On Mon, 21 Oct 2024, Pali Rohár wrote:
See commit e5d1fe451b81 ("ucrt: Don't use the float env functions from the UCRTt library") for more details. --- .../api-ms-win-crt-runtime-l1-1-0.def.in | 18 +++++++------- mingw-w64-crt/lib-common/msvcr120_app.def.in | 24 ++++++++++--------- .../lib-common/ucrtbase-common.def.in | 18 +++++++------- mingw-w64-crt/lib32/msvcr120.def.in | 24 ++++++++++--------- mingw-w64-crt/lib32/msvcr120d.def.in | 24 ++++++++++--------- mingw-w64-crt/lib64/msvcr120.def.in | 24 ++++++++++--------- mingw-w64-crt/lib64/msvcr120d.def.in | 24 ++++++++++--------- mingw-w64-crt/libarm32/msvcr120.def.in | 24 ++++++++++--------- mingw-w64-crt/libarm32/msvcr120d.def.in | 24 ++++++++++--------- 9 files changed, 109 insertions(+), 95 deletions(-)diff --git a/mingw-w64-crt/lib-common/msvcr120_app.def.in b/mingw-w64-crt/lib-common/msvcr120_app.def.in index 4508a43f3370..d91709330583 100644 --- a/mingw-w64-crt/lib-common/msvcr120_app.def.in +++ b/mingw-w64-crt/lib-common/msvcr120_app.def.in @@ -1985,19 +1985,21 @@ fclose fdim fdimf F_ARM32(fdiml) ; Can't use long double functions from the CRT on x86 -feclearexcept -fegetenv -fegetexceptflag -fegetround -feholdexcept +; Don't use the float env functions from msvcr120; fesetround doesn't seem to have +; any effect on the FPU control word as required by other libmingwex math
Did you verify this aspect with all these libraries (testing setting rounding and calling our math functions?), or is this just repeating the same statement that was tested with UCRT?
Technically I think the patch is fine, but I'm a little weary about adding statements saying something about how things behave with a library unless we actually did test that.
// Martin _______________________________________________ Mingw-w64-public mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
