        )
+
+    @nonfunctional_tooslow
+    def test_gegenbauer_complex_general(self):
+        assert_mpmath_equal(
+            lambda n, a, x: sc.eval_gegenbauer(n.real, a.real, x),
+            exception_to_nan(mpmath.gegenbauer),
+            [Arg(-1e3, 1e3), Arg(), ComplexArg()],
+        )
+
+    def test_hankel1(self):
+        assert_mpmath_equal(
+            sc.hankel1,
+            exception_to_nan(lambda v, x: mpmath.hankel1(v, x, **HYPERKW)),
+            [Arg(-1e20, 1e20), Arg()],
+        )
+
+    def test_hankel2(self):
+        assert_mpmath_equal(
+            sc.hankel2,
+            exception_to_nan(lambda v, x: mpmath.hankel2(v, x, **HYPERKW)),
+            [Arg(-1e20, 1e20), Arg()],
+        )
+
+    @pytest.mark.xfail(run=False, reason="issues at intermediately large orders")
+    def test_hermite(self):
+        assert_mpmath_equal(
+            lambda n, x: sc.eval_hermite(int(n), x),
+            exception_to_nan(mpmath.hermite),
+            [IntArg(0, 10000), Arg()],
+        )
+
+    # hurwitz: same as zeta
+
+    def test_hyp0f1(self):
+        # mpmath reports no convergence unless maxterms is large enough
+        KW = dict(maxprec=400, maxterms=1500)
+        # n=500 (non-xslow default) fails for one bad point
+        assert_mpmath_equal(
+            sc.hyp0f1,
+            lambda a, x: mpmath.hyp0f1(a, x, **KW),
+            [Arg(-1e7, 1e7), Arg(0, 1e5)],
+            n=5000,
+        )
+        # NB: The range of the second parameter ("z") is limited from below
+        # because of an overflow in the intermediate calculations. The way
+        # for fix it is to implement an asymptotic expansion for Bessel J
+        # (similar to what is implemented for Bessel I here).
+
+    def test_hyp0f1_complex(self):
+        assert_mpmath_equal(
+            lambda a, z: sc.hyp0f1(a.real, z),
+            exception_to_nan(lambda a, x: mpmath.hyp0f1(a, x, **HYPERKW)),
+            [Arg(-10, 10), ComplexArg(complex(-120, -120), complex(120, 120))],
+        )
+        # NB: The range of the first parameter ("v") are limited by an overflow
+        # in the intermediate calculations. Can be fixed by implementing an
+        # asymptotic expansion for Bessel functions for large order.
+
+    def test_hyp1f1(self):
+        def mpmath_hyp1f1(a, b, x):
+            try:
+                return mpmath.hyp1f1(a, b, x)
+            except ZeroDivisionError:
+                return np.inf
+
+        assert_mpmath_equal(
+            sc.hyp1f1,
+            mpmath_hyp1f1,
+            [Arg(-50, 50), Arg(1, 50, inclusive_a=False), Arg(-50, 50)],
+            n=500,
+            nan_ok=False,
+        )
+
+    @pytest.mark.xfail(run=False)
+    def test_hyp1f1_complex(self):
+        assert_mpmath_equal(
+            inf_to_nan(lambda a, b, x: sc.hyp1f1(a.real, b.real, x)),
+            exception_to_nan(lambda a, b, x: mpmath.hyp1f1(a, b, x, **HYPERKW)),
+            [Arg(-1e3, 1e3), Arg(-1e3, 1e3), ComplexArg()],
+            n=2000,
+        )
+
+    @nonfunctional_tooslow
+    def test_hyp2f1_complex(self):
+        # SciPy's hyp2f1 seems to have performance and accuracy problems
+        assert_mpmath_equal(
+            lambda a, b, c, x: sc.hyp2f1(a.real, b.real, c.real, x),
+            exception_to_nan(lambda a, b, c, x: mpmath.hyp2f1(a, b, c, x, **HYPERKW)),
+            [Arg(-1e2, 1e2), Arg(-1e2, 1e2), Arg(-1e2, 1e2), ComplexArg()],
+            n=10,
+        )
+
+    @pytest.mark.xfail(run=False)
+    def test_hyperu(self):
+        assert_mpmath_equal(
+            sc.hyperu,
+            exception_to_nan(lambda a, b, x: mpmath.hyperu(a, b, x, **HYPERKW)),
+            [Arg(), Arg(), Arg()],
+        )
+
+    @pytest.mark.xfail_on_32bit("mpmath issue gh-342: "
+                                "unsupported operand mpz, long for pow")
+    def test_igam_fac(self):
+        def mp_igam_fac(a, x):
+            return mpmath.power(x, a)*mpmath.exp(-x)/mpmath.gamma(a)
+
+        assert_mpmath_equal(
+            _igam_fac,
+            mp_igam_fac,
+            [Arg(0, 1e14, inclusive_a=False), Arg(0, 1e14)],
+            rtol=1e-10,
+        )
+
+    def test_j0(self):
+        # The Bessel function at large arguments is j0(x) ~ cos(x + phi)/sqrt(x)
+        # and at large arguments the phase of the cosine loses precision.
+        #
+        # This is numerically expected behavior, so we compare only up to
+        # 1e8 = 1e15 * 1e-7
+        assert_mpmath_equal(sc.j0, mpmath.j0, [Arg(-1e3, 1e3)])
+        assert_mpmath_equal(sc.j0, mpmath.j0, [Arg(-1e8, 1e8)], rtol=1e-5)
+
+    def test_j1(self):
+        # See comment in test_j0
+        assert_mpmath_equal(sc.j1, mpmath.j1, [Arg(-1e3, 1e3)])
+        assert_mpmath_equal(sc.j1, mpmath.j1, [Arg(-1e8, 1e8)], rtol=1e-5)
+
+    @pytest.mark.xfail(run=False)
+    def test_jacobi(self):
+        assert_mpmath_equal(
+            sc.eval_jacobi,
+            exception_to_nan(lambda a, b, c, x: mpmath.jacobi(a, b, c, x, **HYPERKW)),
+            [Arg(), Arg(), Arg(), Arg()],
+        )
+        assert_mpmath_equal(
+            lambda n, b, c, x: sc.eval_jacobi(int(n), b, c, x),
+            exception_to_nan(lambda a, b, c, x: mpmath.jacobi(a, b, c, x, **HYPERKW)),
+            [IntArg(), Arg(), Arg(), Arg()],
+        )
+
+    def test_jacobi_int(self):
+        # Redefine functions to deal with numerical + mpmath issues
+        def jacobi(n, a, b, x):
+            # Mpmath does not handle n=0 case always correctly
+            if n == 0:
+                return 1.0
+            return mpmath.jacobi(n, a, b, x)
+        assert_mpmath_equal(
+            lambda n, a, b, x: sc.eval_jacobi(int(n), a, b, x),
+            lambda n, a, b, x: exception_to_nan(jacobi)(n, a, b, x, **HYPERKW),
+            [IntArg(), Arg(), Arg(), Arg()],
+            n=20000,
+            dps=50,
+        )
+
+    def test_kei(self):
+        def kei(x):
+            if x == 0:
+                # work around mpmath issue at x=0
+                return -pi/4
+            return exception_to_nan(mpmath.kei)(0, x, **HYPERKW)
+        assert_mpmath_equal(sc.kei, kei, [Arg(-1e30, 1e30)], n=1000)
+
+    def test_ker(self):
+        assert_mpmath_equal(
+            sc.ker,
+            exception_to_nan(lambda x: mpmath.ker(0, x, **HYPERKW)),
+            [Arg(-1e30, 1e30)],
+            n=1000,
+        )
+
+    @nonfunctional_tooslow
+    def test_laguerre(self):
+        assert_mpmath_equal(
+            trace_args(sc.eval_laguerre),
+            lambda n, x: exception_to_nan(mpmath.laguerre)(n, x, **HYPERKW),
+            [Arg(), Arg()],
+        )
+
+    def test_laguerre_int(self):
+        assert_mpmath_equal(
+            lambda n, x: sc.eval_laguerre(int(n), x),
+            lambda n, x: exception_to_nan(mpmath.laguerre)(n, x, **HYPERKW),
+            [IntArg(), Arg()],
+            n=20000,
+        )
+
+    @pytest.mark.xfail_on_32bit("see gh-3551 for bad points")
+    def test_lambertw_real(self):
+        assert_mpmath_equal(
+            lambda x, k: sc.lambertw(x, int(k.real)),
+            lambda x, k: mpmath.lambertw(x, int(k.real)),
+            [ComplexArg(-np.inf, np.inf), IntArg(0, 10)],
+            rtol=1e-13, nan_ok=False,
+        )
+
+    def test_lanczos_sum_expg_scaled(self):
+        maxgamma = 171.624376956302725
+        e = np.exp(1)
+        g = 6.024680040776729583740234375
+
+        def gamma(x):
+            with np.errstate(over='ignore'):
+                fac = ((x + g - 0.5)/e)**(x - 0.5)
+                if fac != np.inf:
+                    res = fac*_lanczos_sum_expg_scaled(x)
+                else:
+                    fac = ((x + g - 0.5)/e)**(0.5*(x - 0.5))
+                    res = fac*_lanczos_sum_expg_scaled(x)
+                    res *= fac
+            return res
+
+        assert_mpmath_equal(
+            gamma,
+            mpmath.gamma,
+            [Arg(0, maxgamma, inclusive_a=False)],
+            rtol=1e-13,
+        )
+
+    @nonfunctional_tooslow
+    def test_legendre(self):
+        assert_mpmath_equal(sc.eval_legendre, mpmath.legendre, [Arg(), Arg()])
+
+    def test_legendre_int(self):
+        assert_mpmath_equal(
+            lambda n, x: sc.eval_legendre(int(n), x),
+            lambda n, x: exception_to_nan(mpmath.legendre)(n, x, **HYPERKW),
+            [IntArg(), Arg()],
+            n=20000,
+        )
+
+        # Check the small-x expansion
+        assert_mpmath_equal(
+            lambda n, x: sc.eval_legendre(int(n), x),
+            lambda n, x: exception_to_nan(mpmath.legendre)(n, x, **HYPERKW),
+            [IntArg(), FixedArg(np.logspace(-30, -4, 20))],
+        )
+
+    def test_legenp(self):
+        def lpnm(n, m, z):
+            try:
+                v = sc.lpmn(m, n, z)[0][-1,-1]
+            except ValueError:
+                return np.nan
+            if abs(v) > 1e306:
+                # harmonize overflow to inf
+                v = np.inf * np.sign(v.real)
+            return v
+
+        def lpnm_2(n, m, z):
+            v = sc.lpmv(m, n, z)
+            if abs(v) > 1e306:
+                # harmonize overflow to inf
+                v = np.inf * np.sign(v.real)
+            return v
+
+        def legenp(n, m, z):
+            if (z == 1 or z == -1) and int(n) == n:
+                # Special case (mpmath may give inf, we take the limit by
+                # continuity)
+                if m == 0:
+                    if n < 0:
+                        n = -n - 1
+                    return mpmath.power(mpmath.sign(z), n)
+                else:
+                    return 0
+
+            if abs(z) < 1e-15:
+                # mpmath has bad performance here
+                return np.nan
+
+            typ = 2 if abs(z) < 1 else 3
+            v = exception_to_nan(mpmath.legenp)(n, m, z, type=typ)
+
+            if abs(v) > 1e306:
+                # harmonize overflow to inf
+                v = mpmath.inf * mpmath.sign(v.real)
+
+            return v
+
+        assert_mpmath_equal(lpnm, legenp, [IntArg(-100, 100), IntArg(-100, 100), Arg()])
+
+        assert_mpmath_equal(
+            lpnm_2,
+            legenp,
+            [IntArg(-100, 100), Arg(-100, 100), Arg(-1, 1)],
+            atol=1e-10,
+        )
+
+    def test_legenp_complex_2(self):
+        def clpnm(n, m, z):
+            try:
+                return sc.clpmn(m.real, n.real, z, type=2)[0][-1,-1]
+            except ValueError:
+                return np.nan
+
+        def legenp(n, m, z):
+            if abs(z) < 1e-15:
+                # mpmath has bad performance here
+                return np.nan
+            return exception_to_nan(mpmath.legenp)(int(n.real), int(m.real), z, type=2)
+
+        # mpmath is quite slow here
+        x = np.array([-2, -0.99, -0.5, 0, 1e-5, 0.5, 0.99, 20, 2e3])
+        y = np.array([-1e3, -0.5, 0.5, 1.3])
+        z = (x[:,None] + 1j*y[None,:]).ravel()
+
+        assert_mpmath_equal(
+            clpnm,
+            legenp,
+            [FixedArg([-2, -1, 0, 1, 2, 10]),
+             FixedArg([-2, -1, 0, 1, 2, 10]),
+             FixedArg(z)],
+            rtol=1e-6,
+            n=500,
+        )
+
+    def test_legenp_complex_3(self):
+        def clpnm(n, m, z):
+            try:
+                return sc.clpmn(m.real, n.real, z, type=3)[0][-1,-1]
+            except ValueError:
+                return np.nan
+
+        def legenp(n, m, z):
+            if abs(z) < 1e-15:
+                # mpmath has bad performance here
+                return np.nan
+            return exception_to_nan(mpmath.legenp)(int(n.real), int(m.real), z, type=3)
+
+        # mpmath is quite slow here
+        x = np.array([-2, -0.99, -0.5, 0, 1e-5, 0.5, 0.99, 20, 2e3])
+        y = np.array([-1e3, -0.5, 0.5, 1.3])
+        z = (x[:,None] + 1j*y[None,:]).ravel()
+
+        assert_mpmath_equal(
+            clpnm,
+            legenp,
+            [FixedArg([-2, -1, 0, 1, 2, 10]),
+             FixedArg([-2, -1, 0, 1, 2, 10]),
+             FixedArg(z)],
+            rtol=1e-6,
+            n=500,
+        )
+
+    @pytest.mark.xfail(run=False, reason="apparently picks wrong function at |z| > 1")
+    def test_legenq(self):
+        def lqnm(n, m, z):
+            return sc.lqmn(m, n, z)[0][-1,-1]
+
+        def legenq(n, m, z):
+            if abs(z) < 1e-15:
+                # mpmath has bad performance here
+                return np.nan
+            return exception_to_nan(mpmath.legenq)(n, m, z, type=2)
+
+        assert_mpmath_equal(
+            lqnm,
+            legenq,
+            [IntArg(0, 100), IntArg(0, 100), Arg()],
+        )
+
+    @nonfunctional_tooslow
+    def test_legenq_complex(self):
+        def lqnm(n, m, z):
+            return sc.lqmn(int(m.real), int(n.real), z)[0][-1,-1]
+
+        def legenq(n, m, z):
+            if abs(z) < 1e-15:
+                # mpmath has bad performance here
+                return np.nan
+            return exception_to_nan(mpmath.legenq)(int(n.real), int(m.real), z, type=2)
+
+        assert_mpmath_equal(
+            lqnm,
+            legenq,
+            [IntArg(0, 100), IntArg(0, 100), ComplexArg()],
+            n=100,
+        )
+
+    def test_lgam1p(self):
+        def param_filter(x):
+            # Filter the poles
+            return np.where((np.floor(x) == x) & (x <= 0), False, True)
+
+        def mp_lgam1p(z):
+            # The real part of loggamma is log(|gamma(z)|)
+            return mpmath.loggamma(1 + z).real
+
+        assert_mpmath_equal(
+            _lgam1p,
+            mp_lgam1p,
+            [Arg()],
+            rtol=1e-13,
+            dps=100,
+            param_filter=param_filter,
+        )
+
+    def test_loggamma(self):
+        def mpmath_loggamma(z):
+            try:
+                res = mpmath.loggamma(z)
+            except ValueError:
+                res = complex(np.nan, np.nan)
+            return res
+
+        assert_mpmath_equal(
+            sc.loggamma,
+            mpmath_loggamma,
+            [ComplexArg()],
+            nan_ok=False,
+            distinguish_nan_and_inf=False,
+            rtol=5e-14,
+        )
+
+    @pytest.mark.xfail(run=False)
+    def test_pcfd(self):
+        def pcfd(v, x):
+            return sc.pbdv(v, x)[0]
+        assert_mpmath_equal(
+            pcfd,
+            exception_to_nan(lambda v, x: mpmath.pcfd(v, x, **HYPERKW)),
+            [Arg(), Arg()],
+        )
+
+    @pytest.mark.xfail(run=False, reason="it's not the same as the mpmath function --- "
+                                         "maybe different definition?")
+    def test_pcfv(self):
+        def pcfv(v, x):
+            return sc.pbvv(v, x)[0]
+        assert_mpmath_equal(
+            pcfv,
+            lambda v, x: time_limited()(exception_to_nan(mpmath.pcfv))(v, x, **HYPERKW),
+            [Arg(), Arg()],
+            n=1000,
+        )
+
+    def test_pcfw(self):
+        def pcfw(a, x):
+            return sc.pbwa(a, x)[0]
+
+        def dpcfw(a, x):
+            return sc.pbwa(a, x)[1]
+
+        def mpmath_dpcfw(a, x):
+            return mpmath.diff(mpmath.pcfw, (a, x), (0, 1))
+
+        # The Zhang and Jin implementation only uses Taylor series and
+        # is thus accurate in only a very small range.
+        assert_mpmath_equal(
+            pcfw,
+            mpmath.pcfw,
+            [Arg(-5, 5), Arg(-5, 5)],
+            rtol=2e-8,
+            n=100,
+        )
+
+        assert_mpmath_equal(
+            dpcfw,
+            mpmath_dpcfw,
+            [Arg(-5, 5), Arg(-5, 5)],
+            rtol=2e-9,
+            n=100,
+        )
+
+    @pytest.mark.xfail(run=False,
+                       reason="issues at large arguments (atol OK, rtol not) "
+                              "and <eps-close to z=0")
+    def test_polygamma(self):
+        assert_mpmath_equal(
+            sc.polygamma,
+            time_limited()(exception_to_nan(mpmath.polygamma)),
+            [IntArg(0, 1000), Arg()],
+        )
+
+    def test_rgamma(self):
+        assert_mpmath_equal(
+            sc.rgamma,
+            mpmath.rgamma,
+            [Arg(-8000, np.inf)],
+            n=5000,
+            nan_ok=False,
+            ignore_inf_sign=True,
+        )
+
+    def test_rgamma_complex(self):
+        assert_mpmath_equal(
+            sc.rgamma,
+            exception_to_nan(mpmath.rgamma),
+            [ComplexArg()],
+            rtol=5e-13,
+        )
+
+    @pytest.mark.xfail(reason=("see gh-3551 for bad points on 32 bit "
+                               "systems and gh-8095 for another bad "
+                               "point"))
+    def test_rf(self):
+        if _pep440.parse(mpmath.__version__) >= _pep440.Version("1.0.0"):
+            # no workarounds needed
+            mppoch = mpmath.rf
+        else:
+            def mppoch(a, m):
+                # deal with cases where the result in double precision
+                # hits exactly a non-positive integer, but the
+                # corresponding extended-precision mpf floats don't
+                if float(a + m) == int(a + m) and float(a + m) <= 0:
+                    a = mpmath.mpf(a)
+                    m = int(a + m) - a
+                return mpmath.rf(a, m)
+
+        assert_mpmath_equal(sc.poch, mppoch, [Arg(), Arg()], dps=400)
+
+    def test_sinpi(self):
+        eps = np.finfo(float).eps
+        assert_mpmath_equal(
+            _sinpi,
+            mpmath.sinpi,
+            [Arg()],
+            nan_ok=False,
+            rtol=2*eps,
+        )
+
+    def test_sinpi_complex(self):
+        assert_mpmath_equal(
+            _sinpi,
+            mpmath.sinpi,
+            [ComplexArg()],
+            nan_ok=False,
+            rtol=2e-14,
+        )
+
+    def test_shi(self):
+        def shi(x):
+            return sc.shichi(x)[0]
+        assert_mpmath_equal(shi, mpmath.shi, [Arg()])
+        # check asymptotic series cross-over
+        assert_mpmath_equal(shi, mpmath.shi, [FixedArg([88 - 1e-9, 88, 88 + 1e-9])])
+
+    def test_shi_complex(self):
+        def shi(z):
+            return sc.shichi(z)[0]
+        # shi oscillates as Im[z] -> +- inf, so limit range
+        assert_mpmath_equal(
+            shi,
+            mpmath.shi,
+            [ComplexArg(complex(-np.inf, -1e8), complex(np.inf, 1e8))],
+            rtol=1e-12,
+        )
+
+    def test_si(self):
+        def si(x):
+            return sc.sici(x)[0]
+        assert_mpmath_equal(si, mpmath.si, [Arg()])
+
+    def test_si_complex(self):
+        def si(z):
+            return sc.sici(z)[0]
+        # si oscillates as Re[z] -> +- inf, so limit range
+        assert_mpmath_equal(
+            si,
+            mpmath.si,
+            [ComplexArg(complex(-1e8, -np.inf), complex(1e8, np.inf))],
+            rtol=1e-12,
+        )
+
+    def test_spence(self):
+        # mpmath uses a different convention for the dilogarithm
+        def dilog(x):
+            return mpmath.polylog(2, 1 - x)
+        # Spence has a branch cut on the negative real axis
+        assert_mpmath_equal(
+            sc.spence,
+            exception_to_nan(dilog),
+            [Arg(0, np.inf)],
+            rtol=1e-14,
+        )
+
+    def test_spence_complex(self):
+        def dilog(z):
+            return mpmath.polylog(2, 1 - z)
+        assert_mpmath_equal(
+            sc.spence,
+            exception_to_nan(dilog),
+            [ComplexArg()],
+            rtol=1e-14,
+        )
+
+    def test_spherharm(self):
+        def spherharm(l, m, theta, phi):
+            if m > l:
+                return np.nan
+            return sc.sph_harm(m, l, phi, theta)
+        assert_mpmath_equal(
+            spherharm,
+            mpmath.spherharm,
+            [IntArg(0, 100), IntArg(0, 100), Arg(a=0, b=pi), Arg(a=0, b=2*pi)],
+            atol=1e-8,
+            n=6000,
+            dps=150,
+        )
+
+    def test_struveh(self):
+        assert_mpmath_equal(
+            sc.struve,
+            exception_to_nan(mpmath.struveh),
+            [Arg(-1e4, 1e4), Arg(0, 1e4)],
+            rtol=5e-10,
+        )
+
+    def test_struvel(self):
+        def mp_struvel(v, z):
+            if v < 0 and z < -v and abs(v) > 1000:
+                # larger DPS needed for correct results
+                old_dps = mpmath.mp.dps
+                try:
+                    mpmath.mp.dps = 300
+                    return mpmath.struvel(v, z)
+                finally:
+                    mpmath.mp.dps = old_dps
+            return mpmath.struvel(v, z)
+
+        assert_mpmath_equal(
+            sc.modstruve,
+            exception_to_nan(mp_struvel),
+            [Arg(-1e4, 1e4), Arg(0, 1e4)],
+            rtol=5e-10,
+            ignore_inf_sign=True,
+        )
+
+    def test_wrightomega_real(self):
+        def mpmath_wrightomega_real(x):
+            return mpmath.lambertw(mpmath.exp(x), mpmath.mpf('-0.5'))
+
+        # For x < -1000 the Wright Omega function is just 0 to double
+        # precision, and for x > 1e21 it is just x to double
+        # precision.
+        assert_mpmath_equal(
+            sc.wrightomega,
+            mpmath_wrightomega_real,
+            [Arg(-1000, 1e21)],
+            rtol=5e-15,
+            atol=0,
+            nan_ok=False,
+        )
+
+    def test_wrightomega(self):
+        assert_mpmath_equal(
+            sc.wrightomega,
+            lambda z: _mpmath_wrightomega(z, 25),
+            [ComplexArg()],
+            rtol=1e-14,
+            nan_ok=False,
+        )
+
+    def test_hurwitz_zeta(self):
+        assert_mpmath_equal(
+            sc.zeta,
+            exception_to_nan(mpmath.zeta),
+            [Arg(a=1, b=1e10, inclusive_a=False), Arg(a=0, inclusive_a=False)],
+        )
+
+    def test_riemann_zeta(self):
+        assert_mpmath_equal(
+            sc.zeta,
+            lambda x: mpmath.zeta(x) if x != 1 else mpmath.inf,
+            [Arg(-100, 100)],
+            nan_ok=False,
+            rtol=5e-13,
+        )
+
+    def test_zetac(self):
+        assert_mpmath_equal(
+            sc.zetac,
+            lambda x: mpmath.zeta(x) - 1 if x != 1 else mpmath.inf,
+            [Arg(-100, 100)],
+            nan_ok=False,
+            dps=45,
+            rtol=5e-13,
+        )
+
+    def test_boxcox(self):
+
+        def mp_boxcox(x, lmbda):
+            x = mpmath.mp.mpf(x)
+            lmbda = mpmath.mp.mpf(lmbda)
+            if lmbda == 0:
+                return mpmath.mp.log(x)
+            else:
+                return mpmath.mp.powm1(x, lmbda) / lmbda
+
+        assert_mpmath_equal(
+            sc.boxcox,
+            exception_to_nan(mp_boxcox),
+            [Arg(a=0, inclusive_a=False), Arg()],
+            n=200,
+            dps=60,
+            rtol=1e-13,
+        )
+
+    def test_boxcox1p(self):
+
+        def mp_boxcox1p(x, lmbda):
+            x = mpmath.mp.mpf(x)
+            lmbda = mpmath.mp.mpf(lmbda)
+            one = mpmath.mp.mpf(1)
+            if lmbda == 0:
+                return mpmath.mp.log(one + x)
+            else:
+                return mpmath.mp.powm1(one + x, lmbda) / lmbda
+
+        assert_mpmath_equal(
+            sc.boxcox1p,
+            exception_to_nan(mp_boxcox1p),
+            [Arg(a=-1, inclusive_a=False), Arg()],
+            n=200,
+            dps=60,
+            rtol=1e-13,
+        )
+
+    def test_spherical_jn(self):
+        def mp_spherical_jn(n, z):
+            arg = mpmath.mpmathify(z)
+            out = (mpmath.besselj(n + mpmath.mpf(1)/2, arg) /
+                   mpmath.sqrt(2*arg/mpmath.pi))
+            if arg.imag == 0:
+                return out.real
+            else:
+                return out
+
+        assert_mpmath_equal(
+            lambda n, z: sc.spherical_jn(int(n), z),
+            exception_to_nan(mp_spherical_jn),
+            [IntArg(0, 200), Arg(-1e8, 1e8)],
+            dps=300,
+        )
+
+    def test_spherical_jn_complex(self):
+        def mp_spherical_jn(n, z):
+            arg = mpmath.mpmathify(z)
+            out = (mpmath.besselj(n + mpmath.mpf(1)/2, arg) /
+                   mpmath.sqrt(2*arg/mpmath.pi))
+            if arg.imag == 0:
+                return out.real
+            else:
+                return out
+
+        assert_mpmath_equal(
+            lambda n, z: sc.spherical_jn(int(n.real), z),
+            exception_to_nan(mp_spherical_jn),
+            [IntArg(0, 200), ComplexArg()]
+        )
+
+    def test_spherical_yn(self):
+        def mp_spherical_yn(n, z):
+            arg = mpmath.mpmathify(z)
+            out = (mpmath.bessely(n + mpmath.mpf(1)/2, arg) /
+                   mpmath.sqrt(2*arg/mpmath.pi))
+            if arg.imag == 0:
+                return out.real
+            else:
+                return out
+
+        assert_mpmath_equal(
+            lambda n, z: sc.spherical_yn(int(n), z),
+            exception_to_nan(mp_spherical_yn),
+            [IntArg(0, 200), Arg(-1e10, 1e10)],
+            dps=100,
+        )
+
+    def test_spherical_yn_complex(self):
+        def mp_spherical_yn(n, z):
+            arg = mpmath.mpmathify(z)
+            out = (mpmath.bessely(n + mpmath.mpf(1)/2, arg) /
+                   mpmath.sqrt(2*arg/mpmath.pi))
+            if arg.imag == 0:
+                return out.real
+            else:
+                return out
+
+        assert_mpmath_equal(
+            lambda n, z: sc.spherical_yn(int(n.real), z),
+            exception_to_nan(mp_spherical_yn),
+            [IntArg(0, 200), ComplexArg()],
+        )
+
+    def test_spherical_in(self):
+        def mp_spherical_in(n, z):
+            arg = mpmath.mpmathify(z)
+            out = (mpmath.besseli(n + mpmath.mpf(1)/2, arg) /
+                   mpmath.sqrt(2*arg/mpmath.pi))
+            if arg.imag == 0:
+                return out.real
+            else:
+                return out
+
+        assert_mpmath_equal(
+            lambda n, z: sc.spherical_in(int(n), z),
+            exception_to_nan(mp_spherical_in),
+            [IntArg(0, 200), Arg()],
+            dps=200,
+            atol=10**(-278),
+        )
+
+    def test_spherical_in_complex(self):
+        def mp_spherical_in(n, z):
+            arg = mpmath.mpmathify(z)
+            out = (mpmath.besseli(n + mpmath.mpf(1)/2, arg) /
+                   mpmath.sqrt(2*arg/mpmath.pi))
+            if arg.imag == 0:
+                return out.real
+            else:
+                return out
+
+        assert_mpmath_equal(
+            lambda n, z: sc.spherical_in(int(n.real), z),
+            exception_to_nan(mp_spherical_in),
+            [IntArg(0, 200), ComplexArg()],
+        )
+
+    def test_spherical_kn(self):
+        def mp_spherical_kn(n, z):
+            out = (mpmath.besselk(n + mpmath.mpf(1)/2, z) *
+                   mpmath.sqrt(mpmath.pi/(2*mpmath.mpmathify(z))))
+            if mpmath.mpmathify(z).imag == 0:
+                return out.real
+            else:
+                return out
+
+        assert_mpmath_equal(
+            lambda n, z: sc.spherical_kn(int(n), z),
+            exception_to_nan(mp_spherical_kn),
+            [IntArg(0, 150), Arg()],
+            dps=100,
+        )
+
+    @pytest.mark.xfail(run=False,
+                       reason="Accuracy issues near z = -1 inherited from kv.")
+    def test_spherical_kn_complex(self):
+        def mp_spherical_kn(n, z):
+            arg = mpmath.mpmathify(z)
+            out = (mpmath.besselk(n + mpmath.mpf(1)/2, arg) /
+                   mpmath.sqrt(2*arg/mpmath.pi))
+            if arg.imag == 0:
+                return out.real
+            else:
+                return out
+
+        assert_mpmath_equal(
+            lambda n, z: sc.spherical_kn(int(n.real), z),
+            exception_to_nan(mp_spherical_kn),
+            [IntArg(0, 200), ComplexArg()],
+            dps=200,
+        )

.venv/Lib/site-packages/scipy/special/tests/test_nan_inputs.py

@@ -0,0 +1,64 @@
+"""Test how the ufuncs in special handle nan inputs.
+
+"""
+from typing import Callable
+
+import numpy as np
+from numpy.testing import assert_array_equal, assert_, suppress_warnings
+import pytest
+import scipy.special as sc
+
+
+KNOWNFAILURES: dict[str, Callable] = {}
+
+POSTPROCESSING: dict[str, Callable] = {}
+
+
+def _get_ufuncs():
+    ufuncs = []
+    ufunc_names = []
+    for name in sorted(sc.__dict__):
+        obj = sc.__dict__[name]
+        if not isinstance(obj, np.ufunc):
+            continue
+        msg = KNOWNFAILURES.get(obj)
+        if msg is None:
+            ufuncs.append(obj)
+            ufunc_names.append(name)
+        else:
+            fail = pytest.mark.xfail(run=False, reason=msg)
+            ufuncs.append(pytest.param(obj, marks=fail))
+            ufunc_names.append(name)
+    return ufuncs, ufunc_names
+
+
+UFUNCS, UFUNC_NAMES = _get_ufuncs()
+
+
+@pytest.mark.parametrize("func", UFUNCS, ids=UFUNC_NAMES)
+def test_nan_inputs(func):
+    args = (np.nan,)*func.nin
+    with suppress_warnings() as sup:
+        # Ignore warnings about unsafe casts from legacy wrappers
+        sup.filter(RuntimeWarning,
+                   "floating point number truncated to an integer")
+        try:
+            with suppress_warnings() as sup:
+                sup.filter(DeprecationWarning)
+                res = func(*args)
+        except TypeError:
+            # One of the arguments doesn't take real inputs
+            return
+    if func in POSTPROCESSING:
+        res = POSTPROCESSING[func](*res)
+
+    msg = f"got {res} instead of nan"
+    assert_array_equal(np.isnan(res), True, err_msg=msg)
+
+
+def test_legacy_cast():
+    with suppress_warnings() as sup:
+        sup.filter(RuntimeWarning,
+                   "floating point number truncated to an integer")
+        res = sc.bdtrc(np.nan, 1, 0.5)
+        assert_(np.isnan(res))

.venv/Lib/site-packages/scipy/special/tests/test_ndtr.py

@@ -0,0 +1,77 @@
+import numpy as np
+from numpy.testing import assert_equal, assert_allclose
+import scipy.special as sc
+
+
+def test_ndtr():
+    assert_equal(sc.ndtr(0), 0.5)
+    assert_allclose(sc.ndtr(1), 0.8413447460685429)
+
+
+class TestNdtri:
+
+    def test_zero(self):
+        assert sc.ndtri(0.5) == 0.0
+
+    def test_asymptotes(self):
+        assert_equal(sc.ndtri([0.0, 1.0]), [-np.inf, np.inf])
+
+    def test_outside_of_domain(self):
+        assert all(np.isnan(sc.ndtri([-1.5, 1.5])))
+
+
+class TestLogNdtr:
+
+    # The expected values in these tests were computed with mpmath:
+    #
+    #   def log_ndtr_mp(x):
+    #       return mpmath.log(mpmath.ncdf(x))
+    #
+
+    def test_log_ndtr_moderate_le8(self):
+        x = np.array([-0.75, -0.25, 0, 0.5, 1.5, 2.5, 3, 4, 5, 7, 8])
+        expected = np.array([-1.4844482299196562,
+                             -0.9130617648111351,
+                             -0.6931471805599453,
+                             -0.3689464152886564,
+                             -0.06914345561223398,
+                             -0.006229025485860002,
+                             -0.0013508099647481938,
+                             -3.167174337748927e-05,
+                             -2.866516129637636e-07,
+                             -1.279812543886654e-12,
+                             -6.220960574271786e-16])
+        y = sc.log_ndtr(x)
+        assert_allclose(y, expected, rtol=1e-14)
+
+    def test_log_ndtr_values_8_16(self):
+        x = np.array([8.001, 8.06, 8.15, 8.5, 10, 12, 14, 16])
+        expected = [-6.170639424817055e-16,
+                    -3.814722443652823e-16,
+                    -1.819621363526629e-16,
+                    -9.479534822203318e-18,
+                    -7.619853024160525e-24,
+                    -1.776482112077679e-33,
+                    -7.7935368191928e-45,
+                    -6.388754400538087e-58]
+        y = sc.log_ndtr(x)
+        assert_allclose(y, expected, rtol=5e-14)
+
+    def test_log_ndtr_values_16_31(self):
+        x = np.array([16.15, 20.3, 21.4, 26.2, 30.9])
+        expected = [-5.678084565148492e-59,
+                    -6.429244467698346e-92,
+                    -6.680402412553295e-102,
+                    -1.328698078458869e-151,
+                    -5.972288641838264e-210]
+        y = sc.log_ndtr(x)
+        assert_allclose(y, expected, rtol=2e-13)
+
+    def test_log_ndtr_values_gt31(self):
+        x = np.array([31.6, 32.8, 34.9, 37.1])
+        expected = [-1.846036234858162e-219,
+                    -2.9440539964066835e-236,
+                    -3.71721649450857e-267,
+                    -1.4047119663106221e-301]
+        y = sc.log_ndtr(x)
+        assert_allclose(y, expected, rtol=3e-13)

.venv/Lib/site-packages/scipy/special/tests/test_ndtri_exp.py

@@ -0,0 +1,94 @@
+import pytest
+import numpy as np
+from numpy.testing import assert_equal, assert_allclose
+from scipy.special import log_ndtr, ndtri_exp
+from scipy.special._testutils import assert_func_equal
+
+
+def log_ndtr_ndtri_exp(y):
+    return log_ndtr(ndtri_exp(y))
+
+
+@pytest.fixture(scope="class")
+def uniform_random_points():
+    random_state = np.random.RandomState(1234)
+    points = random_state.random_sample(1000)
+    return points
+
+
+class TestNdtriExp:
+    """Tests that ndtri_exp is sufficiently close to an inverse of log_ndtr.
+
+    We have separate tests for the five intervals (-inf, -10),
+    [-10, -2), [-2, -0.14542), [-0.14542, -1e-6), and [-1e-6, 0).
+    ndtri_exp(y) is computed in three different ways depending on if y
+    is in (-inf, -2), [-2, log(1 - exp(-2))], or [log(1 - exp(-2), 0).
+    Each of these intervals is given its own test with two additional tests
+    for handling very small values and values very close to zero.
+    """
+
+    @pytest.mark.parametrize(
+        "test_input", [-1e1, -1e2, -1e10, -1e20, -np.finfo(float).max]
+    )
+    def test_very_small_arg(self, test_input, uniform_random_points):
+        scale = test_input
+        points = scale * (0.5 * uniform_random_points + 0.5)
+        assert_func_equal(
+            log_ndtr_ndtri_exp,
+            lambda y: y, points,
+            rtol=1e-14,
+            nan_ok=True
+        )
+
+    @pytest.mark.parametrize(
+        "interval,expected_rtol",
+        [
+            ((-10, -2), 1e-14),
+            ((-2, -0.14542), 1e-12),
+            ((-0.14542, -1e-6), 1e-10),
+            ((-1e-6, 0), 1e-6),
+        ],
+    )
+    def test_in_interval(self, interval, expected_rtol, uniform_random_points):
+        left, right = interval
+        points = (right - left) * uniform_random_points + left
+        assert_func_equal(
+            log_ndtr_ndtri_exp,
+            lambda y: y, points,
+            rtol=expected_rtol,
+            nan_ok=True
+        )
+
+    def test_extreme(self):
+        # bigneg is not quite the largest negative double precision value.
+        # Here's why:
+        # The round-trip calculation
+        #    y = ndtri_exp(bigneg)
+        #    bigneg2 = log_ndtr(y)
+        # where bigneg is a very large negative value, would--with infinite
+        # precision--result in bigneg2 == bigneg.  When bigneg is large enough,
+        # y is effectively equal to -sqrt(2)*sqrt(-bigneg), and log_ndtr(y) is
+        # effectively -(y/sqrt(2))**2.  If we use bigneg = np.finfo(float).min,
+        # then by construction, the theoretical value is the most negative
+        # finite value that can be represented with 64 bit float point.  This
+        # means tiny changes in how the computation proceeds can result in the
+        # return value being -inf.  (E.g. changing the constant representation
+        # of 1/sqrt(2) from 0.7071067811865475--which is the value returned by
+        # 1/np.sqrt(2)--to 0.7071067811865476--which is the most accurate 64
+        # bit floating point representation of 1/sqrt(2)--results in the
+        # round-trip that starts with np.finfo(float).min returning -inf.  So
+        # we'll move the bigneg value a few ULPs towards 0 to avoid this
+        # sensitivity.
+        # Use the reduce method to apply nextafter four times.
+        bigneg = np.nextafter.reduce([np.finfo(float).min, 0, 0, 0, 0])
+        # tinyneg is approx. -2.225e-308.
+        tinyneg = -np.finfo(float).tiny
+        x = np.array([tinyneg, bigneg])
+        result = log_ndtr_ndtri_exp(x)
+        assert_allclose(result, x, rtol=1e-12)
+
+    def test_asymptotes(self):
+        assert_equal(ndtri_exp([-np.inf, 0.0]), [-np.inf, np.inf])
+
+    def test_outside_domain(self):
+        assert np.isnan(ndtri_exp(1.0))

.venv/Lib/site-packages/scipy/special/tests/test_orthogonal.py

@@ -0,0 +1,804 @@
+import numpy as np
+from numpy import array, sqrt
+from numpy.testing import (assert_array_almost_equal, assert_equal,
+                           assert_almost_equal, assert_allclose)
+from pytest import raises as assert_raises
+
+from scipy import integrate
+import scipy.special as sc
+from scipy.special import gamma
+import scipy.special._orthogonal as orth
+
+
+class TestCheby:
+    def test_chebyc(self):
+        C0 = orth.chebyc(0)
+        C1 = orth.chebyc(1)
+        with np.errstate(all='ignore'):
+            C2 = orth.chebyc(2)
+            C3 = orth.chebyc(3)
+            C4 = orth.chebyc(4)
+            C5 = orth.chebyc(5)
+
+        assert_array_almost_equal(C0.c,[2],13)
+        assert_array_almost_equal(C1.c,[1,0],13)
+        assert_array_almost_equal(C2.c,[1,0,-2],13)
+        assert_array_almost_equal(C3.c,[1,0,-3,0],13)
+        assert_array_almost_equal(C4.c,[1,0,-4,0,2],13)
+        assert_array_almost_equal(C5.c,[1,0,-5,0,5,0],13)
+
+    def test_chebys(self):
+        S0 = orth.chebys(0)
+        S1 = orth.chebys(1)
+        S2 = orth.chebys(2)
+        S3 = orth.chebys(3)
+        S4 = orth.chebys(4)
+        S5 = orth.chebys(5)
+        assert_array_almost_equal(S0.c,[1],13)
+        assert_array_almost_equal(S1.c,[1,0],13)
+        assert_array_almost_equal(S2.c,[1,0,-1],13)
+        assert_array_almost_equal(S3.c,[1,0,-2,0],13)
+        assert_array_almost_equal(S4.c,[1,0,-3,0,1],13)
+        assert_array_almost_equal(S5.c,[1,0,-4,0,3,0],13)
+
+    def test_chebyt(self):
+        T0 = orth.chebyt(0)
+        T1 = orth.chebyt(1)
+        T2 = orth.chebyt(2)
+        T3 = orth.chebyt(3)
+        T4 = orth.chebyt(4)
+        T5 = orth.chebyt(5)
+        assert_array_almost_equal(T0.c,[1],13)
+        assert_array_almost_equal(T1.c,[1,0],13)
+        assert_array_almost_equal(T2.c,[2,0,-1],13)
+        assert_array_almost_equal(T3.c,[4,0,-3,0],13)
+        assert_array_almost_equal(T4.c,[8,0,-8,0,1],13)
+        assert_array_almost_equal(T5.c,[16,0,-20,0,5,0],13)
+
+    def test_chebyu(self):
+        U0 = orth.chebyu(0)
+        U1 = orth.chebyu(1)
+        U2 = orth.chebyu(2)
+        U3 = orth.chebyu(3)
+        U4 = orth.chebyu(4)
+        U5 = orth.chebyu(5)
+        assert_array_almost_equal(U0.c,[1],13)
+        assert_array_almost_equal(U1.c,[2,0],13)
+        assert_array_almost_equal(U2.c,[4,0,-1],13)
+        assert_array_almost_equal(U3.c,[8,0,-4,0],13)
+        assert_array_almost_equal(U4.c,[16,0,-12,0,1],13)
+        assert_array_almost_equal(U5.c,[32,0,-32,0,6,0],13)
+
+
+class TestGegenbauer:
+
+    def test_gegenbauer(self):
+        a = 5*np.random.random() - 0.5
+        if np.any(a == 0):
+            a = -0.2
+        Ca0 = orth.gegenbauer(0,a)
+        Ca1 = orth.gegenbauer(1,a)
+        Ca2 = orth.gegenbauer(2,a)
+        Ca3 = orth.gegenbauer(3,a)
+        Ca4 = orth.gegenbauer(4,a)
+        Ca5 = orth.gegenbauer(5,a)
+
+        assert_array_almost_equal(Ca0.c,array([1]),13)
+        assert_array_almost_equal(Ca1.c,array([2*a,0]),13)
+        assert_array_almost_equal(Ca2.c,array([2*a*(a+1),0,-a]),13)
+        assert_array_almost_equal(Ca3.c,array([4*sc.poch(a,3),0,-6*a*(a+1),
+                                               0])/3.0,11)
+        assert_array_almost_equal(Ca4.c,array([4*sc.poch(a,4),0,-12*sc.poch(a,3),
+                                               0,3*a*(a+1)])/6.0,11)
+        assert_array_almost_equal(Ca5.c,array([4*sc.poch(a,5),0,-20*sc.poch(a,4),
+                                               0,15*sc.poch(a,3),0])/15.0,11)
+
+
+class TestHermite:
+    def test_hermite(self):
+        H0 = orth.hermite(0)
+        H1 = orth.hermite(1)
+        H2 = orth.hermite(2)
+        H3 = orth.hermite(3)
+        H4 = orth.hermite(4)
+        H5 = orth.hermite(5)
+        assert_array_almost_equal(H0.c,[1],13)
+        assert_array_almost_equal(H1.c,[2,0],13)
+        assert_array_almost_equal(H2.c,[4,0,-2],13)
+        assert_array_almost_equal(H3.c,[8,0,-12,0],13)
+        assert_array_almost_equal(H4.c,[16,0,-48,0,12],12)
+        assert_array_almost_equal(H5.c,[32,0,-160,0,120,0],12)
+
+    def test_hermitenorm(self):
+        # He_n(x) = 2**(-n/2) H_n(x/sqrt(2))
+        psub = np.poly1d([1.0/sqrt(2),0])
+        H0 = orth.hermitenorm(0)
+        H1 = orth.hermitenorm(1)
+        H2 = orth.hermitenorm(2)
+        H3 = orth.hermitenorm(3)
+        H4 = orth.hermitenorm(4)
+        H5 = orth.hermitenorm(5)
+        he0 = orth.hermite(0)(psub)
+        he1 = orth.hermite(1)(psub) / sqrt(2)
+        he2 = orth.hermite(2)(psub) / 2.0
+        he3 = orth.hermite(3)(psub) / (2*sqrt(2))
+        he4 = orth.hermite(4)(psub) / 4.0
+        he5 = orth.hermite(5)(psub) / (4.0*sqrt(2))
+
+        assert_array_almost_equal(H0.c,he0.c,13)
+        assert_array_almost_equal(H1.c,he1.c,13)
+        assert_array_almost_equal(H2.c,he2.c,13)
+        assert_array_almost_equal(H3.c,he3.c,13)
+        assert_array_almost_equal(H4.c,he4.c,13)
+        assert_array_almost_equal(H5.c,he5.c,13)
+
+
+class TestShLegendre:
+    def test_sh_legendre(self):
+        # P*_n(x) = P_n(2x-1)
+        psub = np.poly1d([2,-1])
+        Ps0 = orth.sh_legendre(0)
+        Ps1 = orth.sh_legendre(1)
+        Ps2 = orth.sh_legendre(2)
+        Ps3 = orth.sh_legendre(3)
+        Ps4 = orth.sh_legendre(4)
+        Ps5 = orth.sh_legendre(5)
+        pse0 = orth.legendre(0)(psub)
+        pse1 = orth.legendre(1)(psub)
+        pse2 = orth.legendre(2)(psub)
+        pse3 = orth.legendre(3)(psub)
+        pse4 = orth.legendre(4)(psub)
+        pse5 = orth.legendre(5)(psub)
+        assert_array_almost_equal(Ps0.c,pse0.c,13)
+        assert_array_almost_equal(Ps1.c,pse1.c,13)
+        assert_array_almost_equal(Ps2.c,pse2.c,13)
+        assert_array_almost_equal(Ps3.c,pse3.c,13)
+        assert_array_almost_equal(Ps4.c,pse4.c,12)
+        assert_array_almost_equal(Ps5.c,pse5.c,12)
+
+
+class TestShChebyt:
+    def test_sh_chebyt(self):
+        # T*_n(x) = T_n(2x-1)
+        psub = np.poly1d([2,-1])
+        Ts0 = orth.sh_chebyt(0)
+        Ts1 = orth.sh_chebyt(1)
+        Ts2 = orth.sh_chebyt(2)
+        Ts3 = orth.sh_chebyt(3)
+        Ts4 = orth.sh_chebyt(4)
+        Ts5 = orth.sh_chebyt(5)
+        tse0 = orth.chebyt(0)(psub)
+        tse1 = orth.chebyt(1)(psub)
+        tse2 = orth.chebyt(2)(psub)
+        tse3 = orth.chebyt(3)(psub)
+        tse4 = orth.chebyt(4)(psub)
+        tse5 = orth.chebyt(5)(psub)
+        assert_array_almost_equal(Ts0.c,tse0.c,13)
+        assert_array_almost_equal(Ts1.c,tse1.c,13)
+        assert_array_almost_equal(Ts2.c,tse2.c,13)
+        assert_array_almost_equal(Ts3.c,tse3.c,13)
+        assert_array_almost_equal(Ts4.c,tse4.c,12)
+        assert_array_almost_equal(Ts5.c,tse5.c,12)
+
+
+class TestShChebyu:
+    def test_sh_chebyu(self):
+        # U*_n(x) = U_n(2x-1)
+        psub = np.poly1d([2,-1])
+        Us0 = orth.sh_chebyu(0)
+        Us1 = orth.sh_chebyu(1)
+        Us2 = orth.sh_chebyu(2)
+        Us3 = orth.sh_chebyu(3)
+        Us4 = orth.sh_chebyu(4)
+        Us5 = orth.sh_chebyu(5)
+        use0 = orth.chebyu(0)(psub)
+        use1 = orth.chebyu(1)(psub)
+        use2 = orth.chebyu(2)(psub)
+        use3 = orth.chebyu(3)(psub)
+        use4 = orth.chebyu(4)(psub)
+        use5 = orth.chebyu(5)(psub)
+        assert_array_almost_equal(Us0.c,use0.c,13)
+        assert_array_almost_equal(Us1.c,use1.c,13)
+        assert_array_almost_equal(Us2.c,use2.c,13)
+        assert_array_almost_equal(Us3.c,use3.c,13)
+        assert_array_almost_equal(Us4.c,use4.c,12)
+        assert_array_almost_equal(Us5.c,use5.c,11)
+
+
+class TestShJacobi:
+    def test_sh_jacobi(self):
+        # G^(p,q)_n(x) = n! gamma(n+p)/gamma(2*n+p) * P^(p-q,q-1)_n(2*x-1)
+        def conv(n, p):
+            return gamma(n + 1) * gamma(n + p) / gamma(2 * n + p)
+        psub = np.poly1d([2,-1])
+        q = 4 * np.random.random()
+        p = q-1 + 2*np.random.random()
+        # print("shifted jacobi p,q = ", p, q)
+        G0 = orth.sh_jacobi(0,p,q)
+        G1 = orth.sh_jacobi(1,p,q)
+        G2 = orth.sh_jacobi(2,p,q)
+        G3 = orth.sh_jacobi(3,p,q)
+        G4 = orth.sh_jacobi(4,p,q)
+        G5 = orth.sh_jacobi(5,p,q)
+        ge0 = orth.jacobi(0,p-q,q-1)(psub) * conv(0,p)
+        ge1 = orth.jacobi(1,p-q,q-1)(psub) * conv(1,p)
+        ge2 = orth.jacobi(2,p-q,q-1)(psub) * conv(2,p)
+        ge3 = orth.jacobi(3,p-q,q-1)(psub) * conv(3,p)
+        ge4 = orth.jacobi(4,p-q,q-1)(psub) * conv(4,p)
+        ge5 = orth.jacobi(5,p-q,q-1)(psub) * conv(5,p)
+
+        assert_array_almost_equal(G0.c,ge0.c,13)
+        assert_array_almost_equal(G1.c,ge1.c,13)
+        assert_array_almost_equal(G2.c,ge2.c,13)
+        assert_array_almost_equal(G3.c,ge3.c,13)
+        assert_array_almost_equal(G4.c,ge4.c,13)
+        assert_array_almost_equal(G5.c,ge5.c,13)
+
+
+class TestCall:
+    def test_call(self):
+        poly = []
+        for n in range(5):
+            poly.extend([x.strip() for x in
+                ("""
+                orth.jacobi(%(n)d,0.3,0.9)
+                orth.sh_jacobi(%(n)d,0.3,0.9)
+                orth.genlaguerre(%(n)d,0.3)
+                orth.laguerre(%(n)d)
+                orth.hermite(%(n)d)
+                orth.hermitenorm(%(n)d)
+                orth.gegenbauer(%(n)d,0.3)
+                orth.chebyt(%(n)d)
+                orth.chebyu(%(n)d)
+                orth.chebyc(%(n)d)
+                orth.chebys(%(n)d)
+                orth.sh_chebyt(%(n)d)
+                orth.sh_chebyu(%(n)d)
+                orth.legendre(%(n)d)
+                orth.sh_legendre(%(n)d)
+                """ % dict(n=n)).split()
+            ])
+        with np.errstate(all='ignore'):
+            for pstr in poly:
+                p = eval(pstr)
+                assert_almost_equal(p(0.315), np.poly1d(p.coef)(0.315),
+                                    err_msg=pstr)
+
+
+class TestGenlaguerre:
+    def test_regression(self):
+        assert_equal(orth.genlaguerre(1, 1, monic=False)(0), 2.)
+        assert_equal(orth.genlaguerre(1, 1, monic=True)(0), -2.)
+        assert_equal(orth.genlaguerre(1, 1, monic=False), np.poly1d([-1, 2]))
+        assert_equal(orth.genlaguerre(1, 1, monic=True), np.poly1d([1, -2]))
+
+
+def verify_gauss_quad(root_func, eval_func, weight_func, a, b, N,
+                      rtol=1e-15, atol=5e-14):
+    # this test is copied from numpy's TestGauss in test_hermite.py
+    x, w, mu = root_func(N, True)
+
+    n = np.arange(N, dtype=np.dtype("long"))
+    v = eval_func(n[:,np.newaxis], x)
+    vv = np.dot(v*w, v.T)
+    vd = 1 / np.sqrt(vv.diagonal())
+    vv = vd[:, np.newaxis] * vv * vd
+    assert_allclose(vv, np.eye(N), rtol, atol)
+
+    # check that the integral of 1 is correct
+    assert_allclose(w.sum(), mu, rtol, atol)
+
+    # compare the results of integrating a function with quad.
+    def f(x):
+        return x ** 3 - 3 * x ** 2 + x - 2
+    resI = integrate.quad(lambda x: f(x)*weight_func(x), a, b)
+    resG = np.vdot(f(x), w)
+    rtol = 1e-6 if 1e-6 < resI[1] else resI[1] * 10
+    assert_allclose(resI[0], resG, rtol=rtol)
+
+def test_roots_jacobi():
+    def rf(a, b):
+        return lambda n, mu: sc.roots_jacobi(n, a, b, mu)
+    def ef(a, b):
+        return lambda n, x: sc.eval_jacobi(n, a, b, x)
+    def wf(a, b):
+        return lambda x: (1 - x) ** a * (1 + x) ** b
+
+    vgq = verify_gauss_quad
+    vgq(rf(-0.5, -0.75), ef(-0.5, -0.75), wf(-0.5, -0.75), -1., 1., 5)
+    vgq(rf(-0.5, -0.75), ef(-0.5, -0.75), wf(-0.5, -0.75), -1., 1.,
+        25, atol=1e-12)
+    vgq(rf(-0.5, -0.75), ef(-0.5, -0.75), wf(-0.5, -0.75), -1., 1.,
+        100, atol=1e-11)
+
+    vgq(rf(0.5, -0.5), ef(0.5, -0.5), wf(0.5, -0.5), -1., 1., 5)
+    vgq(rf(0.5, -0.5), ef(0.5, -0.5), wf(0.5, -0.5), -1., 1., 25, atol=1.5e-13)
+    vgq(rf(0.5, -0.5), ef(0.5, -0.5), wf(0.5, -0.5), -1., 1., 100, atol=2e-12)
+
+    vgq(rf(1, 0.5), ef(1, 0.5), wf(1, 0.5), -1., 1., 5, atol=2e-13)
+    vgq(rf(1, 0.5), ef(1, 0.5), wf(1, 0.5), -1., 1., 25, atol=2e-13)
+    vgq(rf(1, 0.5), ef(1, 0.5), wf(1, 0.5), -1., 1., 100, atol=1e-12)
+
+    vgq(rf(0.9, 2), ef(0.9, 2), wf(0.9, 2), -1., 1., 5)
+    vgq(rf(0.9, 2), ef(0.9, 2), wf(0.9, 2), -1., 1., 25, atol=1e-13)
+    vgq(rf(0.9, 2), ef(0.9, 2), wf(0.9, 2), -1., 1., 100, atol=3e-13)
+
+    vgq(rf(18.24, 27.3), ef(18.24, 27.3), wf(18.24, 27.3), -1., 1., 5)
+    vgq(rf(18.24, 27.3), ef(18.24, 27.3), wf(18.24, 27.3), -1., 1., 25,
+        atol=1.1e-14)
+    vgq(rf(18.24, 27.3), ef(18.24, 27.3), wf(18.24, 27.3), -1., 1.,
+        100, atol=1e-13)
+
+    vgq(rf(47.1, -0.2), ef(47.1, -0.2), wf(47.1, -0.2), -1., 1., 5, atol=1e-13)
+    vgq(rf(47.1, -0.2), ef(47.1, -0.2), wf(47.1, -0.2), -1., 1., 25, atol=2e-13)
+    vgq(rf(47.1, -0.2), ef(47.1, -0.2), wf(47.1, -0.2), -1., 1.,
+        100, atol=1e-11)
+
+    vgq(rf(1., 658.), ef(1., 658.), wf(1., 658.), -1., 1., 5, atol=2e-13)
+    vgq(rf(1., 658.), ef(1., 658.), wf(1., 658.), -1., 1., 25, atol=1e-12)
+    vgq(rf(1., 658.), ef(1., 658.), wf(1., 658.), -1., 1., 100, atol=1e-11)
+    vgq(rf(1., 658.), ef(1., 658.), wf(1., 658.), -1., 1., 250, atol=1e-11)
+
+    vgq(rf(511., 511.), ef(511., 511.), wf(511., 511.), -1., 1., 5,
+        atol=1e-12)
+    vgq(rf(511., 511.), ef(511., 511.), wf(511., 511.), -1., 1., 25,
+        atol=1e-11)
+    vgq(rf(511., 511.), ef(511., 511.), wf(511., 511.), -1., 1., 100,
+        atol=1e-10)
+
+    vgq(rf(511., 512.), ef(511., 512.), wf(511., 512.), -1., 1., 5,
+        atol=1e-12)
+    vgq(rf(511., 512.), ef(511., 512.), wf(511., 512.), -1., 1., 25,
+        atol=1e-11)
+    vgq(rf(511., 512.), ef(511., 512.), wf(511., 512.), -1., 1., 100,
+        atol=1e-10)
+
+    vgq(rf(1000., 500.), ef(1000., 500.), wf(1000., 500.), -1., 1., 5,
+        atol=1e-12)
+    vgq(rf(1000., 500.), ef(1000., 500.), wf(1000., 500.), -1., 1., 25,
+        atol=1e-11)
+    vgq(rf(1000., 500.), ef(1000., 500.), wf(1000., 500.), -1., 1., 100,
+        atol=1e-10)
+
+    vgq(rf(2.25, 68.9), ef(2.25, 68.9), wf(2.25, 68.9), -1., 1., 5)
+    vgq(rf(2.25, 68.9), ef(2.25, 68.9), wf(2.25, 68.9), -1., 1., 25,
+        atol=1e-13)
+    vgq(rf(2.25, 68.9), ef(2.25, 68.9), wf(2.25, 68.9), -1., 1., 100,
+        atol=1e-13)
+
+    # when alpha == beta == 0, P_n^{a,b}(x) == P_n(x)
+    xj, wj = sc.roots_jacobi(6, 0.0, 0.0)
+    xl, wl = sc.roots_legendre(6)
+    assert_allclose(xj, xl, 1e-14, 1e-14)
+    assert_allclose(wj, wl, 1e-14, 1e-14)
+
+    # when alpha == beta != 0, P_n^{a,b}(x) == C_n^{alpha+0.5}(x)
+    xj, wj = sc.roots_jacobi(6, 4.0, 4.0)
+    xc, wc = sc.roots_gegenbauer(6, 4.5)
+    assert_allclose(xj, xc, 1e-14, 1e-14)
+    assert_allclose(wj, wc, 1e-14, 1e-14)
+
+    x, w = sc.roots_jacobi(5, 2, 3, False)
+    y, v, m = sc.roots_jacobi(5, 2, 3, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(wf(2,3), -1, 1)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_jacobi, 0, 1, 1)
+    assert_raises(ValueError, sc.roots_jacobi, 3.3, 1, 1)
+    assert_raises(ValueError, sc.roots_jacobi, 3, -2, 1)
+    assert_raises(ValueError, sc.roots_jacobi, 3, 1, -2)
+    assert_raises(ValueError, sc.roots_jacobi, 3, -2, -2)
+
+def test_roots_sh_jacobi():
+    def rf(a, b):
+        return lambda n, mu: sc.roots_sh_jacobi(n, a, b, mu)
+    def ef(a, b):
+        return lambda n, x: sc.eval_sh_jacobi(n, a, b, x)
+    def wf(a, b):
+        return lambda x: (1.0 - x) ** (a - b) * x ** (b - 1.0)
+
+    vgq = verify_gauss_quad
+    vgq(rf(-0.5, 0.25), ef(-0.5, 0.25), wf(-0.5, 0.25), 0., 1., 5)
+    vgq(rf(-0.5, 0.25), ef(-0.5, 0.25), wf(-0.5, 0.25), 0., 1.,
+        25, atol=1e-12)
+    vgq(rf(-0.5, 0.25), ef(-0.5, 0.25), wf(-0.5, 0.25), 0., 1.,
+        100, atol=1e-11)
+
+    vgq(rf(0.5, 0.5), ef(0.5, 0.5), wf(0.5, 0.5), 0., 1., 5)
+    vgq(rf(0.5, 0.5), ef(0.5, 0.5), wf(0.5, 0.5), 0., 1., 25, atol=1e-13)
+    vgq(rf(0.5, 0.5), ef(0.5, 0.5), wf(0.5, 0.5), 0., 1., 100, atol=1e-12)
+
+    vgq(rf(1, 0.5), ef(1, 0.5), wf(1, 0.5), 0., 1., 5)
+    vgq(rf(1, 0.5), ef(1, 0.5), wf(1, 0.5), 0., 1., 25, atol=1.5e-13)
+    vgq(rf(1, 0.5), ef(1, 0.5), wf(1, 0.5), 0., 1., 100, atol=2e-12)
+
+    vgq(rf(2, 0.9), ef(2, 0.9), wf(2, 0.9), 0., 1., 5)
+    vgq(rf(2, 0.9), ef(2, 0.9), wf(2, 0.9), 0., 1., 25, atol=1e-13)
+    vgq(rf(2, 0.9), ef(2, 0.9), wf(2, 0.9), 0., 1., 100, atol=1e-12)
+
+    vgq(rf(27.3, 18.24), ef(27.3, 18.24), wf(27.3, 18.24), 0., 1., 5)
+    vgq(rf(27.3, 18.24), ef(27.3, 18.24), wf(27.3, 18.24), 0., 1., 25)
+    vgq(rf(27.3, 18.24), ef(27.3, 18.24), wf(27.3, 18.24), 0., 1.,
+        100, atol=1e-13)
+
+    vgq(rf(47.1, 0.2), ef(47.1, 0.2), wf(47.1, 0.2), 0., 1., 5, atol=1e-12)
+    vgq(rf(47.1, 0.2), ef(47.1, 0.2), wf(47.1, 0.2), 0., 1., 25, atol=1e-11)
+    vgq(rf(47.1, 0.2), ef(47.1, 0.2), wf(47.1, 0.2), 0., 1., 100, atol=1e-10)
+
+    vgq(rf(68.9, 2.25), ef(68.9, 2.25), wf(68.9, 2.25), 0., 1., 5, atol=3.5e-14)
+    vgq(rf(68.9, 2.25), ef(68.9, 2.25), wf(68.9, 2.25), 0., 1., 25, atol=2e-13)
+    vgq(rf(68.9, 2.25), ef(68.9, 2.25), wf(68.9, 2.25), 0., 1.,
+        100, atol=1e-12)
+
+    x, w = sc.roots_sh_jacobi(5, 3, 2, False)
+    y, v, m = sc.roots_sh_jacobi(5, 3, 2, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(wf(3,2), 0, 1)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_sh_jacobi, 0, 1, 1)
+    assert_raises(ValueError, sc.roots_sh_jacobi, 3.3, 1, 1)
+    assert_raises(ValueError, sc.roots_sh_jacobi, 3, 1, 2)    # p - q <= -1
+    assert_raises(ValueError, sc.roots_sh_jacobi, 3, 2, -1)   # q <= 0
+    assert_raises(ValueError, sc.roots_sh_jacobi, 3, -2, -1)  # both
+
+def test_roots_hermite():
+    rootf = sc.roots_hermite
+    evalf = sc.eval_hermite
+    weightf = orth.hermite(5).weight_func
+
+    verify_gauss_quad(rootf, evalf, weightf, -np.inf, np.inf, 5)
+    verify_gauss_quad(rootf, evalf, weightf, -np.inf, np.inf, 25, atol=1e-13)
+    verify_gauss_quad(rootf, evalf, weightf, -np.inf, np.inf, 100, atol=1e-12)
+
+    # Golub-Welsch branch
+    x, w = sc.roots_hermite(5, False)
+    y, v, m = sc.roots_hermite(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, -np.inf, np.inf)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    # Asymptotic branch (switch over at n >= 150)
+    x, w = sc.roots_hermite(200, False)
+    y, v, m = sc.roots_hermite(200, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+    assert_allclose(sum(v), m, 1e-14, 1e-14)
+
+    assert_raises(ValueError, sc.roots_hermite, 0)
+    assert_raises(ValueError, sc.roots_hermite, 3.3)
+
+def test_roots_hermite_asy():
+    # Recursion for Hermite functions
+    def hermite_recursion(n, nodes):
+        H = np.zeros((n, nodes.size))
+        H[0,:] = np.pi**(-0.25) * np.exp(-0.5*nodes**2)
+        if n > 1:
+            H[1,:] = sqrt(2.0) * nodes * H[0,:]
+            for k in range(2, n):
+                H[k,:] = sqrt(2.0/k) * nodes * H[k-1,:] - sqrt((k-1.0)/k) * H[k-2,:]
+        return H
+
+    # This tests only the nodes
+    def test(N, rtol=1e-15, atol=1e-14):
+        x, w = orth._roots_hermite_asy(N)
+        H = hermite_recursion(N+1, x)
+        assert_allclose(H[-1,:], np.zeros(N), rtol, atol)
+        assert_allclose(sum(w), sqrt(np.pi), rtol, atol)
+
+    test(150, atol=1e-12)
+    test(151, atol=1e-12)
+    test(300, atol=1e-12)
+    test(301, atol=1e-12)
+    test(500, atol=1e-12)
+    test(501, atol=1e-12)
+    test(999, atol=1e-12)
+    test(1000, atol=1e-12)
+    test(2000, atol=1e-12)
+    test(5000, atol=1e-12)
+
+def test_roots_hermitenorm():
+    rootf = sc.roots_hermitenorm
+    evalf = sc.eval_hermitenorm
+    weightf = orth.hermitenorm(5).weight_func
+
+    verify_gauss_quad(rootf, evalf, weightf, -np.inf, np.inf, 5)
+    verify_gauss_quad(rootf, evalf, weightf, -np.inf, np.inf, 25, atol=1e-13)
+    verify_gauss_quad(rootf, evalf, weightf, -np.inf, np.inf, 100, atol=1e-12)
+
+    x, w = sc.roots_hermitenorm(5, False)
+    y, v, m = sc.roots_hermitenorm(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, -np.inf, np.inf)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_hermitenorm, 0)
+    assert_raises(ValueError, sc.roots_hermitenorm, 3.3)
+
+def test_roots_gegenbauer():
+    def rootf(a):
+        return lambda n, mu: sc.roots_gegenbauer(n, a, mu)
+    def evalf(a):
+        return lambda n, x: sc.eval_gegenbauer(n, a, x)
+    def weightf(a):
+        return lambda x: (1 - x ** 2) ** (a - 0.5)
+
+    vgq = verify_gauss_quad
+    vgq(rootf(-0.25), evalf(-0.25), weightf(-0.25), -1., 1., 5)
+    vgq(rootf(-0.25), evalf(-0.25), weightf(-0.25), -1., 1., 25, atol=1e-12)
+    vgq(rootf(-0.25), evalf(-0.25), weightf(-0.25), -1., 1., 100, atol=1e-11)
+
+    vgq(rootf(0.1), evalf(0.1), weightf(0.1), -1., 1., 5)
+    vgq(rootf(0.1), evalf(0.1), weightf(0.1), -1., 1., 25, atol=1e-13)
+    vgq(rootf(0.1), evalf(0.1), weightf(0.1), -1., 1., 100, atol=1e-12)
+
+    vgq(rootf(1), evalf(1), weightf(1), -1., 1., 5)
+    vgq(rootf(1), evalf(1), weightf(1), -1., 1., 25, atol=1e-13)
+    vgq(rootf(1), evalf(1), weightf(1), -1., 1., 100, atol=1e-12)
+
+    vgq(rootf(10), evalf(10), weightf(10), -1., 1., 5)
+    vgq(rootf(10), evalf(10), weightf(10), -1., 1., 25, atol=1e-13)
+    vgq(rootf(10), evalf(10), weightf(10), -1., 1., 100, atol=1e-12)
+
+    vgq(rootf(50), evalf(50), weightf(50), -1., 1., 5, atol=1e-13)
+    vgq(rootf(50), evalf(50), weightf(50), -1., 1., 25, atol=1e-12)
+    vgq(rootf(50), evalf(50), weightf(50), -1., 1., 100, atol=1e-11)
+
+    # Alpha=170 is where the approximation used in roots_gegenbauer changes
+    vgq(rootf(170), evalf(170), weightf(170), -1., 1., 5, atol=1e-13)
+    vgq(rootf(170), evalf(170), weightf(170), -1., 1., 25, atol=1e-12)
+    vgq(rootf(170), evalf(170), weightf(170), -1., 1., 100, atol=1e-11)
+    vgq(rootf(170.5), evalf(170.5), weightf(170.5), -1., 1., 5, atol=1.25e-13)
+    vgq(rootf(170.5), evalf(170.5), weightf(170.5), -1., 1., 25, atol=1e-12)
+    vgq(rootf(170.5), evalf(170.5), weightf(170.5), -1., 1., 100, atol=1e-11)
+
+    # Test for failures, e.g. overflows, resulting from large alphas
+    vgq(rootf(238), evalf(238), weightf(238), -1., 1., 5, atol=1e-13)
+    vgq(rootf(238), evalf(238), weightf(238), -1., 1., 25, atol=1e-12)
+    vgq(rootf(238), evalf(238), weightf(238), -1., 1., 100, atol=1e-11)
+    vgq(rootf(512.5), evalf(512.5), weightf(512.5), -1., 1., 5, atol=1e-12)
+    vgq(rootf(512.5), evalf(512.5), weightf(512.5), -1., 1., 25, atol=1e-11)
+    vgq(rootf(512.5), evalf(512.5), weightf(512.5), -1., 1., 100, atol=1e-10)
+
+    # this is a special case that the old code supported.
+    # when alpha = 0, the gegenbauer polynomial is uniformly 0. but it goes
+    # to a scaled down copy of T_n(x) there.
+    vgq(rootf(0), sc.eval_chebyt, weightf(0), -1., 1., 5)
+    vgq(rootf(0), sc.eval_chebyt, weightf(0), -1., 1., 25)
+    vgq(rootf(0), sc.eval_chebyt, weightf(0), -1., 1., 100, atol=1e-12)
+
+    x, w = sc.roots_gegenbauer(5, 2, False)
+    y, v, m = sc.roots_gegenbauer(5, 2, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf(2), -1, 1)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_gegenbauer, 0, 2)
+    assert_raises(ValueError, sc.roots_gegenbauer, 3.3, 2)
+    assert_raises(ValueError, sc.roots_gegenbauer, 3, -.75)
+
+def test_roots_chebyt():
+    weightf = orth.chebyt(5).weight_func
+    verify_gauss_quad(sc.roots_chebyt, sc.eval_chebyt, weightf, -1., 1., 5)
+    verify_gauss_quad(sc.roots_chebyt, sc.eval_chebyt, weightf, -1., 1., 25)
+    verify_gauss_quad(sc.roots_chebyt, sc.eval_chebyt, weightf, -1., 1., 100,
+                      atol=1e-12)
+
+    x, w = sc.roots_chebyt(5, False)
+    y, v, m = sc.roots_chebyt(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, -1, 1)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_chebyt, 0)
+    assert_raises(ValueError, sc.roots_chebyt, 3.3)
+
+def test_chebyt_symmetry():
+    x, w = sc.roots_chebyt(21)
+    pos, neg = x[:10], x[11:]
+    assert_equal(neg, -pos[::-1])
+    assert_equal(x[10], 0)
+
+def test_roots_chebyu():
+    weightf = orth.chebyu(5).weight_func
+    verify_gauss_quad(sc.roots_chebyu, sc.eval_chebyu, weightf, -1., 1., 5)
+    verify_gauss_quad(sc.roots_chebyu, sc.eval_chebyu, weightf, -1., 1., 25)
+    verify_gauss_quad(sc.roots_chebyu, sc.eval_chebyu, weightf, -1., 1., 100)
+
+    x, w = sc.roots_chebyu(5, False)
+    y, v, m = sc.roots_chebyu(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, -1, 1)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_chebyu, 0)
+    assert_raises(ValueError, sc.roots_chebyu, 3.3)
+
+def test_roots_chebyc():
+    weightf = orth.chebyc(5).weight_func
+    verify_gauss_quad(sc.roots_chebyc, sc.eval_chebyc, weightf, -2., 2., 5)
+    verify_gauss_quad(sc.roots_chebyc, sc.eval_chebyc, weightf, -2., 2., 25)
+    verify_gauss_quad(sc.roots_chebyc, sc.eval_chebyc, weightf, -2., 2., 100,
+                      atol=1e-12)
+
+    x, w = sc.roots_chebyc(5, False)
+    y, v, m = sc.roots_chebyc(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, -2, 2)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_chebyc, 0)
+    assert_raises(ValueError, sc.roots_chebyc, 3.3)
+
+def test_roots_chebys():
+    weightf = orth.chebys(5).weight_func
+    verify_gauss_quad(sc.roots_chebys, sc.eval_chebys, weightf, -2., 2., 5)
+    verify_gauss_quad(sc.roots_chebys, sc.eval_chebys, weightf, -2., 2., 25)
+    verify_gauss_quad(sc.roots_chebys, sc.eval_chebys, weightf, -2., 2., 100)
+
+    x, w = sc.roots_chebys(5, False)
+    y, v, m = sc.roots_chebys(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, -2, 2)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_chebys, 0)
+    assert_raises(ValueError, sc.roots_chebys, 3.3)
+
+def test_roots_sh_chebyt():
+    weightf = orth.sh_chebyt(5).weight_func
+    verify_gauss_quad(sc.roots_sh_chebyt, sc.eval_sh_chebyt, weightf, 0., 1., 5)
+    verify_gauss_quad(sc.roots_sh_chebyt, sc.eval_sh_chebyt, weightf, 0., 1., 25)
+    verify_gauss_quad(sc.roots_sh_chebyt, sc.eval_sh_chebyt, weightf, 0., 1.,
+                      100, atol=1e-13)
+
+    x, w = sc.roots_sh_chebyt(5, False)
+    y, v, m = sc.roots_sh_chebyt(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, 0, 1)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_sh_chebyt, 0)
+    assert_raises(ValueError, sc.roots_sh_chebyt, 3.3)
+
+def test_roots_sh_chebyu():
+    weightf = orth.sh_chebyu(5).weight_func
+    verify_gauss_quad(sc.roots_sh_chebyu, sc.eval_sh_chebyu, weightf, 0., 1., 5)
+    verify_gauss_quad(sc.roots_sh_chebyu, sc.eval_sh_chebyu, weightf, 0., 1., 25)
+    verify_gauss_quad(sc.roots_sh_chebyu, sc.eval_sh_chebyu, weightf, 0., 1.,
+                      100, atol=1e-13)
+
+    x, w = sc.roots_sh_chebyu(5, False)
+    y, v, m = sc.roots_sh_chebyu(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, 0, 1)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_sh_chebyu, 0)
+    assert_raises(ValueError, sc.roots_sh_chebyu, 3.3)
+
+def test_roots_legendre():
+    weightf = orth.legendre(5).weight_func
+    verify_gauss_quad(sc.roots_legendre, sc.eval_legendre, weightf, -1., 1., 5)
+    verify_gauss_quad(sc.roots_legendre, sc.eval_legendre, weightf, -1., 1.,
+                      25, atol=1e-13)
+    verify_gauss_quad(sc.roots_legendre, sc.eval_legendre, weightf, -1., 1.,
+                      100, atol=1e-12)
+
+    x, w = sc.roots_legendre(5, False)
+    y, v, m = sc.roots_legendre(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, -1, 1)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_legendre, 0)
+    assert_raises(ValueError, sc.roots_legendre, 3.3)
+
+def test_roots_sh_legendre():
+    weightf = orth.sh_legendre(5).weight_func
+    verify_gauss_quad(sc.roots_sh_legendre, sc.eval_sh_legendre, weightf, 0., 1., 5)
+    verify_gauss_quad(sc.roots_sh_legendre, sc.eval_sh_legendre, weightf, 0., 1.,
+                      25, atol=1e-13)
+    verify_gauss_quad(sc.roots_sh_legendre, sc.eval_sh_legendre, weightf, 0., 1.,
+                      100, atol=1e-12)
+
+    x, w = sc.roots_sh_legendre(5, False)
+    y, v, m = sc.roots_sh_legendre(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, 0, 1)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_sh_legendre, 0)
+    assert_raises(ValueError, sc.roots_sh_legendre, 3.3)
+
+def test_roots_laguerre():
+    weightf = orth.laguerre(5).weight_func
+    verify_gauss_quad(sc.roots_laguerre, sc.eval_laguerre, weightf, 0., np.inf, 5)
+    verify_gauss_quad(sc.roots_laguerre, sc.eval_laguerre, weightf, 0., np.inf,
+                      25, atol=1e-13)
+    verify_gauss_quad(sc.roots_laguerre, sc.eval_laguerre, weightf, 0., np.inf,
+                      100, atol=1e-12)
+
+    x, w = sc.roots_laguerre(5, False)
+    y, v, m = sc.roots_laguerre(5, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf, 0, np.inf)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_laguerre, 0)
+    assert_raises(ValueError, sc.roots_laguerre, 3.3)
+
+def test_roots_genlaguerre():
+    def rootf(a):
+        return lambda n, mu: sc.roots_genlaguerre(n, a, mu)
+    def evalf(a):
+        return lambda n, x: sc.eval_genlaguerre(n, a, x)
+    def weightf(a):
+        return lambda x: x ** a * np.exp(-x)
+
+    vgq = verify_gauss_quad
+    vgq(rootf(-0.5), evalf(-0.5), weightf(-0.5), 0., np.inf, 5)
+    vgq(rootf(-0.5), evalf(-0.5), weightf(-0.5), 0., np.inf, 25, atol=1e-13)
+    vgq(rootf(-0.5), evalf(-0.5), weightf(-0.5), 0., np.inf, 100, atol=1e-12)
+
+    vgq(rootf(0.1), evalf(0.1), weightf(0.1), 0., np.inf, 5)
+    vgq(rootf(0.1), evalf(0.1), weightf(0.1), 0., np.inf, 25, atol=1e-13)
+    vgq(rootf(0.1), evalf(0.1), weightf(0.1), 0., np.inf, 100, atol=1.6e-13)
+
+    vgq(rootf(1), evalf(1), weightf(1), 0., np.inf, 5)
+    vgq(rootf(1), evalf(1), weightf(1), 0., np.inf, 25, atol=1e-13)
+    vgq(rootf(1), evalf(1), weightf(1), 0., np.inf, 100, atol=1.03e-13)
+
+    vgq(rootf(10), evalf(10), weightf(10), 0., np.inf, 5)
+    vgq(rootf(10), evalf(10), weightf(10), 0., np.inf, 25, atol=1e-13)
+    vgq(rootf(10), evalf(10), weightf(10), 0., np.inf, 100, atol=1e-12)
+
+    vgq(rootf(50), evalf(50), weightf(50), 0., np.inf, 5)
+    vgq(rootf(50), evalf(50), weightf(50), 0., np.inf, 25, atol=1e-13)
+    vgq(rootf(50), evalf(50), weightf(50), 0., np.inf, 100, rtol=1e-14, atol=2e-13)
+
+    x, w = sc.roots_genlaguerre(5, 2, False)
+    y, v, m = sc.roots_genlaguerre(5, 2, True)
+    assert_allclose(x, y, 1e-14, 1e-14)
+    assert_allclose(w, v, 1e-14, 1e-14)
+
+    muI, muI_err = integrate.quad(weightf(2.), 0., np.inf)
+    assert_allclose(m, muI, rtol=muI_err)
+
+    assert_raises(ValueError, sc.roots_genlaguerre, 0, 2)
+    assert_raises(ValueError, sc.roots_genlaguerre, 3.3, 2)
+    assert_raises(ValueError, sc.roots_genlaguerre, 3, -1.1)
+
+
+def test_gh_6721():
+    # Regression test for gh_6721. This should not raise.
+    sc.chebyt(65)(0.2)

.venv/Lib/site-packages/scipy/special/tests/test_orthogonal_eval.py

@@ -0,0 +1,272 @@
+import numpy as np
+from numpy.testing import assert_, assert_allclose
+import pytest
+
+from scipy.special import _ufuncs
+import scipy.special._orthogonal as orth
+from scipy.special._testutils import FuncData
+
+
+def test_eval_chebyt():
+    n = np.arange(0, 10000, 7, dtype=np.dtype("long"))
+    x = 2*np.random.rand() - 1
+    v1 = np.cos(n*np.arccos(x))
+    v2 = _ufuncs.eval_chebyt(n, x)
+    assert_(np.allclose(v1, v2, rtol=1e-15))
+
+
+def test_eval_chebyt_gh20129():
+    # https://github.com/scipy/scipy/issues/20129
+    assert _ufuncs.eval_chebyt(7, 2 + 0j) == 5042.0
+
+def test_eval_genlaguerre_restriction():
+    # check it returns nan for alpha <= -1
+    assert_(np.isnan(_ufuncs.eval_genlaguerre(0, -1, 0)))
+    assert_(np.isnan(_ufuncs.eval_genlaguerre(0.1, -1, 0)))
+
+
+def test_warnings():
+    # ticket 1334
+    with np.errstate(all='raise'):
+        # these should raise no fp warnings
+        _ufuncs.eval_legendre(1, 0)
+        _ufuncs.eval_laguerre(1, 1)
+        _ufuncs.eval_gegenbauer(1, 1, 0)
+
+
+class TestPolys:
+    """
+    Check that the eval_* functions agree with the constructed polynomials
+
+    """
+
+    def check_poly(self, func, cls, param_ranges=[], x_range=[], nn=10,
+                   nparam=10, nx=10, rtol=1e-8):
+        np.random.seed(1234)
+
+        dataset = []
+        for n in np.arange(nn):
+            params = [a + (b-a)*np.random.rand(nparam) for a,b in param_ranges]
+            params = np.asarray(params).T
+            if not param_ranges:
+                params = [0]
+            for p in params:
+                if param_ranges:
+                    p = (n,) + tuple(p)
+                else:
+                    p = (n,)
+                x = x_range[0] + (x_range[1] - x_range[0])*np.random.rand(nx)
+                x[0] = x_range[0]  # always include domain start point
+                x[1] = x_range[1]  # always include domain end point
+                poly = np.poly1d(cls(*p).coef)
+                z = np.c_[np.tile(p, (nx,1)), x, poly(x)]
+                dataset.append(z)
+
+        dataset = np.concatenate(dataset, axis=0)
+
+        def polyfunc(*p):
+            p = (p[0].astype(np.dtype("long")),) + p[1:]
+            return func(*p)
+
+        with np.errstate(all='raise'):
+            ds = FuncData(polyfunc, dataset, list(range(len(param_ranges)+2)), -1,
+                          rtol=rtol)
+            ds.check()
+
+    def test_jacobi(self):
+        self.check_poly(_ufuncs.eval_jacobi, orth.jacobi,
+                        param_ranges=[(-0.99, 10), (-0.99, 10)],
+                        x_range=[-1, 1], rtol=1e-5)
+
+    def test_sh_jacobi(self):
+        self.check_poly(_ufuncs.eval_sh_jacobi, orth.sh_jacobi,
+                        param_ranges=[(1, 10), (0, 1)], x_range=[0, 1],
+                        rtol=1e-5)
+
+    def test_gegenbauer(self):
+        self.check_poly(_ufuncs.eval_gegenbauer, orth.gegenbauer,
+                        param_ranges=[(-0.499, 10)], x_range=[-1, 1],
+                        rtol=1e-7)
+
+    def test_chebyt(self):
+        self.check_poly(_ufuncs.eval_chebyt, orth.chebyt,
+                        param_ranges=[], x_range=[-1, 1])
+
+    def test_chebyu(self):
+        self.check_poly(_ufuncs.eval_chebyu, orth.chebyu,
+                        param_ranges=[], x_range=[-1, 1])
+
+    def test_chebys(self):
+        self.check_poly(_ufuncs.eval_chebys, orth.chebys,
+                        param_ranges=[], x_range=[-2, 2])
+
+    def test_chebyc(self):
+        self.check_poly(_ufuncs.eval_chebyc, orth.chebyc,
+                        param_ranges=[], x_range=[-2, 2])
+
+    def test_sh_chebyt(self):
+        with np.errstate(all='ignore'):
+            self.check_poly(_ufuncs.eval_sh_chebyt, orth.sh_chebyt,
+                            param_ranges=[], x_range=[0, 1])
+
+    def test_sh_chebyu(self):
+        self.check_poly(_ufuncs.eval_sh_chebyu, orth.sh_chebyu,
+                        param_ranges=[], x_range=[0, 1])
+
+    def test_legendre(self):
+        self.check_poly(_ufuncs.eval_legendre, orth.legendre,
+                        param_ranges=[], x_range=[-1, 1])
+
+    def test_sh_legendre(self):
+        with np.errstate(all='ignore'):
+            self.check_poly(_ufuncs.eval_sh_legendre, orth.sh_legendre,
+                            param_ranges=[], x_range=[0, 1])
+
+    def test_genlaguerre(self):
+        self.check_poly(_ufuncs.eval_genlaguerre, orth.genlaguerre,
+                        param_ranges=[(-0.99, 10)], x_range=[0, 100])
+
+    def test_laguerre(self):
+        self.check_poly(_ufuncs.eval_laguerre, orth.laguerre,
+                        param_ranges=[], x_range=[0, 100])
+
+    def test_hermite(self):
+        self.check_poly(_ufuncs.eval_hermite, orth.hermite,
+                        param_ranges=[], x_range=[-100, 100])
+
+    def test_hermitenorm(self):
+        self.check_poly(_ufuncs.eval_hermitenorm, orth.hermitenorm,
+                        param_ranges=[], x_range=[-100, 100])
+
+
+class TestRecurrence:
+    """
+    Check that the eval_* functions sig='ld->d' and 'dd->d' agree.
+
+    """
+
+    def check_poly(self, func, param_ranges=[], x_range=[], nn=10,
+                   nparam=10, nx=10, rtol=1e-8):
+        np.random.seed(1234)
+
+        dataset = []
+        for n in np.arange(nn):
+            params = [a + (b-a)*np.random.rand(nparam) for a,b in param_ranges]
+            params = np.asarray(params).T
+            if not param_ranges:
+                params = [0]
+            for p in params:
+                if param_ranges:
+                    p = (n,) + tuple(p)
+                else:
+                    p = (n,)
+                x = x_range[0] + (x_range[1] - x_range[0])*np.random.rand(nx)
+                x[0] = x_range[0]  # always include domain start point
+                x[1] = x_range[1]  # always include domain end point
+                kw = dict(sig=(len(p)+1)*'d'+'->d')
+                z = np.c_[np.tile(p, (nx,1)), x, func(*(p + (x,)), **kw)]
+                dataset.append(z)
+
+        dataset = np.concatenate(dataset, axis=0)
+
+        def polyfunc(*p):
+            p = (p[0].astype(int),) + p[1:]
+            kw = dict(sig='l'+(len(p)-1)*'d'+'->d')
+            return func(*p, **kw)
+
+        with np.errstate(all='raise'):
+            ds = FuncData(polyfunc, dataset, list(range(len(param_ranges)+2)), -1,
+                          rtol=rtol)
+            ds.check()
+
+    def test_jacobi(self):
+        self.check_poly(_ufuncs.eval_jacobi,
+                        param_ranges=[(-0.99, 10), (-0.99, 10)],
+                        x_range=[-1, 1])
+
+    def test_sh_jacobi(self):
+        self.check_poly(_ufuncs.eval_sh_jacobi,
+                        param_ranges=[(1, 10), (0, 1)], x_range=[0, 1])
+
+    def test_gegenbauer(self):
+        self.check_poly(_ufuncs.eval_gegenbauer,
+                        param_ranges=[(-0.499, 10)], x_range=[-1, 1])
+
+    def test_chebyt(self):
+        self.check_poly(_ufuncs.eval_chebyt,
+                        param_ranges=[], x_range=[-1, 1])
+
+    def test_chebyu(self):
+        self.check_poly(_ufuncs.eval_chebyu,
+                        param_ranges=[], x_range=[-1, 1])
+
+    def test_chebys(self):
+        self.check_poly(_ufuncs.eval_chebys,
+                        param_ranges=[], x_range=[-2, 2])
+
+    def test_chebyc(self):
+        self.check_poly(_ufuncs.eval_chebyc,
+                        param_ranges=[], x_range=[-2, 2])
+
+    def test_sh_chebyt(self):
+        self.check_poly(_ufuncs.eval_sh_chebyt,
+                        param_ranges=[], x_range=[0, 1])
+
+    def test_sh_chebyu(self):
+        self.check_poly(_ufuncs.eval_sh_chebyu,
+                        param_ranges=[], x_range=[0, 1])
+
+    def test_legendre(self):
+        self.check_poly(_ufuncs.eval_legendre,
+                        param_ranges=[], x_range=[-1, 1])
+
+    def test_sh_legendre(self):
+        self.check_poly(_ufuncs.eval_sh_legendre,
+                        param_ranges=[], x_range=[0, 1])
+
+    def test_genlaguerre(self):
+        self.check_poly(_ufuncs.eval_genlaguerre,
+                        param_ranges=[(-0.99, 10)], x_range=[0, 100])
+
+    def test_laguerre(self):
+        self.check_poly(_ufuncs.eval_laguerre,
+                        param_ranges=[], x_range=[0, 100])
+
+    def test_hermite(self):
+        v = _ufuncs.eval_hermite(70, 1.0)
+        a = -1.457076485701412e60
+        assert_allclose(v, a)
+
+
+def test_hermite_domain():
+    # Regression test for gh-11091.
+    assert np.isnan(_ufuncs.eval_hermite(-1, 1.0))
+    assert np.isnan(_ufuncs.eval_hermitenorm(-1, 1.0))
+
+
+@pytest.mark.parametrize("n", [0, 1, 2])
+@pytest.mark.parametrize("x", [0, 1, np.nan])
+def test_hermite_nan(n, x):
+    # Regression test for gh-11369.
+    assert np.isnan(_ufuncs.eval_hermite(n, x)) == np.any(np.isnan([n, x]))
+    assert np.isnan(_ufuncs.eval_hermitenorm(n, x)) == np.any(np.isnan([n, x]))
+
+
+@pytest.mark.parametrize('n', [0, 1, 2, 3.2])
+@pytest.mark.parametrize('alpha', [1, np.nan])
+@pytest.mark.parametrize('x', [2, np.nan])
+def test_genlaguerre_nan(n, alpha, x):
+    # Regression test for gh-11361.
+    nan_laguerre = np.isnan(_ufuncs.eval_genlaguerre(n, alpha, x))
+    nan_arg = np.any(np.isnan([n, alpha, x]))
+    assert nan_laguerre == nan_arg
+
+
+@pytest.mark.parametrize('n', [0, 1, 2, 3.2])
+@pytest.mark.parametrize('alpha', [0.0, 1, np.nan])
+@pytest.mark.parametrize('x', [1e-6, 2, np.nan])
+def test_gegenbauer_nan(n, alpha, x):
+    # Regression test for gh-11370.
+    nan_gegenbauer = np.isnan(_ufuncs.eval_gegenbauer(n, alpha, x))
+    nan_arg = np.any(np.isnan([n, alpha, x]))
+    assert nan_gegenbauer == nan_arg

.venv/Lib/site-packages/scipy/special/tests/test_owens_t.py

@@ -0,0 +1,53 @@
+import numpy as np
+from numpy.testing import assert_equal, assert_allclose
+
+import scipy.special as sc
+
+
+def test_symmetries():
+    np.random.seed(1234)
+    a, h = np.random.rand(100), np.random.rand(100)
+    assert_equal(sc.owens_t(h, a), sc.owens_t(-h, a))
+    assert_equal(sc.owens_t(h, a), -sc.owens_t(h, -a))
+
+
+def test_special_cases():
+    assert_equal(sc.owens_t(5, 0), 0)
+    assert_allclose(sc.owens_t(0, 5), 0.5*np.arctan(5)/np.pi,
+                    rtol=5e-14)
+    # Target value is 0.5*Phi(5)*(1 - Phi(5)) for Phi the CDF of the
+    # standard normal distribution
+    assert_allclose(sc.owens_t(5, 1), 1.4332574485503512543e-07,
+                    rtol=5e-14)
+
+
+def test_nans():
+    assert_equal(sc.owens_t(20, np.nan), np.nan)
+    assert_equal(sc.owens_t(np.nan, 20), np.nan)
+    assert_equal(sc.owens_t(np.nan, np.nan), np.nan)
+
+
+def test_infs():
+    h, a = 0, np.inf
+    # T(0, a) = 1/2π * arctan(a)
+    res = 1/(2*np.pi) * np.arctan(a)
+    assert_allclose(sc.owens_t(h, a), res, rtol=5e-14)
+    assert_allclose(sc.owens_t(h, -a), -res, rtol=5e-14)
+
+    h = 1
+    # Refer Owens T function definition in Wikipedia
+    # https://en.wikipedia.org/wiki/Owen%27s_T_function
+    # Value approximated through Numerical Integration
+    # using scipy.integrate.quad
+    # quad(lambda x: 1/(2*pi)*(exp(-0.5*(1*1)*(1+x*x))/(1+x*x)), 0, inf)
+    res = 0.07932762696572854
+    assert_allclose(sc.owens_t(h, np.inf), res, rtol=5e-14)
+    assert_allclose(sc.owens_t(h, -np.inf), -res, rtol=5e-14)
+
+    assert_equal(sc.owens_t(np.inf, 1), 0)
+    assert_equal(sc.owens_t(-np.inf, 1), 0)
+
+    assert_equal(sc.owens_t(np.inf, np.inf), 0)
+    assert_equal(sc.owens_t(-np.inf, np.inf), 0)
+    assert_equal(sc.owens_t(np.inf, -np.inf), -0.0)
+    assert_equal(sc.owens_t(-np.inf, -np.inf), -0.0)

.venv/Lib/site-packages/scipy/special/tests/test_pcf.py

@@ -0,0 +1,24 @@
+"""Tests for parabolic cylinder functions.
+
+"""
+import numpy as np
+from numpy.testing import assert_allclose, assert_equal
+import scipy.special as sc
+
+
+def test_pbwa_segfault():
+    # Regression test for https://github.com/scipy/scipy/issues/6208.
+    #
+    # Data generated by mpmath.
+    #
+    w = 1.02276567211316867161
+    wp = -0.48887053372346189882
+    assert_allclose(sc.pbwa(0, 0), (w, wp), rtol=1e-13, atol=0)
+
+
+def test_pbwa_nan():
+    # Check that NaN's are returned outside of the range in which the
+    # implementation is accurate.
+    pts = [(-6, -6), (-6, 6), (6, -6), (6, 6)]
+    for p in pts:
+        assert_equal(sc.pbwa(*p), (np.nan, np.nan))

.venv/Lib/site-packages/scipy/special/tests/test_pdtr.py

@@ -0,0 +1,48 @@
+import numpy as np
+import scipy.special as sc
+from numpy.testing import assert_almost_equal, assert_array_equal
+
+
+class TestPdtr:
+    def test(self):
+        val = sc.pdtr(0, 1)
+        assert_almost_equal(val, np.exp(-1))
+
+    def test_m_zero(self):
+        val = sc.pdtr([0, 1, 2], 0)
+        assert_array_equal(val, [1, 1, 1])
+
+    def test_rounding(self):
+        double_val = sc.pdtr([0.1, 1.1, 2.1], 1.0)
+        int_val = sc.pdtr([0, 1, 2], 1.0)
+        assert_array_equal(double_val, int_val)
+
+    def test_inf(self):
+        val = sc.pdtr(np.inf, 1.0)
+        assert_almost_equal(val, 1.0)
+
+    def test_domain(self):
+        val = sc.pdtr(-1.1, 1.0)
+        assert np.isnan(val)
+
+class TestPdtrc:
+    def test_value(self):
+        val = sc.pdtrc(0, 1)
+        assert_almost_equal(val, 1 - np.exp(-1))
+
+    def test_m_zero(self):
+        val = sc.pdtrc([0, 1, 2], 0.0)
+        assert_array_equal(val, [0, 0, 0])
+
+    def test_rounding(self):
+        double_val = sc.pdtrc([0.1, 1.1, 2.1], 1.0)
+        int_val = sc.pdtrc([0, 1, 2], 1.0)
+        assert_array_equal(double_val, int_val)
+
+    def test_inf(self):
+        val = sc.pdtrc(np.inf, 1.0)
+        assert_almost_equal(val, 0.0)
+
+    def test_domain(self):
+        val = sc.pdtrc(-1.1, 1.0)
+        assert np.isnan(val)

.venv/Lib/site-packages/scipy/special/tests/test_powm1.py

@@ -0,0 +1,65 @@
+import pytest
+import numpy as np
+from numpy.testing import assert_allclose
+from scipy.special import powm1
+
+
+# Expected values were computed with mpmath, e.g.
+#
+#   >>> import mpmath
+#   >>> mpmath.np.dps = 200
+#   >>> print(float(mpmath.powm1(2.0, 1e-7))
+#   6.931472045825965e-08
+#
+powm1_test_cases = [
+    (1.25, 0.75, 0.18217701125396976, 1e-15),
+    (2.0, 1e-7, 6.931472045825965e-08, 1e-15),
+    (25.0, 5e-11, 1.6094379125636148e-10, 1e-15),
+    (0.99996, 0.75, -3.0000150002530058e-05, 1e-15),
+    (0.9999999999990905, 20, -1.81898940353014e-11, 1e-15),
+    (-1.25, 751.0, -6.017550852453444e+72, 2e-15)
+]
+
+
+@pytest.mark.parametrize('x, y, expected, rtol', powm1_test_cases)
+def test_powm1(x, y, expected, rtol):
+    p = powm1(x, y)
+    assert_allclose(p, expected, rtol=rtol)
+
+
+@pytest.mark.parametrize('x, y, expected',
+                         [(0.0, 0.0, 0.0),
+                          (0.0, -1.5, np.inf),
+                          (0.0, 1.75, -1.0),
+                          (-1.5, 2.0, 1.25),
+                          (-1.5, 3.0, -4.375),
+                          (np.nan, 0.0, 0.0),
+                          (1.0, np.nan, 0.0),
+                          (1.0, np.inf, 0.0),
+                          (1.0, -np.inf, 0.0),
+                          (np.inf, 7.5, np.inf),
+                          (np.inf, -7.5, -1.0),
+                          (3.25, np.inf, np.inf),
+                          (np.inf, np.inf, np.inf),
+                          (np.inf, -np.inf, -1.0),
+                          (np.inf, 0.0, 0.0),
+                          (-np.inf, 0.0, 0.0),
+                          (-np.inf, 2.0, np.inf),
+                          (-np.inf, 3.0, -np.inf),
+                          (-1.0, float(2**53 - 1), -2.0)])
+def test_powm1_exact_cases(x, y, expected):
+    # Test cases where we have an exact expected value.
+    p = powm1(x, y)
+    assert p == expected
+
+
+@pytest.mark.parametrize('x, y',
+                         [(-1.25, 751.03),
+                          (-1.25, np.inf),
+                          (np.nan, np.nan),
+                          (-np.inf, -np.inf),
+                          (-np.inf, 2.5)])
+def test_powm1_return_nan(x, y):
+    # Test cases where the expected return value is nan.
+    p = powm1(x, y)
+    assert np.isnan(p)

.venv/Lib/site-packages/scipy/special/tests/test_precompute_expn_asy.py

@@ -0,0 +1,24 @@
+from numpy.testing import assert_equal
+
+from scipy.special._testutils import check_version, MissingModule
+from scipy.special._precompute.expn_asy import generate_A
+
+try:
+    import sympy
+    from sympy import Poly
+except ImportError:
+    sympy = MissingModule("sympy")
+
+
+@check_version(sympy, "1.0")
+def test_generate_A():
+    # Data from DLMF 8.20.5
+    x = sympy.symbols('x')
+    Astd = [Poly(1, x),
+            Poly(1, x),
+            Poly(1 - 2*x),
+            Poly(1 - 8*x + 6*x**2)]
+    Ares = generate_A(len(Astd))
+
+    for p, q in zip(Astd, Ares):
+        assert_equal(p, q)

.venv/Lib/site-packages/scipy/special/tests/test_precompute_gammainc.py

@@ -0,0 +1,108 @@
+import pytest
+
+from scipy.special._testutils import MissingModule, check_version
+from scipy.special._mptestutils import (
+    Arg, IntArg, mp_assert_allclose, assert_mpmath_equal)
+from scipy.special._precompute.gammainc_asy import (
+    compute_g, compute_alpha, compute_d)
+from scipy.special._precompute.gammainc_data import gammainc, gammaincc
+
+try:
+    import sympy
+except ImportError:
+    sympy = MissingModule('sympy')
+
+try:
+    import mpmath as mp
+except ImportError:
+    mp = MissingModule('mpmath')
+
+
+@check_version(mp, '0.19')
+def test_g():
+    # Test data for the g_k. See DLMF 5.11.4.
+    with mp.workdps(30):
+        g = [mp.mpf(1), mp.mpf(1)/12, mp.mpf(1)/288,
+             -mp.mpf(139)/51840, -mp.mpf(571)/2488320,
+             mp.mpf(163879)/209018880, mp.mpf(5246819)/75246796800]
+        mp_assert_allclose(compute_g(7), g)
+
+
+@pytest.mark.slow
+@check_version(mp, '0.19')
+@check_version(sympy, '0.7')
+@pytest.mark.xfail_on_32bit("rtol only 2e-11, see gh-6938")
+def test_alpha():
+    # Test data for the alpha_k. See DLMF 8.12.14.
+    with mp.workdps(30):
+        alpha = [mp.mpf(0), mp.mpf(1), mp.mpf(1)/3, mp.mpf(1)/36,
+                 -mp.mpf(1)/270, mp.mpf(1)/4320, mp.mpf(1)/17010,
+                 -mp.mpf(139)/5443200, mp.mpf(1)/204120]
+        mp_assert_allclose(compute_alpha(9), alpha)
+
+
+@pytest.mark.xslow
+@check_version(mp, '0.19')
+@check_version(sympy, '0.7')
+def test_d():
+    # Compare the d_{k, n} to the results in appendix F of [1].
+    #
+    # Sources
+    # -------
+    # [1] DiDonato and Morris, Computation of the Incomplete Gamma
+    #     Function Ratios and their Inverse, ACM Transactions on
+    #     Mathematical Software, 1986.
+
+    with mp.workdps(50):
+        dataset = [(0, 0, -mp.mpf('0.333333333333333333333333333333')),
+                   (0, 12, mp.mpf('0.102618097842403080425739573227e-7')),
+                   (1, 0, -mp.mpf('0.185185185185185185185185185185e-2')),
+                   (1, 12, mp.mpf('0.119516285997781473243076536700e-7')),
+                   (2, 0, mp.mpf('0.413359788359788359788359788360e-2')),
+                   (2, 12, -mp.mpf('0.140925299108675210532930244154e-7')),
+                   (3, 0, mp.mpf('0.649434156378600823045267489712e-3')),
+                   (3, 12, -mp.mpf('0.191111684859736540606728140873e-7')),
+                   (4, 0, -mp.mpf('0.861888290916711698604702719929e-3')),
+                   (4, 12, mp.mpf('0.288658297427087836297341274604e-7')),
+                   (5, 0, -mp.mpf('0.336798553366358150308767592718e-3')),
+                   (5, 12, mp.mpf('0.482409670378941807563762631739e-7')),
+                   (6, 0, mp.mpf('0.531307936463992223165748542978e-3')),
+                   (6, 12, -mp.mpf('0.882860074633048352505085243179e-7')),
+                   (7, 0, mp.mpf('0.344367606892377671254279625109e-3')),
+                   (7, 12, -mp.mpf('0.175629733590604619378669693914e-6')),
+                   (8, 0, -mp.mpf('0.652623918595309418922034919727e-3')),
+                   (8, 12, mp.mpf('0.377358774161109793380344937299e-6')),
+                   (9, 0, -mp.mpf('0.596761290192746250124390067179e-3')),
+                   (9, 12, mp.mpf('0.870823417786464116761231237189e-6'))]
+        d = compute_d(10, 13)
+        res = [d[k][n] for k, n, std in dataset]
+        std = [x[2] for x in dataset]
+        mp_assert_allclose(res, std)
+
+
+@check_version(mp, '0.19')
+def test_gammainc():
+    # Quick check that the gammainc in
+    # special._precompute.gammainc_data agrees with mpmath's
+    # gammainc.
+    assert_mpmath_equal(gammainc,
+                        lambda a, x: mp.gammainc(a, b=x, regularized=True),
+                        [Arg(0, 100, inclusive_a=False), Arg(0, 100)],
+                        nan_ok=False, rtol=1e-17, n=50, dps=50)
+
+
+@pytest.mark.xslow
+@check_version(mp, '0.19')
+def test_gammaincc():
+    # Check that the gammaincc in special._precompute.gammainc_data
+    # agrees with mpmath's gammainc.
+    assert_mpmath_equal(lambda a, x: gammaincc(a, x, dps=1000),
+                        lambda a, x: mp.gammainc(a, a=x, regularized=True),
+                        [Arg(20, 100), Arg(20, 100)],
+                        nan_ok=False, rtol=1e-17, n=50, dps=1000)
+
+    # Test the fast integer path
+    assert_mpmath_equal(gammaincc,
+                        lambda a, x: mp.gammainc(a, a=x, regularized=True),
+                        [IntArg(1, 100), Arg(0, 100)],
+                        nan_ok=False, rtol=1e-17, n=50, dps=50)

.venv/Lib/site-packages/scipy/special/tests/test_precompute_utils.py

@@ -0,0 +1,36 @@
+import pytest
+
+from scipy.special._testutils import MissingModule, check_version
+from scipy.special._mptestutils import mp_assert_allclose
+from scipy.special._precompute.utils import lagrange_inversion
+
+try:
+    import sympy
+except ImportError:
+    sympy = MissingModule('sympy')
+
+try:
+    import mpmath as mp
+except ImportError:
+    mp = MissingModule('mpmath')
+
+
+@pytest.mark.slow
+@check_version(sympy, '0.7')
+@check_version(mp, '0.19')
+class TestInversion:
+    @pytest.mark.xfail_on_32bit("rtol only 2e-9, see gh-6938")
+    def test_log(self):
+        with mp.workdps(30):
+            logcoeffs = mp.taylor(lambda x: mp.log(1 + x), 0, 10)
+            expcoeffs = mp.taylor(lambda x: mp.exp(x) - 1, 0, 10)
+            invlogcoeffs = lagrange_inversion(logcoeffs)
+            mp_assert_allclose(invlogcoeffs, expcoeffs)
+
+    @pytest.mark.xfail_on_32bit("rtol only 1e-15, see gh-6938")
+    def test_sin(self):
+        with mp.workdps(30):
+            sincoeffs = mp.taylor(mp.sin, 0, 10)
+            asincoeffs = mp.taylor(mp.asin, 0, 10)
+            invsincoeffs = lagrange_inversion(sincoeffs)
+            mp_assert_allclose(invsincoeffs, asincoeffs, atol=1e-30)

.venv/Lib/site-packages/scipy/special/tests/test_round.py

@@ -0,0 +1,16 @@
+import numpy as np
+import pytest
+
+from scipy.special import _test_internal
+
+
+@pytest.mark.skipif(not _test_internal.have_fenv(), reason="no fenv()")
+def test_add_round_up():
+    np.random.seed(1234)
+    _test_internal.test_add_round(10**5, 'up')
+
+
+@pytest.mark.skipif(not _test_internal.have_fenv(), reason="no fenv()")
+def test_add_round_down():
+    np.random.seed(1234)
+    _test_internal.test_add_round(10**5, 'down')

.venv/Lib/site-packages/scipy/special/tests/test_sf_error.py

@@ -0,0 +1,134 @@
+import sys
+import warnings
+
+import numpy as np
+from numpy.testing import assert_, assert_equal, IS_PYPY
+import pytest
+from pytest import raises as assert_raises
+
+import scipy.special as sc
+from scipy.special._ufuncs import _sf_error_test_function
+
+_sf_error_code_map = {
+    # skip 'ok'
+    'singular': 1,
+    'underflow': 2,
+    'overflow': 3,
+    'slow': 4,
+    'loss': 5,
+    'no_result': 6,
+    'domain': 7,
+    'arg': 8,
+    'other': 9
+}
+
+_sf_error_actions = [
+    'ignore',
+    'warn',
+    'raise'
+]
+
+
+def _check_action(fun, args, action):
+    # TODO: special expert should correct
+    # the coercion at the true location?
+    args = np.asarray(args, dtype=np.dtype("long"))
+    if action == 'warn':
+        with pytest.warns(sc.SpecialFunctionWarning):
+            fun(*args)
+    elif action == 'raise':
+        with assert_raises(sc.SpecialFunctionError):
+            fun(*args)
+    else:
+        # action == 'ignore', make sure there are no warnings/exceptions
+        with warnings.catch_warnings():
+            warnings.simplefilter("error")
+            fun(*args)
+
+
+def test_geterr():
+    err = sc.geterr()
+    for key, value in err.items():
+        assert_(key in _sf_error_code_map)
+        assert_(value in _sf_error_actions)
+
+
+def test_seterr():
+    entry_err = sc.geterr()
+    try:
+        for category, error_code in _sf_error_code_map.items():
+            for action in _sf_error_actions:
+                geterr_olderr = sc.geterr()
+                seterr_olderr = sc.seterr(**{category: action})
+                assert_(geterr_olderr == seterr_olderr)
+                newerr = sc.geterr()
+                assert_(newerr[category] == action)
+                geterr_olderr.pop(category)
+                newerr.pop(category)
+                assert_(geterr_olderr == newerr)
+                _check_action(_sf_error_test_function, (error_code,), action)
+    finally:
+        sc.seterr(**entry_err)
+
+
+@pytest.mark.skipif(IS_PYPY, reason="Test not meaningful on PyPy")
+def test_sf_error_special_refcount():
+    # Regression test for gh-16233.
+    # Check that the reference count of scipy.special is not increased
+    # when a SpecialFunctionError is raised.
+    refcount_before = sys.getrefcount(sc)
+    with sc.errstate(all='raise'):
+        with pytest.raises(sc.SpecialFunctionError, match='domain error'):
+            sc.ndtri(2.0)
+    refcount_after = sys.getrefcount(sc)
+    assert refcount_after == refcount_before
+
+
+def test_errstate_pyx_basic():
+    olderr = sc.geterr()
+    with sc.errstate(singular='raise'):
+        with assert_raises(sc.SpecialFunctionError):
+            sc.loggamma(0)
+    assert_equal(olderr, sc.geterr())
+
+
+def test_errstate_c_basic():
+    olderr = sc.geterr()
+    with sc.errstate(domain='raise'):
+        with assert_raises(sc.SpecialFunctionError):
+            sc.spence(-1)
+    assert_equal(olderr, sc.geterr())
+
+
+def test_errstate_cpp_basic():
+    olderr = sc.geterr()
+    with sc.errstate(underflow='raise'):
+        with assert_raises(sc.SpecialFunctionError):
+            sc.wrightomega(-1000)
+    assert_equal(olderr, sc.geterr())
+
+
+def test_errstate_cpp_scipy_special():
+    olderr = sc.geterr()
+    with sc.errstate(singular='raise'):
+        with assert_raises(sc.SpecialFunctionError):
+            sc.lambertw(0, 1)
+    assert_equal(olderr, sc.geterr())
+
+
+def test_errstate():
+    for category, error_code in _sf_error_code_map.items():
+        for action in _sf_error_actions:
+            olderr = sc.geterr()
+            with sc.errstate(**{category: action}):
+                _check_action(_sf_error_test_function, (error_code,), action)
+            assert_equal(olderr, sc.geterr())
+
+
+def test_errstate_all_but_one():
+    olderr = sc.geterr()
+    with sc.errstate(all='raise', singular='ignore'):
+        sc.gammaln(0)
+        with assert_raises(sc.SpecialFunctionError):
+            sc.spence(-1.0)
+    assert_equal(olderr, sc.geterr())

.venv/Lib/site-packages/scipy/special/tests/test_sici.py

@@ -0,0 +1,36 @@
+import numpy as np
+
+import scipy.special as sc
+from scipy.special._testutils import FuncData
+
+
+def test_sici_consistency():
+    # Make sure the implementation of sici for real arguments agrees
+    # with the implementation of sici for complex arguments.
+
+    # On the negative real axis Cephes drops the imaginary part in ci
+    def sici(x):
+        si, ci = sc.sici(x + 0j)
+        return si.real, ci.real
+    
+    x = np.r_[-np.logspace(8, -30, 200), 0, np.logspace(-30, 8, 200)]
+    si, ci = sc.sici(x)
+    dataset = np.column_stack((x, si, ci))
+    FuncData(sici, dataset, 0, (1, 2), rtol=1e-12).check()
+
+
+def test_shichi_consistency():
+    # Make sure the implementation of shichi for real arguments agrees
+    # with the implementation of shichi for complex arguments.
+
+    # On the negative real axis Cephes drops the imaginary part in chi
+    def shichi(x):
+        shi, chi = sc.shichi(x + 0j)
+        return shi.real, chi.real
+
+    # Overflow happens quickly, so limit range
+    x = np.r_[-np.logspace(np.log10(700), -30, 200), 0,
+              np.logspace(-30, np.log10(700), 200)]
+    shi, chi = sc.shichi(x)
+    dataset = np.column_stack((x, shi, chi))
+    FuncData(shichi, dataset, 0, (1, 2), rtol=1e-14).check()

.venv/Lib/site-packages/scipy/special/tests/test_specfun.py

@@ -0,0 +1,36 @@
+"""
+Various made-up tests to hit different branches of the code in specfun.c
+"""
+
+import numpy as np
+from numpy.testing import assert_allclose
+from scipy import special
+
+
+def test_cchg_branches():
+    res = special.hyp1f1(0.1, 1, 7.0-24.0j)
+    assert_allclose(res, (-3.7659844658568016+4.970311359851648j))
+
+
+def test_cva2_cv0_branches():
+    res, resp = special.mathieu_cem([40, 129], [13, 14], [30, 45])
+    assert_allclose(res, np.array([-0.3741211, 0.74441928]))
+    assert_allclose(resp, np.array([-37.02872758, -86.13549877]))
+
+    res, resp = special.mathieu_sem([40, 129], [13, 14], [30, 45])
+    assert_allclose(res, np.array([0.92955551, 0.66771207]))
+    assert_allclose(resp, np.array([-14.91073448, 96.02954185]))
+
+
+def test_chgm_branches():
+    res = special.eval_genlaguerre(-3.2, 3, 2.5)
+    assert_allclose(res, -0.7077721935779854)
+
+
+def test_hygfz_branches():
+    """(z == 1.0) && (c-a-b > 0.0)"""
+    res = special.hyp2f1(1.5, 2.5, 4.5, 1.+0.j)
+    assert_allclose(res, 10.30835089459151+0j)
+    """(cabs(z+1) < eps) && (fabs(c-a+b - 1.0) < eps)"""
+    res = special.hyp2f1(5+5e-16, 2, 2, -1.0 + 5e-16j)
+    assert_allclose(res, 0.031249999999999986+3.9062499999999994e-17j)

.venv/Lib/site-packages/scipy/special/tests/test_spence.py

@@ -0,0 +1,32 @@
+import numpy as np
+from numpy import sqrt, log, pi
+from scipy.special._testutils import FuncData
+from scipy.special import spence
+
+
+def test_consistency():
+    # Make sure the implementation of spence for real arguments
+    # agrees with the implementation of spence for imaginary arguments.
+
+    x = np.logspace(-30, 300, 200)
+    dataset = np.vstack((x + 0j, spence(x))).T
+    FuncData(spence, dataset, 0, 1, rtol=1e-14).check()
+
+
+def test_special_points():
+    # Check against known values of Spence's function.
+
+    phi = (1 + sqrt(5))/2
+    dataset = [(1, 0),
+               (2, -pi**2/12),
+               (0.5, pi**2/12 - log(2)**2/2),
+               (0, pi**2/6),
+               (-1, pi**2/4 - 1j*pi*log(2)),
+               ((-1 + sqrt(5))/2, pi**2/15 - log(phi)**2),
+               ((3 - sqrt(5))/2, pi**2/10 - log(phi)**2),
+               (phi, -pi**2/15 + log(phi)**2/2),
+               # Corrected from Zagier, "The Dilogarithm Function"
+               ((3 + sqrt(5))/2, -pi**2/10 - log(phi)**2)]
+
+    dataset = np.asarray(dataset)
+    FuncData(spence, dataset, 0, 1, rtol=1e-14).check()

.venv/Lib/site-packages/scipy/special/tests/test_spfun_stats.py

@@ -0,0 +1,61 @@
+import numpy as np
+from numpy.testing import (assert_array_equal,
+        assert_array_almost_equal_nulp, assert_almost_equal)
+from pytest import raises as assert_raises
+
+from scipy.special import gammaln, multigammaln
+
+
+class TestMultiGammaLn:
+
+    def test1(self):
+        # A test of the identity
+        #     Gamma_1(a) = Gamma(a)
+        np.random.seed(1234)
+        a = np.abs(np.random.randn())
+        assert_array_equal(multigammaln(a, 1), gammaln(a))
+
+    def test2(self):
+        # A test of the identity
+        #     Gamma_2(a) = sqrt(pi) * Gamma(a) * Gamma(a - 0.5)
+        a = np.array([2.5, 10.0])
+        result = multigammaln(a, 2)
+        expected = np.log(np.sqrt(np.pi)) + gammaln(a) + gammaln(a - 0.5)
+        assert_almost_equal(result, expected)
+
+    def test_bararg(self):
+        assert_raises(ValueError, multigammaln, 0.5, 1.2)
+
+
+def _check_multigammaln_array_result(a, d):
+    # Test that the shape of the array returned by multigammaln
+    # matches the input shape, and that all the values match
+    # the value computed when multigammaln is called with a scalar.
+    result = multigammaln(a, d)
+    assert_array_equal(a.shape, result.shape)
+    a1 = a.ravel()
+    result1 = result.ravel()
+    for i in range(a.size):
+        assert_array_almost_equal_nulp(result1[i], multigammaln(a1[i], d))
+
+
+def test_multigammaln_array_arg():
+    # Check that the array returned by multigammaln has the correct
+    # shape and contains the correct values.  The cases have arrays
+    # with several different shapes.
+    # The cases include a regression test for ticket #1849
+    # (a = np.array([2.0]), an array with a single element).
+    np.random.seed(1234)
+
+    cases = [
+        # a, d
+        (np.abs(np.random.randn(3, 2)) + 5, 5),
+        (np.abs(np.random.randn(1, 2)) + 5, 5),
+        (np.arange(10.0, 18.0).reshape(2, 2, 2), 3),
+        (np.array([2.0]), 3),
+        (np.float64(2.0), 3),
+    ]
+
+    for a, d in cases:
+        _check_multigammaln_array_result(a, d)
+

.venv/Lib/site-packages/scipy/special/tests/test_sph_harm.py

@@ -0,0 +1,37 @@
+import numpy as np
+from numpy.testing import assert_allclose
+import scipy.special as sc
+
+
+def test_first_harmonics():
+    # Test against explicit representations of the first four
+    # spherical harmonics which use `theta` as the azimuthal angle,
+    # `phi` as the polar angle, and include the Condon-Shortley
+    # phase.
+
+    # Notation is Ymn
+    def Y00(theta, phi):
+        return 0.5*np.sqrt(1/np.pi)
+
+    def Yn11(theta, phi):
+        return 0.5*np.sqrt(3/(2*np.pi))*np.exp(-1j*theta)*np.sin(phi)
+
+    def Y01(theta, phi):
+        return 0.5*np.sqrt(3/np.pi)*np.cos(phi)
+
+    def Y11(theta, phi):
+        return -0.5*np.sqrt(3/(2*np.pi))*np.exp(1j*theta)*np.sin(phi)
+
+    harms = [Y00, Yn11, Y01, Y11]
+    m = [0, -1, 0, 1]
+    n = [0, 1, 1, 1]
+
+    theta = np.linspace(0, 2*np.pi)
+    phi = np.linspace(0, np.pi)
+    theta, phi = np.meshgrid(theta, phi)
+
+    for harm, m, n in zip(harms, m, n):
+        assert_allclose(sc.sph_harm(m, n, theta, phi),
+                        harm(theta, phi),
+                        rtol=1e-15, atol=1e-15,
+                        err_msg=f"Y^{m}_{n} incorrect")

.venv/Lib/site-packages/scipy/special/tests/test_spherical_bessel.py

@@ -0,0 +1,385 @@
+#
+# Tests of spherical Bessel functions.
+#
+import numpy as np
+from numpy.testing import (assert_almost_equal, assert_allclose,
+                           assert_array_almost_equal, suppress_warnings)
+import pytest
+from numpy import sin, cos, sinh, cosh, exp, inf, nan, r_, pi
+
+from scipy.special import spherical_jn, spherical_yn, spherical_in, spherical_kn
+from scipy.integrate import quad
+
+
+class TestSphericalJn:
+    def test_spherical_jn_exact(self):
+        # https://dlmf.nist.gov/10.49.E3
+        # Note: exact expression is numerically stable only for small
+        # n or z >> n.
+        x = np.array([0.12, 1.23, 12.34, 123.45, 1234.5])
+        assert_allclose(spherical_jn(2, x),
+                        (-1/x + 3/x**3)*sin(x) - 3/x**2*cos(x))
+
+    def test_spherical_jn_recurrence_complex(self):
+        # https://dlmf.nist.gov/10.51.E1
+        n = np.array([1, 2, 3, 7, 12])
+        x = 1.1 + 1.5j
+        assert_allclose(spherical_jn(n - 1, x) + spherical_jn(n + 1, x),
+                        (2*n + 1)/x*spherical_jn(n, x))
+
+    def test_spherical_jn_recurrence_real(self):
+        # https://dlmf.nist.gov/10.51.E1
+        n = np.array([1, 2, 3, 7, 12])
+        x = 0.12
+        assert_allclose(spherical_jn(n - 1, x) + spherical_jn(n + 1,x),
+                        (2*n + 1)/x*spherical_jn(n, x))
+
+    def test_spherical_jn_inf_real(self):
+        # https://dlmf.nist.gov/10.52.E3
+        n = 6
+        x = np.array([-inf, inf])
+        assert_allclose(spherical_jn(n, x), np.array([0, 0]))
+
+    def test_spherical_jn_inf_complex(self):
+        # https://dlmf.nist.gov/10.52.E3
+        n = 7
+        x = np.array([-inf + 0j, inf + 0j, inf*(1+1j)])
+        with suppress_warnings() as sup:
+            sup.filter(RuntimeWarning, "invalid value encountered in multiply")
+            assert_allclose(spherical_jn(n, x), np.array([0, 0, inf*(1+1j)]))
+
+    def test_spherical_jn_large_arg_1(self):
+        # https://github.com/scipy/scipy/issues/2165
+        # Reference value computed using mpmath, via
+        # besselj(n + mpf(1)/2, z)*sqrt(pi/(2*z))
+        assert_allclose(spherical_jn(2, 3350.507), -0.00029846226538040747)
+
+    def test_spherical_jn_large_arg_2(self):
+        # https://github.com/scipy/scipy/issues/1641
+        # Reference value computed using mpmath, via
+        # besselj(n + mpf(1)/2, z)*sqrt(pi/(2*z))
+        assert_allclose(spherical_jn(2, 10000), 3.0590002633029811e-05)
+
+    def test_spherical_jn_at_zero(self):
+        # https://dlmf.nist.gov/10.52.E1
+        # But note that n = 0 is a special case: j0 = sin(x)/x -> 1
+        n = np.array([0, 1, 2, 5, 10, 100])
+        x = 0
+        assert_allclose(spherical_jn(n, x), np.array([1, 0, 0, 0, 0, 0]))
+
+
+class TestSphericalYn:
+    def test_spherical_yn_exact(self):
+        # https://dlmf.nist.gov/10.49.E5
+        # Note: exact expression is numerically stable only for small
+        # n or z >> n.
+        x = np.array([0.12, 1.23, 12.34, 123.45, 1234.5])
+        assert_allclose(spherical_yn(2, x),
+                        (1/x - 3/x**3)*cos(x) - 3/x**2*sin(x))
+
+    def test_spherical_yn_recurrence_real(self):
+        # https://dlmf.nist.gov/10.51.E1
+        n = np.array([1, 2, 3, 7, 12])
+        x = 0.12
+        assert_allclose(spherical_yn(n - 1, x) + spherical_yn(n + 1,x),
+                        (2*n + 1)/x*spherical_yn(n, x))
+
+    def test_spherical_yn_recurrence_complex(self):
+        # https://dlmf.nist.gov/10.51.E1
+        n = np.array([1, 2, 3, 7, 12])
+        x = 1.1 + 1.5j
+        assert_allclose(spherical_yn(n - 1, x) + spherical_yn(n + 1, x),
+                        (2*n + 1)/x*spherical_yn(n, x))
+
+    def test_spherical_yn_inf_real(self):
+        # https://dlmf.nist.gov/10.52.E3
+        n = 6
+        x = np.array([-inf, inf])
+        assert_allclose(spherical_yn(n, x), np.array([0, 0]))
+
+    def test_spherical_yn_inf_complex(self):
+        # https://dlmf.nist.gov/10.52.E3
+        n = 7
+        x = np.array([-inf + 0j, inf + 0j, inf*(1+1j)])
+        with suppress_warnings() as sup:
+            sup.filter(RuntimeWarning, "invalid value encountered in multiply")
+            assert_allclose(spherical_yn(n, x), np.array([0, 0, inf*(1+1j)]))
+
+    def test_spherical_yn_at_zero(self):
+        # https://dlmf.nist.gov/10.52.E2
+        n = np.array([0, 1, 2, 5, 10, 100])
+        x = 0
+        assert_allclose(spherical_yn(n, x), np.full(n.shape, -inf))
+
+    def test_spherical_yn_at_zero_complex(self):
+        # Consistently with numpy:
+        # >>> -np.cos(0)/0
+        # -inf
+        # >>> -np.cos(0+0j)/(0+0j)
+        # (-inf + nan*j)
+        n = np.array([0, 1, 2, 5, 10, 100])
+        x = 0 + 0j
+        assert_allclose(spherical_yn(n, x), np.full(n.shape, nan))
+
+
+class TestSphericalJnYnCrossProduct:
+    def test_spherical_jn_yn_cross_product_1(self):
+        # https://dlmf.nist.gov/10.50.E3
+        n = np.array([1, 5, 8])
+        x = np.array([0.1, 1, 10])
+        left = (spherical_jn(n + 1, x) * spherical_yn(n, x) -
+                spherical_jn(n, x) * spherical_yn(n + 1, x))
+        right = 1/x**2
+        assert_allclose(left, right)
+
+    def test_spherical_jn_yn_cross_product_2(self):
+        # https://dlmf.nist.gov/10.50.E3
+        n = np.array([1, 5, 8])
+        x = np.array([0.1, 1, 10])
+        left = (spherical_jn(n + 2, x) * spherical_yn(n, x) -
+                spherical_jn(n, x) * spherical_yn(n + 2, x))
+        right = (2*n + 3)/x**3
+        assert_allclose(left, right)
+
+
+class TestSphericalIn:
+    def test_spherical_in_exact(self):
+        # https://dlmf.nist.gov/10.49.E9
+        x = np.array([0.12, 1.23, 12.34, 123.45])
+        assert_allclose(spherical_in(2, x),
+                        (1/x + 3/x**3)*sinh(x) - 3/x**2*cosh(x))
+
+    def test_spherical_in_recurrence_real(self):
+        # https://dlmf.nist.gov/10.51.E4
+        n = np.array([1, 2, 3, 7, 12])
+        x = 0.12
+        assert_allclose(spherical_in(n - 1, x) - spherical_in(n + 1,x),
+                        (2*n + 1)/x*spherical_in(n, x))
+
+    def test_spherical_in_recurrence_complex(self):
+        # https://dlmf.nist.gov/10.51.E1
+        n = np.array([1, 2, 3, 7, 12])
+        x = 1.1 + 1.5j
+        assert_allclose(spherical_in(n - 1, x) - spherical_in(n + 1,x),
+                        (2*n + 1)/x*spherical_in(n, x))
+
+    def test_spherical_in_inf_real(self):
+        # https://dlmf.nist.gov/10.52.E3
+        n = 5
+        x = np.array([-inf, inf])
+        assert_allclose(spherical_in(n, x), np.array([-inf, inf]))
+
+    def test_spherical_in_inf_complex(self):
+        # https://dlmf.nist.gov/10.52.E5
+        # Ideally, i1n(n, 1j*inf) = 0 and i1n(n, (1+1j)*inf) = (1+1j)*inf, but
+        # this appears impossible to achieve because C99 regards any complex
+        # value with at least one infinite  part as a complex infinity, so
+        # 1j*inf cannot be distinguished from (1+1j)*inf.  Therefore, nan is
+        # the correct return value.
+        n = 7
+        x = np.array([-inf + 0j, inf + 0j, inf*(1+1j)])
+        assert_allclose(spherical_in(n, x), np.array([-inf, inf, nan]))
+
+    def test_spherical_in_at_zero(self):
+        # https://dlmf.nist.gov/10.52.E1
+        # But note that n = 0 is a special case: i0 = sinh(x)/x -> 1
+        n = np.array([0, 1, 2, 5, 10, 100])
+        x = 0
+        assert_allclose(spherical_in(n, x), np.array([1, 0, 0, 0, 0, 0]))
+
+
+class TestSphericalKn:
+    def test_spherical_kn_exact(self):
+        # https://dlmf.nist.gov/10.49.E13
+        x = np.array([0.12, 1.23, 12.34, 123.45])
+        assert_allclose(spherical_kn(2, x),
+                        pi/2*exp(-x)*(1/x + 3/x**2 + 3/x**3))
+
+    def test_spherical_kn_recurrence_real(self):
+        # https://dlmf.nist.gov/10.51.E4
+        n = np.array([1, 2, 3, 7, 12])
+        x = 0.12
+        assert_allclose(
+            (-1)**(n - 1)*spherical_kn(n - 1, x) - (-1)**(n + 1)*spherical_kn(n + 1,x),
+            (-1)**n*(2*n + 1)/x*spherical_kn(n, x)
+        )
+
+    def test_spherical_kn_recurrence_complex(self):
+        # https://dlmf.nist.gov/10.51.E4
+        n = np.array([1, 2, 3, 7, 12])
+        x = 1.1 + 1.5j
+        assert_allclose(
+            (-1)**(n - 1)*spherical_kn(n - 1, x) - (-1)**(n + 1)*spherical_kn(n + 1,x),
+            (-1)**n*(2*n + 1)/x*spherical_kn(n, x)
+        )
+
+    def test_spherical_kn_inf_real(self):
+        # https://dlmf.nist.gov/10.52.E6
+        n = 5
+        x = np.array([-inf, inf])
+        assert_allclose(spherical_kn(n, x), np.array([-inf, 0]))
+
+    def test_spherical_kn_inf_complex(self):
+        # https://dlmf.nist.gov/10.52.E6
+        # The behavior at complex infinity depends on the sign of the real
+        # part: if Re(z) >= 0, then the limit is 0; if Re(z) < 0, then it's
+        # z*inf.  This distinction cannot be captured, so we return nan.
+        n = 7
+        x = np.array([-inf + 0j, inf + 0j, inf*(1+1j)])
+        assert_allclose(spherical_kn(n, x), np.array([-inf, 0, nan]))
+
+    def test_spherical_kn_at_zero(self):
+        # https://dlmf.nist.gov/10.52.E2
+        n = np.array([0, 1, 2, 5, 10, 100])
+        x = 0
+        assert_allclose(spherical_kn(n, x), np.full(n.shape, inf))
+
+    def test_spherical_kn_at_zero_complex(self):
+        # https://dlmf.nist.gov/10.52.E2
+        n = np.array([0, 1, 2, 5, 10, 100])
+        x = 0 + 0j
+        assert_allclose(spherical_kn(n, x), np.full(n.shape, nan))
+
+
+class SphericalDerivativesTestCase:
+    def fundamental_theorem(self, n, a, b):
+        integral, tolerance = quad(lambda z: self.df(n, z), a, b)
+        assert_allclose(integral,
+                        self.f(n, b) - self.f(n, a),
+                        atol=tolerance)
+
+    @pytest.mark.slow
+    def test_fundamental_theorem_0(self):
+        self.fundamental_theorem(0, 3.0, 15.0)
+
+    @pytest.mark.slow
+    def test_fundamental_theorem_7(self):
+        self.fundamental_theorem(7, 0.5, 1.2)
+
+
+class TestSphericalJnDerivatives(SphericalDerivativesTestCase):
+    def f(self, n, z):
+        return spherical_jn(n, z)
+
+    def df(self, n, z):
+        return spherical_jn(n, z, derivative=True)
+
+    def test_spherical_jn_d_zero(self):
+        n = np.array([0, 1, 2, 3, 7, 15])
+        assert_allclose(spherical_jn(n, 0, derivative=True),
+                        np.array([0, 1/3, 0, 0, 0, 0]))
+
+
+class TestSphericalYnDerivatives(SphericalDerivativesTestCase):
+    def f(self, n, z):
+        return spherical_yn(n, z)
+
+    def df(self, n, z):
+        return spherical_yn(n, z, derivative=True)
+
+
+class TestSphericalInDerivatives(SphericalDerivativesTestCase):
+    def f(self, n, z):
+        return spherical_in(n, z)
+
+    def df(self, n, z):
+        return spherical_in(n, z, derivative=True)
+
+    def test_spherical_in_d_zero(self):
+        n = np.array([0, 1, 2, 3, 7, 15])
+        spherical_in(n, 0, derivative=False)
+        assert_allclose(spherical_in(n, 0, derivative=True),
+                        np.array([0, 1/3, 0, 0, 0, 0]))
+
+
+class TestSphericalKnDerivatives(SphericalDerivativesTestCase):
+    def f(self, n, z):
+        return spherical_kn(n, z)
+
+    def df(self, n, z):
+        return spherical_kn(n, z, derivative=True)
+
+
+class TestSphericalOld:
+    # These are tests from the TestSpherical class of test_basic.py,
+    # rewritten to use spherical_* instead of sph_* but otherwise unchanged.
+
+    def test_sph_in(self):
+        # This test reproduces test_basic.TestSpherical.test_sph_in.
+        i1n = np.empty((2,2))
+        x = 0.2
+
+        i1n[0][0] = spherical_in(0, x)
+        i1n[0][1] = spherical_in(1, x)
+        i1n[1][0] = spherical_in(0, x, derivative=True)
+        i1n[1][1] = spherical_in(1, x, derivative=True)
+
+        inp0 = (i1n[0][1])
+        inp1 = (i1n[0][0] - 2.0/0.2 * i1n[0][1])
+        assert_array_almost_equal(i1n[0],np.array([1.0066800127054699381,
+                                                0.066933714568029540839]),12)
+        assert_array_almost_equal(i1n[1],[inp0,inp1],12)
+
+    def test_sph_in_kn_order0(self):
+        x = 1.
+        sph_i0 = np.empty((2,))
+        sph_i0[0] = spherical_in(0, x)
+        sph_i0[1] = spherical_in(0, x, derivative=True)
+        sph_i0_expected = np.array([np.sinh(x)/x,
+                                    np.cosh(x)/x-np.sinh(x)/x**2])
+        assert_array_almost_equal(r_[sph_i0], sph_i0_expected)
+
+        sph_k0 = np.empty((2,))
+        sph_k0[0] = spherical_kn(0, x)
+        sph_k0[1] = spherical_kn(0, x, derivative=True)
+        sph_k0_expected = np.array([0.5*pi*exp(-x)/x,
+                                    -0.5*pi*exp(-x)*(1/x+1/x**2)])
+        assert_array_almost_equal(r_[sph_k0], sph_k0_expected)
+
+    def test_sph_jn(self):
+        s1 = np.empty((2,3))
+        x = 0.2
+
+        s1[0][0] = spherical_jn(0, x)
+        s1[0][1] = spherical_jn(1, x)
+        s1[0][2] = spherical_jn(2, x)
+        s1[1][0] = spherical_jn(0, x, derivative=True)
+        s1[1][1] = spherical_jn(1, x, derivative=True)
+        s1[1][2] = spherical_jn(2, x, derivative=True)
+
+        s10 = -s1[0][1]
+        s11 = s1[0][0]-2.0/0.2*s1[0][1]
+        s12 = s1[0][1]-3.0/0.2*s1[0][2]
+        assert_array_almost_equal(s1[0],[0.99334665397530607731,
+                                      0.066400380670322230863,
+                                      0.0026590560795273856680],12)
+        assert_array_almost_equal(s1[1],[s10,s11,s12],12)
+
+    def test_sph_kn(self):
+        kn = np.empty((2,3))
+        x = 0.2
+
+        kn[0][0] = spherical_kn(0, x)
+        kn[0][1] = spherical_kn(1, x)
+        kn[0][2] = spherical_kn(2, x)
+        kn[1][0] = spherical_kn(0, x, derivative=True)
+        kn[1][1] = spherical_kn(1, x, derivative=True)
+        kn[1][2] = spherical_kn(2, x, derivative=True)
+
+        kn0 = -kn[0][1]
+        kn1 = -kn[0][0]-2.0/0.2*kn[0][1]
+        kn2 = -kn[0][1]-3.0/0.2*kn[0][2]
+        assert_array_almost_equal(kn[0],[6.4302962978445670140,
+                                         38.581777787067402086,
+                                         585.15696310385559829],12)
+        assert_array_almost_equal(kn[1],[kn0,kn1,kn2],9)
+
+    def test_sph_yn(self):
+        sy1 = spherical_yn(2, 0.2)
+        sy2 = spherical_yn(0, 0.2)
+        assert_almost_equal(sy1,-377.52483,5)  # previous values in the system
+        assert_almost_equal(sy2,-4.9003329,5)
+        sphpy = (spherical_yn(0, 0.2) - 2*spherical_yn(2, 0.2))/3
+        sy3 = spherical_yn(1, 0.2, derivative=True)
+        # compare correct derivative val. (correct =-system val).
+        assert_almost_equal(sy3,sphpy,4)

.venv/Lib/site-packages/scipy/special/tests/test_support_alternative_backends.py

@@ -0,0 +1,67 @@
+import pytest
+from hypothesis import given, strategies, reproduce_failure  # noqa: F401
+import hypothesis.extra.numpy as npst
+
+from scipy.special._support_alternative_backends import (get_array_special_func,
+                                                         array_special_func_map)
+from scipy.conftest import array_api_compatible
+from scipy import special
+from scipy._lib._array_api import xp_assert_close
+from scipy._lib.array_api_compat import numpy as np
+
+try:
+    import array_api_strict
+    HAVE_ARRAY_API_STRICT = True
+except ImportError:
+    HAVE_ARRAY_API_STRICT = False
+
+
+@pytest.mark.skipif(not HAVE_ARRAY_API_STRICT,
+                    reason="`array_api_strict` not installed")
+def test_dispatch_to_unrecognize_library():
+    xp = array_api_strict
+    f = get_array_special_func('ndtr', xp=xp, n_array_args=1)
+    x = [1, 2, 3]
+    res = f(xp.asarray(x))
+    ref = xp.asarray(special.ndtr(np.asarray(x)))
+    xp_assert_close(res, ref, xp=xp)
+
+
+@array_api_compatible
+@given(data=strategies.data())
+@pytest.mark.parametrize('f_name_n_args', array_special_func_map.items())
+def test_support_alternative_backends(xp, data, f_name_n_args):
+    f_name, n_args = f_name_n_args
+    f = getattr(special, f_name)
+
+    mbs = npst.mutually_broadcastable_shapes(num_shapes=n_args)
+    shapes, final_shape = data.draw(mbs)
+
+    dtype = data.draw(strategies.sampled_from(['float32', 'float64']))
+    dtype_np = getattr(np, dtype)
+    dtype_xp = getattr(xp, dtype)
+
+    elements = dict(min_value=dtype_np(-10), max_value=dtype_np(10),
+                    allow_subnormal=False)
+    args_np = [np.asarray(data.draw(npst.arrays(dtype_np, shape, elements=elements)))
+               for shape in shapes]
+
+    # `torch.asarray(np.asarray(1.))` produces
+    # TypeError: can't convert np.ndarray of type numpy.object_.
+    # So we extract the scalar from 0d arrays.
+    args_xp = [xp.asarray(arg[()], dtype=dtype_xp) for arg in args_np]
+
+    ref = np.asarray(f(*args_np))
+    res = f(*args_xp)
+
+    eps = np.finfo(dtype).eps
+    # PyTorch seems to struggle with precision near the poles of `gammaln`,
+    # so the tolerance needs to be quite loose (eps**0.2) - see gh-19935.
+    # To compensate, we also check that the root-mean-square error is
+    # less than eps**0.5.
+    ref = xp.asarray(ref, dtype=dtype_xp)
+    xp_assert_close(res, ref, rtol=eps**0.2, atol=eps*10,
+                    check_namespace=True, check_shape=True, check_dtype=True,)
+    xp_assert_close(xp.sqrt(xp.mean(res**2)), xp.sqrt(xp.mean(ref**2)),
+                    rtol=eps**0.5, atol=eps*10,
+                    check_namespace=False, check_shape=False, check_dtype=False,)

.venv/Lib/site-packages/scipy/special/tests/test_trig.py

@@ -0,0 +1,72 @@
+import pytest
+import numpy as np
+from numpy.testing import assert_equal, assert_allclose, suppress_warnings
+
+from scipy.special._ufuncs import _sinpi as sinpi
+from scipy.special._ufuncs import _cospi as cospi
+
+
+def test_integer_real_part():
+    x = np.arange(-100, 101)
+    y = np.hstack((-np.linspace(310, -30, 10), np.linspace(-30, 310, 10)))
+    x, y = np.meshgrid(x, y)
+    z = x + 1j*y
+    # In the following we should be *exactly* right
+    res = sinpi(z)
+    assert_equal(res.real, 0.0)
+    res = cospi(z)
+    assert_equal(res.imag, 0.0)
+
+
+def test_half_integer_real_part():
+    x = np.arange(-100, 101) + 0.5
+    y = np.hstack((-np.linspace(310, -30, 10), np.linspace(-30, 310, 10)))
+    x, y = np.meshgrid(x, y)
+    z = x + 1j*y
+    # In the following we should be *exactly* right
+    res = sinpi(z)
+    assert_equal(res.imag, 0.0)
+    res = cospi(z)
+    assert_equal(res.real, 0.0)
+
+
+@pytest.mark.skip("Temporary skip while gh-19526 is being resolved")
+def test_intermediate_overlow():
+    # Make sure we avoid overflow in situations where cosh/sinh would
+    # overflow but the product with sin/cos would not
+    sinpi_pts = [complex(1 + 1e-14, 227),
+                 complex(1e-35, 250),
+                 complex(1e-301, 445)]
+    # Data generated with mpmath
+    sinpi_std = [complex(-8.113438309924894e+295, -np.inf),
+                 complex(1.9507801934611995e+306, np.inf),
+                 complex(2.205958493464539e+306, np.inf)]
+    with suppress_warnings() as sup:
+        sup.filter(RuntimeWarning, "invalid value encountered in multiply")
+        for p, std in zip(sinpi_pts, sinpi_std):
+            res = sinpi(p)
+            assert_allclose(res.real, std.real)
+            assert_allclose(res.imag, std.imag)
+
+    # Test for cosine, less interesting because cos(0) = 1.
+    p = complex(0.5 + 1e-14, 227)
+    std = complex(-8.113438309924894e+295, -np.inf)
+    with suppress_warnings() as sup:
+        sup.filter(RuntimeWarning, "invalid value encountered in multiply")
+        res = cospi(p)
+        assert_allclose(res.real, std.real)
+        assert_allclose(res.imag, std.imag)
+
+
+def test_zero_sign():
+    y = sinpi(-0.0)
+    assert y == 0.0
+    assert np.signbit(y)
+
+    y = sinpi(0.0)
+    assert y == 0.0
+    assert not np.signbit(y)
+
+    y = cospi(0.5)
+    assert y == 0.0
+    assert not np.signbit(y)

.venv/Lib/site-packages/scipy/special/tests/test_wrightomega.py

@@ -0,0 +1,117 @@
+import pytest
+import numpy as np
+from numpy.testing import assert_, assert_equal, assert_allclose
+
+import scipy.special as sc
+from scipy.special._testutils import assert_func_equal
+
+
+def test_wrightomega_nan():
+    pts = [complex(np.nan, 0),
+           complex(0, np.nan),
+           complex(np.nan, np.nan),
+           complex(np.nan, 1),
+           complex(1, np.nan)]
+    for p in pts:
+        res = sc.wrightomega(p)
+        assert_(np.isnan(res.real))
+        assert_(np.isnan(res.imag))
+
+
+def test_wrightomega_inf_branch():
+    pts = [complex(-np.inf, np.pi/4),
+           complex(-np.inf, -np.pi/4),
+           complex(-np.inf, 3*np.pi/4),
+           complex(-np.inf, -3*np.pi/4)]
+    expected_results = [complex(0.0, 0.0),
+                        complex(0.0, -0.0),
+                        complex(-0.0, 0.0),
+                        complex(-0.0, -0.0)]
+    for p, expected in zip(pts, expected_results):
+        res = sc.wrightomega(p)
+        # We can't use assert_equal(res, expected) because in older versions of
+        # numpy, assert_equal doesn't check the sign of the real and imaginary
+        # parts when comparing complex zeros. It does check the sign when the
+        # arguments are *real* scalars.
+        assert_equal(res.real, expected.real)
+        assert_equal(res.imag, expected.imag)
+
+
+def test_wrightomega_inf():
+    pts = [complex(np.inf, 10),
+           complex(-np.inf, 10),
+           complex(10, np.inf),
+           complex(10, -np.inf)]
+    for p in pts:
+        assert_equal(sc.wrightomega(p), p)
+
+
+def test_wrightomega_singular():
+    pts = [complex(-1.0, np.pi),
+           complex(-1.0, -np.pi)]
+    for p in pts:
+        res = sc.wrightomega(p)
+        assert_equal(res, -1.0)
+        assert_(np.signbit(res.imag) == np.bool_(False))
+
+
+@pytest.mark.parametrize('x, desired', [
+    (-np.inf, 0),
+    (np.inf, np.inf),
+])
+def test_wrightomega_real_infinities(x, desired):
+    assert sc.wrightomega(x) == desired
+
+
+def test_wrightomega_real_nan():
+    assert np.isnan(sc.wrightomega(np.nan))
+
+
+def test_wrightomega_real_series_crossover():
+    desired_error = 2 * np.finfo(float).eps
+    crossover = 1e20
+    x_before_crossover = np.nextafter(crossover, -np.inf)
+    x_after_crossover = np.nextafter(crossover, np.inf)
+    # Computed using Mpmath
+    desired_before_crossover = 99999999999999983569.948
+    desired_after_crossover = 100000000000000016337.948
+    assert_allclose(
+        sc.wrightomega(x_before_crossover),
+        desired_before_crossover,
+        atol=0,
+        rtol=desired_error,
+    )
+    assert_allclose(
+        sc.wrightomega(x_after_crossover),
+        desired_after_crossover,
+        atol=0,
+        rtol=desired_error,
+    )
+
+
+def test_wrightomega_exp_approximation_crossover():
+    desired_error = 2 * np.finfo(float).eps
+    crossover = -50
+    x_before_crossover = np.nextafter(crossover, np.inf)
+    x_after_crossover = np.nextafter(crossover, -np.inf)
+    # Computed using Mpmath
+    desired_before_crossover = 1.9287498479639314876e-22
+    desired_after_crossover = 1.9287498479639040784e-22
+    assert_allclose(
+        sc.wrightomega(x_before_crossover),
+        desired_before_crossover,
+        atol=0,
+        rtol=desired_error,
+    )
+    assert_allclose(
+        sc.wrightomega(x_after_crossover),
+        desired_after_crossover,
+        atol=0,
+        rtol=desired_error,
+    )
+
+
+def test_wrightomega_real_versus_complex():
+    x = np.linspace(-500, 500, 1001)
+    results = sc.wrightomega(x + 0j).real
+    assert_func_equal(sc.wrightomega, results, x, atol=0, rtol=1e-14)

.venv/Lib/site-packages/torch/lib/cublasLt64_12.dll

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:111614ef49d83141a8556416fb9faf9019990f21d40f777fe5d47af65489e90d
+size 538818048

.venv/Lib/site-packages/torch/lib/cudnn_engines_precompiled64_9.dll

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:680625354e9f2ac80d573cc94fc5190848198262557226edd482aaca4ccb9113
+size 588910632