STL
STL copied to clipboard
microsoft
Clang 18+ `/fp:fast` emits `-Wnan-infinity-disabled` in `VSO_0000000_vector_algorithms_floats`
All 8 clang-cl /fp:fast configs have the same issue. For example:
Error log for clang-cl /fp:fast /MT without /D_USE_STD_VECTOR_ALGORITHMS=0
FAIL: std :: tests/VSO_0000000_vector_algorithms_floats:16 (6 of 24)
******************** TEST 'std :: tests/VSO_0000000_vector_algorithms_floats:16' FAILED ********************
Build setup steps:
Build steps:
Build step failed unexpectedly.
Command: "c:\Program Files\LLVM19\bin\clang-cl.EXE" "C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp" "-IC:\STL\out\x64\out\inc" "-IC:\STL\llvm-project\libcxx\test\support" "-IC:\STL\tests\std\include" "/nologo" "/Od" "/W4" "/w14061" "/w14242" "/w14265" "/w14582" "/w14583" "/w14587" "/w14588" "/w14749" "/w14841" "/w14842" "/w15038" "/w15214" "/w15215" "/w15216" "/w15217" "/w15262" "/sdl" "/WX" "/D_ENABLE_STL_INTERNAL_CHECK" "/bigobj" "/FIforce_include.hpp" "/w14365" "/w14668" "/w15267" "/D_ENFORCE_FACET_SPECIALIZATIONS=1" "/D_STL_CALL_ABORT_INSTEAD_OF_INVALID_PARAMETER" "-fno-ms-compatibility" "-fno-delayed-template-parsing" "-Wno-unqualified-std-cast-call" "/EHsc" "/MT" "/std:c++20" "/permissive-" "/w14640" "/Zc:threadSafeInit-" "--start-no-unused-arguments" "/fp:fast" "-m64" "-FeC:\STL\out\x64\tests\std\tests\VSO_0000000_vector_algorithms_floats\Output\16\VSO_0000000_vector_algorithms_floats.exe" "-link" "-LIBPATH:C:\STL\out\x64\out\lib\amd64" "-LIBPATH:C:\Program Files\Microsoft Visual Studio\2022\Preview\VC\Tools\MSVC\14.42.34226\lib\x64" "/MANIFEST:EMBED"
Exit Code: 1 (0x1)
Standard Error:
--
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:5:
C:\STL\out\x64\out\inc\limits(883,16): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
883 | return __builtin_nanf("0");
| ^~~~~~~~~~~~~~~~~~~
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(671,1): error: use of infinity is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
671 | _CLANG_BUILTIN1(isfinite)
| ^~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(662,5): note: expanded from macro '_CLANG_BUILTIN1'
662 | _CLANG_BUILTIN1_ARG(NAME, float) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(605,16): note: expanded from macro '_CLANG_BUILTIN1_ARG'
605 | return __builtin_##NAME(_Xx); \
| ^~~~~~~~~~~~~~~~~~~~~
(196,1): note: expanded from here
196 | __builtin_isfinite
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(671,1): error: use of infinity is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
671 | _CLANG_BUILTIN1(isfinite)
| ^~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(663,5): note: expanded from macro '_CLANG_BUILTIN1'
663 | _CLANG_BUILTIN1_ARG(NAME, double) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(605,16): note: expanded from macro '_CLANG_BUILTIN1_ARG'
605 | return __builtin_##NAME(_Xx); \
| ^~~~~~~~~~~~~~~~~~~~~
(199,1): note: expanded from here
199 | __builtin_isfinite
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(671,1): error: use of infinity is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
671 | _CLANG_BUILTIN1(isfinite)
| ^~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(664,5): note: expanded from macro '_CLANG_BUILTIN1'
664 | _CLANG_BUILTIN1_ARG(NAME, long double)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(605,16): note: expanded from macro '_CLANG_BUILTIN1_ARG'
605 | return __builtin_##NAME(_Xx); \
| ^~~~~~~~~~~~~~~~~~~~~
(202,1): note: expanded from here
202 | __builtin_isfinite
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(672,1): error: use of infinity is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
672 | _CLANG_BUILTIN1(isinf)
| ^~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(662,5): note: expanded from macro '_CLANG_BUILTIN1'
662 | _CLANG_BUILTIN1_ARG(NAME, float) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(605,16): note: expanded from macro '_CLANG_BUILTIN1_ARG'
605 | return __builtin_##NAME(_Xx); \
| ^~~~~~~~~~~~~~~~~~~~~
(205,1): note: expanded from here
205 | __builtin_isinf
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(672,1): error: use of infinity is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
672 | _CLANG_BUILTIN1(isinf)
| ^~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(663,5): note: expanded from macro '_CLANG_BUILTIN1'
663 | _CLANG_BUILTIN1_ARG(NAME, double) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(605,16): note: expanded from macro '_CLANG_BUILTIN1_ARG'
605 | return __builtin_##NAME(_Xx); \
| ^~~~~~~~~~~~~~~~~~~~~
(208,1): note: expanded from here
208 | __builtin_isinf
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(672,1): error: use of infinity is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
672 | _CLANG_BUILTIN1(isinf)
| ^~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(664,5): note: expanded from macro '_CLANG_BUILTIN1'
664 | _CLANG_BUILTIN1_ARG(NAME, long double)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(605,16): note: expanded from macro '_CLANG_BUILTIN1_ARG'
605 | return __builtin_##NAME(_Xx); \
| ^~~~~~~~~~~~~~~~~~~~~
(211,1): note: expanded from here
211 | __builtin_isinf
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(673,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
673 | _CLANG_BUILTIN1(isnan)
| ^~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(662,5): note: expanded from macro '_CLANG_BUILTIN1'
662 | _CLANG_BUILTIN1_ARG(NAME, float) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(605,16): note: expanded from macro '_CLANG_BUILTIN1_ARG'
605 | return __builtin_##NAME(_Xx); \
| ^~~~~~~~~~~~~~~~~~~~~
(214,1): note: expanded from here
214 | __builtin_isnan
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(673,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
673 | _CLANG_BUILTIN1(isnan)
| ^~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(663,5): note: expanded from macro '_CLANG_BUILTIN1'
663 | _CLANG_BUILTIN1_ARG(NAME, double) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(605,16): note: expanded from macro '_CLANG_BUILTIN1_ARG'
605 | return __builtin_##NAME(_Xx); \
| ^~~~~~~~~~~~~~~~~~~~~
(217,1): note: expanded from here
217 | __builtin_isnan
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(673,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
673 | _CLANG_BUILTIN1(isnan)
| ^~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(664,5): note: expanded from macro '_CLANG_BUILTIN1'
664 | _CLANG_BUILTIN1_ARG(NAME, long double)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(605,16): note: expanded from macro '_CLANG_BUILTIN1_ARG'
605 | return __builtin_##NAME(_Xx); \
| ^~~~~~~~~~~~~~~~~~~~~
(220,1): note: expanded from here
220 | __builtin_isnan
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(681,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
681 | _CLANG_BUILTIN2(isunordered)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(667,5): note: expanded from macro '_CLANG_BUILTIN2'
667 | _CLANG_BUILTIN2_ARG(NAME, float) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(611,16): note: expanded from macro '_CLANG_BUILTIN2_ARG'
611 | return __builtin_##NAME(_Xx, _Yx); \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
(293,1): note: expanded from here
293 | __builtin_isunordered
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(681,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
681 | _CLANG_BUILTIN2(isunordered)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(668,5): note: expanded from macro '_CLANG_BUILTIN2'
668 | _CLANG_BUILTIN2_ARG(NAME, double) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(611,16): note: expanded from macro '_CLANG_BUILTIN2_ARG'
611 | return __builtin_##NAME(_Xx, _Yx); \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
(297,1): note: expanded from here
297 | __builtin_isunordered
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(681,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
681 | _CLANG_BUILTIN2(isunordered)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(669,5): note: expanded from macro '_CLANG_BUILTIN2'
669 | _CLANG_BUILTIN2_ARG(NAME, long double)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(611,16): note: expanded from macro '_CLANG_BUILTIN2_ARG'
611 | return __builtin_##NAME(_Xx, _Yx); \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
(301,1): note: expanded from here
301 | __builtin_isunordered
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(688,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
688 | _CLANG_BUILTIN2_TEMPLATED(isunordered)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(641,5): note: expanded from macro '_CLANG_BUILTIN2_TEMPLATED'
641 | _CLANG_BUILTIN2_ARG_TEMPLATED(NAME, float) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(617,16): note: expanded from macro '_CLANG_BUILTIN2_ARG_TEMPLATED'
617 | return __builtin_##NAME(static_cast(_Xx), static_cast(_Yx)); \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(81,1): note: expanded from here
81 | __builtin_isunordered
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(688,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
688 | _CLANG_BUILTIN2_TEMPLATED(isunordered)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(642,5): note: expanded from macro '_CLANG_BUILTIN2_TEMPLATED'
642 | _CLANG_BUILTIN2_ARG_TEMPLATED(NAME, double) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(617,16): note: expanded from macro '_CLANG_BUILTIN2_ARG_TEMPLATED'
617 | return __builtin_##NAME(static_cast(_Xx), static_cast(_Yx)); \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(85,1): note: expanded from here
85 | __builtin_isunordered
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(688,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
688 | _CLANG_BUILTIN2_TEMPLATED(isunordered)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(643,5): note: expanded from macro '_CLANG_BUILTIN2_TEMPLATED'
643 | _CLANG_BUILTIN2_ARG_TEMPLATED(NAME, long double) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(617,16): note: expanded from macro '_CLANG_BUILTIN2_ARG_TEMPLATED'
617 | return __builtin_##NAME(static_cast(_Xx), static_cast(_Yx)); \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(89,1): note: expanded from here
89 | __builtin_isunordered
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(688,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
688 | _CLANG_BUILTIN2_TEMPLATED(isunordered)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(644,5): note: expanded from macro '_CLANG_BUILTIN2_TEMPLATED'
644 | _CLANG_BUILTIN2_ARG_TEMPLATED(NAME, signed char) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(617,16): note: expanded from macro '_CLANG_BUILTIN2_ARG_TEMPLATED'
617 | return __builtin_##NAME(static_cast(_Xx), static_cast(_Yx)); \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(93,1): note: expanded from here
93 | __builtin_isunordered
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(688,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
688 | _CLANG_BUILTIN2_TEMPLATED(isunordered)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(645,5): note: expanded from macro '_CLANG_BUILTIN2_TEMPLATED'
645 | _CLANG_BUILTIN2_ARG_TEMPLATED(NAME, short) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(617,16): note: expanded from macro '_CLANG_BUILTIN2_ARG_TEMPLATED'
617 | return __builtin_##NAME(static_cast(_Xx), static_cast(_Yx)); \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(97,1): note: expanded from here
97 | __builtin_isunordered
| ^
In file included from C:\STL\tests\std\tests\VSO_0000000_vector_algorithms_floats\test.cpp:6:
In file included from C:\STL\out\x64\out\inc\random:13:
C:\STL\out\x64\out\inc\cmath(688,1): error: use of NaN is undefined behavior due to the currently enabled floating-point options [-Werror,-Wnan-infinity-disabled]
688 | _CLANG_BUILTIN2_TEMPLATED(isunordered)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(646,5): note: expanded from macro '_CLANG_BUILTIN2_TEMPLATED'
646 | _CLANG_BUILTIN2_ARG_TEMPLATED(NAME, int) \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C:\STL\out\x64\out\inc\cmath(617,16): note: expanded from macro '_CLANG_BUILTIN2_ARG_TEMPLATED'
617 | return __builtin_##NAME(static_cast(_Xx), static_cast(_Yx)); \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(101,1): note: expanded from here
101 | __builtin_isunordered
| ^
fatal error: too many errors emitted, stopping now [-ferror-limit=]
20 errors generated.
--
#4932 will silence these warnings for the time being.
I don't know what you would do about it other than suppression.
/fp:fast is indeed asking for UB for NANs.
I don't know what you would do about it other than suppression.
/fp:fastis indeed asking for UB for NANs.
Ideally, we don't test undefined behavior. It tends to be unreliable. That could mean adding a macro to the /fp:fast configs that tells the tests to avoid problematic code. Or it could mean suppression, or it could mean not running those tests with /fp:fast.
I see.
The whole point of separating VSO_0000000_vector_algorithms_floats from VSO_0000000_vector_algorithms was testing the cursed floating modes with the same exhaustive coverage as the standard /fp:precise mode.
Out of the cursed modes, /fp:fast is the main goal for testing, as it changes the control flow for the algorithm in question.
We can possibly drop infinities from /fp:fast coverage, that is, remove these elements:
https://github.com/microsoft/STL/blob/c7c5ca7d6adde7c1554b1e52c0b97cb2a7f7ca60/tests/std/tests/VSO_0000000_vector_algorithms_floats/test.cpp#L33-L34
As for NANs. the test doesn't even try them in the standard /fp:precise mode. We've concluded that we don't support NANs in the sorting algorithms.
But most of these warnings (actually all quoted in the description) originate from <cmath> and <limits> headers.
Need to suppress them in these headers!
After one minute of my not-especially-clueful attempting during the weekly maintainer meeting, we weren't able to repro warnings from <limits> with a from-scratch source file. We should extract a minimal repro to understand exactly what about this test is continuing to emit Clang warnings. As usual, if merely including the header and doing not-risky things (like mentioning numeric_limits<double>::stuff() but not specifically inf/nan directly) emits a warning, then we should suppress it in the STL headers, but not if the user is asking us to do risky things on their behalf.
When NaN and Inf are not properly supported, should we make the corresponding function return 0.0 and has_* give false?
When NaN and Inf are not properly supported, should we make the corresponding function return
0.0andhas_*givefalse?
I don't think so. Something still works with NaN and Inf as expected even in /fp:fast.
As usual, if merely including the header and doing not-risky things (like mentioning
numeric_limits<double>::stuff()but not specifically inf/nan directly) emits a warning, then we should suppress it in the STL headers, but not if the user is asking us to do risky things on their behalf.
I've verified that it's the case (although for libc++ with #pragma GCC system_header removed; Godbolt link).