On 01/04/21 16:27 +0200, Jakub Jelinek via Libstdc++ wrote:
On Thu, Apr 01, 2021 at 03:25:56PM +0100, Jonathan Wakely via Gcc-patches wrote:
On 01/04/21 16:04 +0200, Eric Botcazou wrote:
> > Oops, and I also forgot to update the other headers that were using
> > those autoconf-generated macros!
> >
> > Fixed by this patch. Tested on mingw-w64, pushed to trunk.
>
> The original patch was apparently backported onto the gcc-10 branch, but not
> this follow-up fix, so Windows builds are now broken on the branch.
Sorry. I did try to test a mingw-w64 cross, but I think I probably
tested it against trunk not the branch. So didn't test it at all.
It should be fixed by the attached patch for gcc-10. I'm testing it
now on linux and mingw.
I'll wait for RM approval before pushing.
Ok for 10.3, thanks.
Thanks, pushed.
I was unable to reproduce the problem on the gcc-10 branch or with the
gcc-10.3.0-RC-20210401.tar.xz release candidate. That's because I
didn't backport r11-6226-gf5feee6adc8350a292c235eb21e31a5082350d94
either. The follow-up fix referred to above was only needed because
r11-6226 regenerated libstdc++-v3/configure and libstdc++-v3/config.h.in. But
because I didn't backport that (or the follow-up fix) to gcc-10, it still
checks for the errno macros during configure, and so the HAVE_EXXX macros that
the Windows headers need should be defined.
Eric, are you building the RC with --enable-maintainer-mode maybe? Or
regenerating the autoconf files yourself?
