[PATCH v2] math.h: Do use __MATH_TG_BUILTIN_CLASSIFY for C++
H.J. Lu
hjl.tools@gmail.com
Wed Dec 31 05:26:02 GMT 2025
On Wed, Dec 31, 2025 at 6:56 AM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> On Wed, Dec 31, 2025 at 1:25 AM Adhemerval Zanella
> <adhemerval.zanella@linaro.org> wrote:
> >
> > The __MATH_TG_BUILTIN_CLASSIFY macro can not be used in C++
> > mode because it might use __builtin_types_compatible_p, which
> > is a C-only builtin.
>
> The subject says "math.h: Do use __MATH_TG_BUILTIN_CLASSIFY for C++".
>
> > The C++ version uses the built-in; using a different scheme,
> > such as function overloading (like iszero), breaks libstdc++,
> > which expects isinf to be implemented as a macro. The x86
> > test-builtin-denormal.c test has been adjusted to include the
> > relevant gcc/llvm bug information.
> >
> > Checked with bmg targeting x86_64-linux-gnu and on
> > x86_64-linux-gnu and aarch64 with gcc-15 and clang-18.
> > ---
> > include/array_length.h | 4 +
> > math/Makefile | 4 +
> > math/math.h | 38 ++---
> > math/test-math-fpclassify.cc | 149 +++++++++++++++++++
> > math/test-math-isinf.cc | 145 ++++++++++++++++++
> > scripts/sysd-rules.awk | 28 +++-
> > sysdeps/x86/fpu/Makefile | 5 +
> > sysdeps/x86/fpu/test-builtin-denormal-cxx.cc | 1 +
> > sysdeps/x86/fpu/test-builtin-denormal.c | 10 +-
> > 9 files changed, 363 insertions(+), 21 deletions(-)
> > create mode 100644 math/test-math-fpclassify.cc
> > create mode 100644 math/test-math-isinf.cc
> > create mode 100644 sysdeps/x86/fpu/test-builtin-denormal-cxx.cc
> >
> > diff --git a/include/array_length.h b/include/array_length.h
> > index 2a4f09fd37..2f5ab7b219 100644
> > --- a/include/array_length.h
> > +++ b/include/array_length.h
> > @@ -21,6 +21,9 @@
> >
> > /* array_length (VAR) is the number of elements in the array VAR. VAR
> > must evaluate to an array, not a pointer. */
> > +#ifdef __cplusplus
> > +#define array_length(var) (sizeof (var) / sizeof ((var)[0]))
> > +#else
> > #define array_length(var) \
> > (sizeof (var) / sizeof ((var)[0]) \
> > + 0 * sizeof (struct { \
> > @@ -28,6 +31,7 @@
> > (__typeof (var), __typeof (&(var)[0])), \
> > "argument must be an array"); \
> > }))
> > +#endif
> >
> > /* array_end (VAR) is a pointer one past the end of the array VAR.
> > VAR must evaluate to an array, not a pointer. */
> > diff --git a/math/Makefile b/math/Makefile
> > index 8fe2540ce0..fa9193e75b 100644
> > --- a/math/Makefile
> > +++ b/math/Makefile
> > @@ -620,8 +620,10 @@ tests-static += \
> > ifneq (,$(CXX))
> > tests += \
> > test-math-cxx11 \
> > + test-math-fpclassify \
> > test-math-iscanonical \
> > test-math-iseqsig \
> > + test-math-isinf \
> > test-math-isinff \
> > test-math-issignaling \
> > test-math-iszero \
> > @@ -1151,10 +1153,12 @@ CFLAGS-test-signgam-ullong-init-static.c += -std=c99
> >
> > CFLAGS-test-math-cxx11.cc += -std=c++11
> >
> > +CFLAGS-test-math-isinf.cc += -std=gnu++11
> > CFLAGS-test-math-isinff.cc += -std=gnu++11
> > CFLAGS-test-math-iszero.cc += -std=gnu++11
> > CFLAGS-test-math-issignaling.cc += -std=gnu++11
> > CFLAGS-test-math-iscanonical.cc += -std=gnu++11
> > +CFLAGS-test-math-fpclassify.cc += -std=gnu++11
> >
> > CFLAGS-test-iszero-excess-precision.c += -fexcess-precision=standard
> > CFLAGS-test-iseqsig-excess-precision.c += -fexcess-precision=standard
> > diff --git a/math/math.h b/math/math.h
> > index 96b64379f7..05f46b7116 100644
> > --- a/math/math.h
> > +++ b/math/math.h
> > @@ -1173,21 +1173,20 @@ enum
> >
> > /* Return number of classification appropriate for X. */
> > # if ((__GNUC_PREREQ (4,4) && !defined __SUPPORT_SNAN__) \
> > - || __glibc_clang_prereq (2,8)) \
> > - && (!defined __OPTIMIZE_SIZE__ || defined __cplusplus)
> > - /* The check for __cplusplus allows the use of the builtin, even
> > - when optimization for size is on. This is provided for
> > - libstdc++, only to let its configure test work when it is built
> > - with -Os. No further use of this definition of fpclassify is
> > - expected in C++ mode, since libstdc++ provides its own version
> > - of fpclassify in cmath (which undefines fpclassify). */
> > -# define fpclassify(x) \
> > - __MATH_TG_BUILTIN_CLASSIFY ((x), \
> > - __builtin_fpclassify, (FP_NAN, FP_INFINITE, \
> > - FP_NORMAL, FP_SUBNORMAL, \
> > - FP_ZERO, x), \
> > - fpclassify, (x), \
> > - __FP_BUILTIN_FPCLASSIFY_DENORMAL)
> > + || __glibc_clang_prereq (2,8)) \
> > + && !defined __cplusplus
> > +/* The __MATH_TG_BUILTIN_CLASSIFY macro can not be used in C++ mode because
Should we provide a C++ __MATH_TG_BUILTIN_CLASSIFY?
> > + it might use __builtin_types_compatible_p, which is C-only builtin. */
> > +# define fpclassify(x) \
> > + __MATH_TG_BUILTIN_CLASSIFY ((x), \
> > + __builtin_fpclassify, (FP_NAN, FP_INFINITE, \
> > + FP_NORMAL, FP_SUBNORMAL, \
> > + FP_ZERO, x), \
> > + fpclassify, (x), \
> > + __FP_BUILTIN_FPCLASSIFY_DENORMAL)
> > +# elif defined __cplusplus
> > +# define fpclassify(x) __builtin_fpclassify (FP_NAN, FP_INFINITE, FP_NORMAL, \
> > + FP_SUBNORMAL, FP_ZERO, x)
> > # else
> > # define fpclassify(x) __MATH_TG ((x), __fpclassify, (x))
> > # endif
> > @@ -1247,12 +1246,17 @@ enum
> > # define isinf(x) \
> > (__builtin_types_compatible_p (__typeof (x), _Float128) \
> > ? __isinff128 (x) : __builtin_isinf_sign (x))
> > -# elif (__GNUC_PREREQ (4,4) && !defined __SUPPORT_SNAN__) \
> > - || __glibc_clang_prereq (3,7)
> > + /* The __MATH_TG_BUILTIN_CLASSIFY macro can not be used in C++ mode because
> > + it might use __builtin_types_compatible_p, which is C-only builtin. */
> > +# elif ((__GNUC_PREREQ (4,4) && !defined __SUPPORT_SNAN__) \
> > + || __glibc_clang_prereq (3,7)) \
> > + && !defined __cplusplus
> > # define isinf(x) __MATH_TG_BUILTIN_CLASSIFY ((x), \
> > __builtin_isinf_sign, (x), \
> > isinf, (x), \
> > __FP_BUILTIN_ISINF_SIGN_DENORMAL)
> > +# elif defined __cplusplus
> > +# define isinf(x) __builtin_isinf_sign (x)
> > # else
> > # define isinf(x) __MATH_TG ((x), __isinf, (x))
> > # endif
> > diff --git a/math/test-math-fpclassify.cc b/math/test-math-fpclassify.cc
> > new file mode 100644
> > index 0000000000..f8dbf4fe7d
> > --- /dev/null
> > +++ b/math/test-math-fpclassify.cc
> > @@ -0,0 +1,149 @@
> > +/* Test for the C++ implementation of fpclassify.
> > + Copyright (C) 2025 Free Software Foundation, Inc.
> > + This file is part of the GNU C Library.
> > +
> > + The GNU C Library is free software; you can redistribute it and/or
> > + modify it under the terms of the GNU Lesser General Public
> > + License as published by the Free Software Foundation; either
> > + version 2.1 of the License, or (at your option) any later version.
> > +
> > + The GNU C Library is distributed in the hope that it will be useful,
> > + but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > + Lesser General Public License for more details.
> > +
> > + You should have received a copy of the GNU Lesser General Public
> > + License along with the GNU C Library; if not, see
> > + <https://www.gnu.org/licenses/>. */
> > +
> > +#define _GNU_SOURCE 1
> > +#include <math.h>
> > +#include <stdio.h>
> > +#include <limits>
> > +
> > +#include <support/check.h>
> > +
> > +/* Support for _Float128 in std::numeric_limits is limited.
> > + Include ieee754_float128.h and use the bitfields in the union
> > + ieee854_float128.ieee_nan to build corner-case inputs. */
> > +#if __HAVE_DISTINCT_FLOAT128
> > +# include <ieee754_float128.h>
> > +#endif
> > +
> > +template <class T>
> > +static void
> > +check_type ()
> > +{
> > + typedef std::numeric_limits<T> limits;
> > + TEST_COMPARE (fpclassify (T{}), FP_ZERO);
> > + TEST_COMPARE (fpclassify (T{0}), FP_ZERO);
> > + TEST_COMPARE (fpclassify (T{-0.0}), FP_ZERO);
> > + TEST_COMPARE (fpclassify (T{1}), FP_NORMAL);
> > + TEST_COMPARE (fpclassify (T{-1}), FP_NORMAL);
> > + TEST_COMPARE (fpclassify (limits::min ()), FP_NORMAL);
> > + TEST_COMPARE (fpclassify (-limits::min ()), FP_NORMAL);
> > + TEST_COMPARE (fpclassify (limits::max ()), FP_NORMAL);
> > + TEST_COMPARE (fpclassify (-limits::max ()), FP_NORMAL);
> > + if (limits::has_infinity)
> > + {
> > + TEST_COMPARE (fpclassify (limits::infinity ()), FP_INFINITE);
> > + TEST_COMPARE (fpclassify (-limits::infinity ()), FP_INFINITE);
> > + }
> > + TEST_COMPARE (fpclassify (limits::epsilon ()), FP_NORMAL);
> > + TEST_COMPARE (fpclassify (-limits::epsilon ()), FP_NORMAL);
> > + if (limits::has_quiet_NaN)
> > + TEST_COMPARE (fpclassify (limits::quiet_NaN ()), FP_NAN);
> > + if (limits::has_signaling_NaN)
> > + TEST_COMPARE (fpclassify (limits::signaling_NaN ()), FP_NAN);
> > + if (limits::has_signaling_NaN)
> > + TEST_COMPARE (fpclassify (limits::signaling_NaN ()), FP_NAN);
> > + TEST_COMPARE (fpclassify (limits::denorm_min ()),
> > + std::numeric_limits<T>::has_denorm == std::denorm_absent
> > + ? FP_NORMAL : FP_SUBNORMAL);
> > + TEST_COMPARE (fpclassify (-limits::denorm_min ()),
> > + std::numeric_limits<T>::has_denorm == std::denorm_absent
> > + ? FP_NORMAL : FP_SUBNORMAL);
> > +}
> > +
> > +#if __HAVE_DISTINCT_FLOAT128
> > +static void
> > +check_float128 ()
> > +{
> > + ieee854_float128 q;
> > +
> > + q.d = 0.0Q;
> > + TEST_COMPARE (fpclassify (q.d), FP_ZERO);
> > + q.d = -0.0Q;
> > + TEST_COMPARE (fpclassify (q.d), FP_ZERO);
> > + q.d = 1.0Q;
> > + TEST_COMPARE (fpclassify (q.d), FP_NORMAL);
> > + q.d = -1.0Q;
> > + TEST_COMPARE (fpclassify (q.d), FP_NORMAL);
> > +
> > + /* Normal min. */
> > + q.ieee.negative = 0;
> > + q.ieee.exponent = 0x0001;
> > + q.ieee.mantissa0 = 0x0000;
> > + q.ieee.mantissa1 = 0x00000000;
> > + q.ieee.mantissa2 = 0x00000000;
> > + q.ieee.mantissa3 = 0x00000000;
> > + TEST_COMPARE (fpclassify (q.d), FP_NORMAL);
> > + q.ieee.negative = 1;
> > + TEST_COMPARE (fpclassify (q.d), FP_NORMAL);
> > +
> > + /* Normal max. */
> > + q.ieee.negative = 0;
> > + q.ieee.exponent = 0x7FFE;
> > + q.ieee.mantissa0 = 0xFFFF;
> > + q.ieee.mantissa1 = 0xFFFFFFFF;
> > + q.ieee.mantissa2 = 0xFFFFFFFF;
> > + q.ieee.mantissa3 = 0xFFFFFFFF;
> > + TEST_COMPARE (fpclassify (q.d), FP_NORMAL);
> > + q.ieee.negative = 1;
> > + TEST_COMPARE (fpclassify (q.d), FP_NORMAL);
> > +
> > + /* Infinity. */
> > + q.ieee.negative = 0;
> > + q.ieee.exponent = 0x7FFF;
> > + q.ieee.mantissa0 = 0x0000;
> > + q.ieee.mantissa1 = 0x00000000;
> > + q.ieee.mantissa2 = 0x00000000;
> > + q.ieee.mantissa3 = 0x00000000;
> > + TEST_COMPARE (fpclassify (q.d), FP_INFINITE);
> > +
> > + /* Quiet NaN. */
> > + q.ieee_nan.quiet_nan = 1;
> > + q.ieee_nan.mantissa0 = 0x0000;
> > + TEST_COMPARE (fpclassify (q.d), FP_NAN);
> > +
> > + /* Signaling NaN. */
> > + q.ieee_nan.quiet_nan = 0;
> > + q.ieee_nan.mantissa0 = 0x4000;
> > + TEST_COMPARE (fpclassify (q.d), FP_NAN);
> > +
> > + /* Denormal min. */
> > + q.ieee.negative = 0;
> > + q.ieee.exponent = 0x0000;
> > + q.ieee.mantissa0 = 0x0000;
> > + q.ieee.mantissa1 = 0x00000000;
> > + q.ieee.mantissa2 = 0x00000000;
> > + q.ieee.mantissa3 = 0x00000001;
> > + TEST_COMPARE (fpclassify (q.d), FP_SUBNORMAL);
> > + q.ieee.negative = 1;
> > + TEST_COMPARE (fpclassify (q.d), FP_SUBNORMAL);
> > +}
> > +#endif
> > +
> > +static int
> > +do_test (void)
> > +{
> > + check_type<float> ();
> > + check_type<double> ();
> > + check_type<long double> ();
> > +#if __HAVE_DISTINCT_FLOAT128
> > + check_float128 ();
> > +#endif
> > + return 0;
> > +}
> > +
> > +#include <support/test-driver.c>
> > diff --git a/math/test-math-isinf.cc b/math/test-math-isinf.cc
> > new file mode 100644
> > index 0000000000..4cdc146583
> > --- /dev/null
> > +++ b/math/test-math-isinf.cc
> > @@ -0,0 +1,145 @@
> > +/* Test for the C++ implementation of isinf.
> > + Copyright (C) 2025 Free Software Foundation, Inc.
> > + This file is part of the GNU C Library.
> > +
> > + The GNU C Library is free software; you can redistribute it and/or
> > + modify it under the terms of the GNU Lesser General Public
> > + License as published by the Free Software Foundation; either
> > + version 2.1 of the License, or (at your option) any later version.
> > +
> > + The GNU C Library is distributed in the hope that it will be useful,
> > + but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > + Lesser General Public License for more details.
> > +
> > + You should have received a copy of the GNU Lesser General Public
> > + License along with the GNU C Library; if not, see
> > + <https://www.gnu.org/licenses/>. */
> > +
> > +#define _GNU_SOURCE 1
> > +#include <math.h>
> > +#include <stdio.h>
> > +#include <limits>
> > +
> > +#include <support/check.h>
> > +
> > +/* Support for _Float128 in std::numeric_limits is limited.
> > + Include ieee754_float128.h and use the bitfields in the union
> > + ieee854_float128.ieee_nan to build corner-case inputs. */
> > +#if __HAVE_DISTINCT_FLOAT128
> > +# include <ieee754_float128.h>
> > +#endif
> > +
> > +template <class T>
> > +static void
> > +check_type ()
> > +{
> > + typedef std::numeric_limits<T> limits;
> > + TEST_COMPARE (isinf (T{}), 0);
> > + TEST_COMPARE (isinf (T{0}), 0);
> > + TEST_COMPARE (isinf (T{-0.0}), 0);
> > + TEST_COMPARE (isinf (T{1}), 0);
> > + TEST_COMPARE (isinf (T{-1}), 0);
> > + TEST_COMPARE (isinf (limits::min ()), 0);
> > + TEST_COMPARE (isinf (-limits::min ()), 0);
> > + TEST_COMPARE (isinf (limits::max ()), 0);
> > + TEST_COMPARE (isinf (-limits::max ()), 0);
> > + if (limits::has_infinity)
> > + {
> > + TEST_COMPARE (isinf (limits::infinity ()), 1);
> > + TEST_COMPARE (isinf (-limits::infinity ()), -1);
> > + }
> > + TEST_COMPARE (isinf (limits::epsilon ()), 0);
> > + TEST_COMPARE (isinf (-limits::epsilon ()), 0);
> > + if (limits::has_quiet_NaN)
> > + TEST_COMPARE (isinf (limits::quiet_NaN ()), 0);
> > + if (limits::has_signaling_NaN)
> > + TEST_COMPARE (isinf (limits::signaling_NaN ()), 0);
> > + if (limits::has_signaling_NaN)
> > + TEST_COMPARE (isinf (limits::signaling_NaN ()), 0);
> > + TEST_COMPARE (isinf (limits::denorm_min ()), 0);
> > + TEST_COMPARE (isinf (-limits::denorm_min ()), 0);
> > +}
> > +
> > +#if __HAVE_DISTINCT_FLOAT128
> > +static void
> > +check_float128 ()
> > +{
> > + ieee854_float128 q;
> > +
> > + q.d = 0.0Q;
> > + TEST_COMPARE (isinf (q.d), 0);
> > + q.d = -0.0Q;
> > + TEST_COMPARE (isinf (q.d), 0);
> > + q.d = 1.0Q;
> > + TEST_COMPARE (isinf (q.d), 0);
> > + q.d = -1.0Q;
> > + TEST_COMPARE (isinf (q.d), 0);
> > +
> > + /* Normal min. */
> > + q.ieee.negative = 0;
> > + q.ieee.exponent = 0x0001;
> > + q.ieee.mantissa0 = 0x0000;
> > + q.ieee.mantissa1 = 0x00000000;
> > + q.ieee.mantissa2 = 0x00000000;
> > + q.ieee.mantissa3 = 0x00000000;
> > + TEST_COMPARE (isinf (q.d), 0);
> > + q.ieee.negative = 1;
> > + TEST_COMPARE (isinf (q.d), 0);
> > +
> > + /* Normal max. */
> > + q.ieee.negative = 0;
> > + q.ieee.exponent = 0x7FFE;
> > + q.ieee.mantissa0 = 0xFFFF;
> > + q.ieee.mantissa1 = 0xFFFFFFFF;
> > + q.ieee.mantissa2 = 0xFFFFFFFF;
> > + q.ieee.mantissa3 = 0xFFFFFFFF;
> > + TEST_COMPARE (isinf (q.d), 0);
> > + q.ieee.negative = 1;
> > + TEST_COMPARE (isinf (q.d), 0);
> > +
> > + /* Infinity. */
> > + q.ieee.negative = 0;
> > + q.ieee.exponent = 0x7FFF;
> > + q.ieee.mantissa0 = 0x0000;
> > + q.ieee.mantissa1 = 0x00000000;
> > + q.ieee.mantissa2 = 0x00000000;
> > + q.ieee.mantissa3 = 0x00000000;
> > + TEST_COMPARE (isinf (q.d), 1);
> > +
> > + /* Quiet NaN. */
> > + q.ieee_nan.quiet_nan = 1;
> > + q.ieee_nan.mantissa0 = 0x0000;
> > + TEST_COMPARE (isinf (q.d), 0);
> > +
> > + /* Signaling NaN. */
> > + q.ieee_nan.quiet_nan = 0;
> > + q.ieee_nan.mantissa0 = 0x4000;
> > + TEST_COMPARE (isinf (q.d), 0);
> > +
> > + /* Denormal min. */
> > + q.ieee.negative = 0;
> > + q.ieee.exponent = 0x0000;
> > + q.ieee.mantissa0 = 0x0000;
> > + q.ieee.mantissa1 = 0x00000000;
> > + q.ieee.mantissa2 = 0x00000000;
> > + q.ieee.mantissa3 = 0x00000001;
> > + TEST_COMPARE (isinf (q.d), 0);
> > + q.ieee.negative = 1;
> > + TEST_COMPARE (isinf (q.d), 0);
> > +}
> > +#endif
> > +
> > +static int
> > +do_test (void)
> > +{
> > + check_type<float> ();
> > + check_type<double> ();
> > + check_type<long double> ();
> > +#if __HAVE_DISTINCT_FLOAT128
> > + check_float128 ();
> > +#endif
> > + return 0;
> > +}
> > +
> > +#include <support/test-driver.c>
> > diff --git a/scripts/sysd-rules.awk b/scripts/sysd-rules.awk
> > index c82e8fd607..cdc892b397 100644
> > --- a/scripts/sysd-rules.awk
> > +++ b/scripts/sysd-rules.awk
> > @@ -35,6 +35,7 @@ BEGIN {
> >
> > for (sysdir_idx = 1; sysdir_idx <= nsysdirs; ++sysdir_idx) {
> > dir = sysdirs[sysdir_idx];
> > + sysdep_dir = dir
> > if (dir !~ /^\//) dir = "$(..)" dir;
> > asm_rules = 1;
> > for (i = 1; i <= ninhibit_asm; ++i) {
> > @@ -43,6 +44,20 @@ BEGIN {
> > break;
> > }
> > }
> > + # Check if there are any *.S files.
> > + if (asm_rules == 1 && system("ls " sysdep_dir "/*.S > /dev/null 2>&1") != 0) {
> > + asm_rules = 0;
> > + }
> > + # Check if there are any *.c files.
> > + c_rules = 1
> > + if (system("ls " sysdep_dir "/*.c > /dev/null 2>&1") != 0) {
> > + c_rules = 0;
> > + }
> > + # Check if there are any *.cc files.
> > + cc_rules = 1
> > + if (system("ls " sysdep_dir "/*.cc > /dev/null 2>&1") != 0) {
> > + cc_rules = 0;
> > + }
> > for (suffix_idx = 1; suffix_idx <= nsuffixes; ++suffix_idx) {
> > o = suffixes[suffix_idx];
> > for (pattern_idx = 1; pattern_idx <= npatterns; ++pattern_idx) {
> > @@ -66,9 +81,16 @@ BEGIN {
> > print target, dep, "$(before-compile)";
> > print "\t$(compile-command.S)" command_suffix;
> > }
> > - dep = dir "/" dep_pattern ".c";
> > - print target, dep, "$(before-compile)";
> > - print "\t$(compile-command.c)" command_suffix;
> > + if (c_rules) {
> > + dep = dir "/" dep_pattern ".c";
> > + print target, dep, "$(before-compile)";
> > + print "\t$(compile-command.c)" command_suffix;
> > + }
> > + if (cc_rules) {
> > + dep = dir "/" dep_pattern ".cc";
> > + print target, dep, "$(before-compile)";
> > + print "\t$(compile-command.cc)" command_suffix;
> > + }
> > }
> > }
> > print "$(inst_includedir)/%.h:", dir "/%.h", "$(+force)";
> > diff --git a/sysdeps/x86/fpu/Makefile b/sysdeps/x86/fpu/Makefile
> > index 8566d3c80f..cb0879dd5f 100644
> > --- a/sysdeps/x86/fpu/Makefile
> > +++ b/sysdeps/x86/fpu/Makefile
> > @@ -11,6 +11,11 @@ tests += \
> > test-fenv-x87 \
> > test-flt-eval-method-sse \
> > # tests
> > +ifneq (,$(CXX))
> > +tests += \
> > + test-builtin-denormal-cxx \
> > +# tests
> > +endif
> > ifeq ($(have-test-cc-cflags-mfpmath-387),yes)
> > tests += test-flt-eval-method-387
> > endif
> > diff --git a/sysdeps/x86/fpu/test-builtin-denormal-cxx.cc b/sysdeps/x86/fpu/test-builtin-denormal-cxx.cc
> > new file mode 100644
> > index 0000000000..0b7e8e8728
> > --- /dev/null
> > +++ b/sysdeps/x86/fpu/test-builtin-denormal-cxx.cc
> > @@ -0,0 +1 @@
> > +#include "test-builtin-denormal.c"
> > diff --git a/sysdeps/x86/fpu/test-builtin-denormal.c b/sysdeps/x86/fpu/test-builtin-denormal.c
> > index bddcd59387..789f1e3535 100644
> > --- a/sysdeps/x86/fpu/test-builtin-denormal.c
> > +++ b/sysdeps/x86/fpu/test-builtin-denormal.c
> > @@ -40,15 +40,23 @@ static const ieee_long_double_shape_type inputs[] = {
> > static int
> > do_test (void)
> > {
> > - for (int i = 0; i < array_length (inputs); i++)
> > + for (unsigned int i = 0; i < array_length (inputs); i++)
> > {
> > TEST_COMPARE (feclearexcept (FE_INVALID), 0);
> > TEST_COMPARE (fpclassify (inputs[i].value), FP_NAN);
> > +#ifndef __cplusplus
> > + /* C++ uses the builtin, which does not trigger FE_INVALID for
> > + pseudo-numbers (GCC BZ 123161 and LLVM issue 172533. */
> > TEST_COMPARE (fetestexcept (FE_INVALID), 0);
>
> Shouldn't the C++ builtin be disabled?
We should add a C++ test and make it XFAIL.
>
> > +#endif
> >
> > TEST_COMPARE (feclearexcept (FE_INVALID), 0);
> > TEST_COMPARE (isinf (inputs[i].value), 0);
> > +#ifndef __cplusplus
> > + /* C++ uses the builtin, which does not trigger FE_INVALID for
> > + pseudo-numbers (GCC BZ 123173 and LLVM issue 172651. */
> > TEST_COMPARE (fetestexcept (FE_INVALID), 0);
> > +#endif
> > }
> >
> > return 0;
> > --
> > 2.43.0
> >
>
>
> --
> H.J.
--
H.J.
More information about the Libc-alpha
mailing list