On Wed, 26 Jun 2024, Pali Rohár wrote:
On Wednesday 26 June 2024 15:54:42 Martin Storsjö wrote:
On Tue, 25 Jun 2024, Pali Rohár wrote:
---
mingw-w64-crt/lib-common/ucrtbase.def.in | 194 +++++++++++------------
1 file changed, 95 insertions(+), 99 deletions(-)
nextafter
nextafterf
-nextafterl F_X86_ANY(DATA)
-; All of the nexttoward functions take the second parameter as long doubke,
-; making them unusable for x86.
-nexttoward F_X86_ANY(DATA)
-nexttowardf F_X86_ANY(DATA)
-nexttowardl F_X86_ANY(DATA)
+F_ARM_ANY(nextafterl) ; Can't use long double functions from the CRT on x86
+F_ARM_ANY(nexttoward) ; Can't use long double functions from the CRT on x86
+F_ARM_ANY(nexttowardf) ; Can't use long double functions from the CRT on x86
+F_ARM_ANY(nexttowardl) ; Can't use long double functions from the CRT on x86
I think the existing comment here, about why a function that isn't the
regular -l suffixed function is treated the same way, is useful to keep.
I thought that it is not needed to have that comment as basically newly
added comments explains that. Basically every commented function either
takes long double parameter or function returns long double type.
Yes, that's the general principle, but when looking at lists of functions,
like nextafter, nextafterf, nextafterl, then nexttoward is surprising -
it's not immediately obvious that these functions have a second parameter
that always is long double, unless you look up the docs for the functions.
// Martin
_______________________________________________
Mingw-w64-public mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
