> With the change to require lld, it's likely the test runs on a smaller set of buildbots though.
Yes and maybe they would also be surprised to know that but ok, good news is we only have one bot of ours to fix. > I am guessing behavior is intended as LLVM_ENABLE_LLD=ON is not setting CLANG_D? Presumably EFAULT_LINKER for stage 2. Should we expect clang to use just built lld in stage 2 by default without even setting CLANG_DEFAULT_LINKER? Clang is usually a drop in replacement for gcc, so I'd expect using lld to always be opt in somehow during build or compile time. I just thought that the buildbot builders would opt into using as much llvm stuff as possible. Perhaps you can try enabling lld in the second stage and see what tests fail? There could be more than just this new one. On Mon, 1 Jul 2024 at 18:21, mingmingl <minglot...@gmail.com> wrote: > Thank you both for the quick responses! > > On Mon, Jul 1, 2024 at 5:03 AM Omair Javaid <omair.jav...@linaro.org> > wrote: > >> On Mon, 1 Jul 2024 at 14:56, David Spickett <david.spick...@linaro.org> >> wrote: >> >>> This is one for @Omair Javaid <omair.jav...@linaro.org> who is on bots >>> this week. >>> >>> But to save him some confusion, I noticed that the next build >>> https://lab.llvm.org/buildbot/#/builders/122/builds/151 passes. This is >>> because >>> https://github.com/llvm/llvm-project/commit/038bc1c18c786e14cc306401b00144265f8860f5 >>> required lld which surprisingly, makes the test unsupported? >>> >>> UNSUPPORTED: Profile-armhf :: Linux/instrprof-vtable-value-prof.cpp >>> (74341 of 77518) >>> >>> Which is odd because it's enabled in both stages and used to link in >>> stage 2. >>> >> I am guessing behavior is intended as LLVM_ENABLE_LLD=ON is not setting >> CLANG_DEFAULT_LINKER for stage 2. Should we expect clang to use just built >> lld in stage 2 by default without even setting CLANG_DEFAULT_LINKER? >> > >>> Did you get any other notifications of the same include issue, >>> particularly from Linaro's other builders? >>> >> > Not really. I only received two buildbot failure email notifications, one > is 'fatal error: 'cstdio' file not found' as being discussed here, and the > other <https://lab.llvm.org/buildbot/#/builders/95/builds/672> failure is > from buildbot clang-ppc64le-linux-test-suite and fixed by commit 038bc1c > <https://github.com/llvm/llvm-project/commit/038bc1c18c786e14cc306401b00144265f8860f5>. > With the change to require lld, it's likely the test runs on a smaller set > of buildbots though. > > On Mon, 1 Jul 2024 at 10:39, mingmingl <minglot...@gmail.com> wrote: >>> >>>> Greetings! >>>> This is Mingming and I'm an LLVM contributor. I have received a lot >>>> of >>>> useful code review feedback from aarch64 experts. Thank you for all of >>>> that! >>>> >>>> I'm writing to report a failure ( >>>> https://lab.llvm.org/buildbot/#/builders/122/builds/150) on buildbot >>>> clang-armv8-lld-2stage triggered by my recent patch. The issue seems to >>>> stem from the C++ standard libraries not being configured in the >>>> buildbot >>>> environment. >>>> >>>> While using a c-style header is a workaround (which pr 97245 >>>> <https://github.com/llvm/llvm-project/pull/97245> did), @petrhosek >>>> reminded >>>> me that it would be good to inform maintainers so they can take a look, >>>> which makes sense to me. >>>> >>>> Could someone please assist in resolving this configuration problem? >>>> Thank you in advance and let me know if I miss anything. >>>> >>>> Thanks, Mingming >>>> _______________________________________________ >>>> linaro-toolchain mailing list -- linaro-toolchain@lists.linaro.org >>>> To unsubscribe send an email to linaro-toolchain-le...@lists.linaro.org >>>> >>> _______________________________________________ linaro-toolchain mailing list -- linaro-toolchain@lists.linaro.org To unsubscribe send an email to linaro-toolchain-le...@lists.linaro.org
