On Mon, 13 Oct 2025, Peter0x44 wrote:
On 2025-10-13 10:13, Martin Storsjö wrote:
This code was used for initial armv7 target bringup - but it hasn't
been used since it was made disabled by default in
ae0e8298da4e02606b389e988ae866994f0d1e3b.
The armv7 target (ARMNT) has got support for hardware float in
its baseline, so there's no need for this code. There are no known
cases in the wild where users build mingw-w64-crt with this enabled.
Signed-off-by: Martin Storsjö <[email protected]>
I'm not against this in principle, but could it maybe affect some users of
maybe, windows 9x on a 486sx or something like that?
Not saying that's a great reason to keep it. But it could perhaps be a use
case?
In theory, I guess the code could be used for that - but it currently
really isn't. This code isn't generic softfloat code, it's specifically
softfloat-for-arm32; the source files are added to src_libmingwexarm32 in
Makefile.am only.
And back when I started to work on mingw on arm32, when these functions
were still in use, I ran into a number of cases where they either gave bad
results, or even just caused crashes for some specific input values. So
the code isn't really mature for that case either.
So if someone wants to pick up the code, brush it off and use it for
486sx, it still exists in git history though.
// Martin
_______________________________________________
Mingw-w64-public mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
