https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122172

--- Comment #29 from Hans-Peter Nilsson <hp at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #28)
> (In reply to Hans-Peter Nilsson from comment #27)
> > There were still (22-11=)11 regressions compared to r16-3809 for cris-elf at
> > r16-4373-g385984f555 (still at r16-4386-g5b57da59c12e69).  I'll open a
> > separate PR for:
> >  g++.sum g++-dg-lto-devirt-2-01.exe
> >  g++.sum g++.dg/ipa/devirt-2.C
> > but these 9 seem attributable to the same r16-3810 commit:
> >  libstdc++.sum 24_iterators/reverse_iterator/100639.cc
> >  libstdc++.sum 27_io/print/1.cc
> >  libstdc++.sum 27_io/print/2.cc
> >  libstdc++.sum 27_io/print/3.cc
> >  libstdc++.sum 29_atomics/atomic_ref/requirements.cc
> >  libstdc++.sum std/ranges/iota/93267.cc
> >  libstdc++.sum std/ranges/iota/96042.cc
> >  libstdc++.sum std/ranges/iota/size.cc
> >  libstdc++.sum std/ranges/subrange/96042.cc
> > All but one are compilation or linking errors for a missing funlockfile
> > function (some of the errors are at compilation stage, others for linking
> > state).
> 
> Those are all new in r16-4350-g8bd872f1ea7414 so unrelated to the
> _Atomic_word stuff.

Responding to that statement at face value (sp?): sorry, but that's not
correct.  I re-checked my double-checking.

But no need to take my word for it; the first one,
24_iterators/reverse_iterator/100639.cc for example, you can see that it has a
commit-date in 2020 (no, not author-date; I've learned). The last one,
std/ranges/subrange/96042.cc, from 2000.

But, I realized all by myself that you could mean that r16-4350 changed the
reason for the failure.  Looking at the log entries for the first and the last
failure, that seems to be it: r16-4350 needs to handle newlib targets.

> > The atomic one is
> > /gccobj/cris-elf/libstdc++-v3/include/bits/atomic_base.h:1555: \
> > error: static assertion failed: atomic operations on volatile T must be
> > lock-free
> 
> That's Bug 122267

Thanks, that does look similar.  A different test though.

> I think they're two new regressions from two separate commits, unrelated to
> this one.

I guess I'll go bother the author of those commits instead. :)
(I'll reference this PR though.)

Reply via email to