On Mon, Mar 07, 2022 at 02:15:49AM -0600, Glenn Washburn wrote:
> On Fri, 4 Mar 2022 18:01:10 -0600
> Glenn Washburn <developm...@efficientek.com> wrote:
>
> > On Wed,  2 Mar 2022 14:08:23 -0500
> > Robbie Harwood <rharw...@redhat.com> wrote:
> >
> > > Changes this version:
> > >
> > > - Reorder last two commits so that warning fixes come after the change 
> > > that
> > >   introduces them.
> > > - Fix comment formatting to comply with grub2 style.
> >
> > Either I missed it before or something changed. But I'm getting this
> > build error now for x86_64-efi, and I'm not getting it without this patch
> > series.
> >
> > In file included from 
> > /root/grub-tests.update-gnulib/grub/include/grub/disk.h:31
> > ,
> >                  from 
> > /root/grub-tests.update-gnulib/grub/include/grub/file.h:26
> > ,
> >                  from 
> > /root/grub-tests.update-gnulib/grub/include/grub/loader.h:
> > 23,
> >                  from 
> > /root/grub-tests.update-gnulib/grub/grub-core/loader/i386/
> > bsd.c:19:
> > /root/grub-tests.update-gnulib/grub/grub-core/loader/i386/bsd.c: In 
> > function ‘gr
> > ub_freebsd_add_meta_module’:
> > /root/grub-tests.update-gnulib/grub/include/grub/misc.h:71:10: error: ‘ptr’ 
> > may be used uninitialized in this function [-Werror=maybe-uninitialized]
> >    71 |   return grub_memmove (dest, src, n);
> >       |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~
> > /root/grub-tests.update-gnulib/grub/grub-core/loader/i386/bsd.c:266:9: 
> > note: ‘ptr’ was declared here
> >   266 |   void *ptr;
> >       |         ^~~
> >
> > Reviewing the code it doesn't look like ptr can actually be used
> > uninitialized, so it seems like GCC 10.1.0 isn't smart enough to figure
> > that out. Initializing to NULL fixes the build issue.
>
> For completeness, I just realized that I wasn't using the compiler I
> thought I was using. Turns out this issue is with x86_64 Debian 11's
> GCC, which is 10.2.1. I don't think this is relevant to the issue, but
> may help in reproducing if needed.
>
> After some more testing, I've narrowed it down to passing -O2 to gcc,
> without which the series builds fine.

Thanks for the report. I am just running build tests for all architectures
and platforms with GCC 10/11. I have spotted and fixed 2 additional problems.
I hope I will post fixes for all of them tomorrow.

Robbie and I discovered another gnulib build issue for Windows platforms.
I prepared a patch and will share it with Robbie to send it in updated
gnulib v9 patchset.

Daniel

_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Reply via email to