Hi,
On Thu, 29 Oct 2020 01:24:07 -0400
George Koehler wrote:
> On Thu, 8 Oct 2020 01:08:05 +0200
> Charlene Wendling <julianaito@posteo.jp> wrote:
>
> > Hi,
> >
> > Here are the current gnustep/libobjc failures:
> >
> > http://build-failures.rhaalovely.net/powerpc/2020-08-23/www/sope.log
> > http://build-failures.rhaalovely.net/powerpc/2020-08-23/x11/gnustep/dbuskit.log
> > http://build-failures.rhaalovely.net/powerpc/2020-08-23/x11/gnustep/gui.log
> > http://build-failures.rhaalovely.net/powerpc/2020-08-23/x11/gnustep/performance.log
> >
> > So i'm asking again if i can go ahead so it doesn't waste bulk time.
> >
> > Charlène.
>
> I'm sorry for replying 3 weeks after your mail. Your diff (from
> April!), quoted below, would have marked gnustep-base BROKEN-powerpc
> and changed gnustep-make on powerpc.
>
> I now have a clang diff (https://reviews.llvm.org/D90329) that
> unbreaks gnustep-base and gnustep-make without needing your diff.
> Last week, Anthony Richardby had found and reported the bug that
> broke clang's va_arg(_, id) in Objective-C code. With the clang diff,
> and using FETCH_PACKAGES=, I built x11/gnustep/{libobjc2,make,base,
> gui,gmines} and played gmines.
Thanks a lot to you two! I've let it flow since i'm not much qualified
to assert if it was right or not.
> Later, I will mail the clang diff to tech@, but now, it isn't in
> OpenBSD's clang, so ok gkoehler@ to mark base BROKEN-powerpc if you
> don't want to wait for the fix.
With that new patch you're preparing, marking it BROKEN would be extra
work for a not so long time. I think it's fine to let it like this.
Especially that now spidermonkey requires rust, it will shorten a bit
the bulk time anyway.
> I didn't ok your gnustep/make diff because I'm not sure. I don't know
> why amd64 and i386 have the different LIBRARY_COMBO. Also, AS=llvm-as
> looks wrong because ports-clang (for llvm-as) might not be installed.
Honestly, i'm not sure about that as well. jca expressed the same
concerns. I added that because the two base-clang archs that built
gnustep at that time were concerned by that part. If it works without
this change and your clang diff on powerpc, we should left that part
as-is in my opinion.
> The clang diff affects only 32-bit powerpc. Something else is wrong
> on powerpc64 (where most gnustep packages are missing), but I have
> not yet tried a powerpc64 build. --George
Charlène.
No comments:
Post a Comment