[cfe-dev] [Release-testers] [llvm-dev] 10.0.1-rc4 tagged

[Brian Cain via cfe-dev]

> -----Original Message-----
> From: Tom Stellard <tstellar at redhat.com>
> Sent: Wednesday, July 15, 2020 1:17 PM
> To: bcain at codeaurora.org; 'Brian Cain' <brian.cain at gmail.com>
> Cc: 'Release-testers' <release-testers at lists.llvm.org>; 'Clang Dev' <cfe-
> dev at lists.llvm.org>; 'openmp-dev' <openmp-dev at lists.llvm.org>; 'LLDB Dev'
> <lldb-dev at lists.llvm.org>
> Subject: Re: [Release-testers] [llvm-dev] 10.0.1-rc4 tagged
> 
> On 7/15/20 1:07 PM, Tom Stellard via Release-testers wrote:
> > On 7/14/20 8:35 PM, Brian Cain wrote:
> >>> -----Original Message-----
> >>> From: llvm-dev <llvm-dev-bounces at lists.llvm.org> On Behalf Of Tom
> >>> Stellard via llvm-dev
> >> ...
> >>> On 7/9/20 7:42 PM, Brian Cain wrote:
> >>>> Uploaded:
> >>>> dcf43e25a77a2ffb4ffaa5a04babe409b2b3ffac07bc989a1dca730ecacb43c2
> >>>>    clang+llvm-10.0.1-rc4-x86_64-linux-sles12.4.tar.xz
> >>>>
> 6db54d9f55fb41877b23f4b89e7b68d44fcc5378d615d232c820307237dc33f7
> >>>>    clang+llvm-10.0.1-rc2-x86_64-linux-sles12.4.tar.xz
> >>>>
> 2d907945d8d8d5d2969c6947c50d91e100d5dd09ccb37214a811c11cbfa635af
> >>>>    clang+llvm-10.0.1-rc3-x86_64-linux-sles12.4.tar.xz
> >>>>
> >>>> ccdfda65932661e5fd4faba1fde17fe3800f39b4c21687242b7cdfdbea4cf131
> >>>>    clang+llvm-10.0.1-rc3-x86_64-linux-gnu-ubuntu-16.04.tar.xz
> >>>>
> 761a6f4658545f733f14dcd6c50a16b49994a4c448c779616c5716eed80b90f1
> >>>>    clang+llvm-10.0.1-rc4-x86_64-linux-gnu-ubuntu-16.04.tar.xz
> >>>>
> >>>> I saw phase2/3 mismatches that seem concerning.  Are these the only
> >>>> platforms that encountered them?
> >>>>
> >>>> I can open a bug, let me know if there's particular data beyond the
> >>>> logs I should gather.
> >>>>
> >>>
> >>> Can you attach 2 of the binary files the are different?
> >>>
> >>
> >> Attached a few from each impacted platform.  Additional notes below.
> >>
> >> # for ubuntu16, 10.0.1-rc4:
> >> $ grep 'differs between' rc4/logs/testing.10.0.1-rc4.log |wc -l
> >> 14
> >>
> >> $ grep 'differs between' rc4/logs/testing.10.0.1-rc4.log file
> >> omptarget-nvptx_intermediate_link.o differs between phase 2 and phase
> >> 3 file omptarget-nvptx_generated_target_impl.cu.o differs between
> >> phase
> >> 2 and phase 3
> >> file omptarget-nvptx_generated_reduction.cu.o differs between phase 2
> >> and phase 3 file omptarget-nvptx_generated_omptarget.cu.o differs
> >> between phase 2 and phase 3 file
> >> omptarget-nvptx_generated_support.cu.o differs between phase 2 and
> >> phase 3 file omptarget-nvptx_generated_libcall.cu.o differs between
> >> phase 2 and phase 3 file omptarget-nvptx_generated_critical.cu.o
> >> differs between phase 2 and phase 3 file
> >> omptarget-nvptx_generated_data_sharing.cu.o differs between phase
> >> 2 and phase 3
> >> file omptarget-nvptx_generated_task.cu.o differs between phase 2 and
> >> phase 3 file omptarget-nvptx_generated_cancel.cu.o differs between
> >> phase 2 and phase 3 file omptarget-nvptx_generated_sync.cu.o differs
> >> between phase 2 and phase 3 file omptarget-nvptx_generated_loop.cu.o
> >> differs between phase 2 and phase 3 file
> >> omptarget-nvptx_generated_omp_data.cu.o differs between phase 2 and
> >> phase 3 file omptarget-nvptx_generated_parallel.cu.o differs between
> >> phase 2 and phase 3
> >>
> >> ~~~~~~~~~~
> >
> > The difference in ubuntu 10.0.1-rc4 are in the '__nv_module_id'
> > section of binaries, which contains a GUID that is unique to the
> > binary, so this is expected.
> >
> >>
> >> # for sles12, 10.0.1-rc3:
> >> $ grep 'differs between' rc3/logs/testing.10.0.1-rc3.log |wc -l
> >> 4860
> >>
> >> $ tail rc3/logs/testing.10.0.1-rc3.log file interface.cpp.o differs
> >> between phase 2 and phase 3 file rtl.cpp.o differs between phase 2
> >> and phase 3 file omptarget.cpp.o differs between phase 2 and phase 3
> >> file ompt-tsan.cpp.o differs between phase 2 and phase 3 file
> >> ompt-tsan.cpp.o differs between phase 2 and phase 3 file Bye.cpp.o
> >> differs between phase 2 and phase 3 file TestPlugin.cpp.o differs
> >> between phase 2 and phase 3 file PipSqueak.cpp.o differs between
> >> phase 2 and phase 3 file PipSqueak.cpp.o differs between phase 2 and
> >> phase 3 file ExportedFuncs.cpp.o differs between phase 2 and phase 3
> >>
> >
> > Do you get these same failures on sles with rc4?

No, I did not.

> It looks like the difference in the objects on sles is because the compiler ID
> string in phase 2 and phase 3 have a different git commit hash. The git hash in
> the phase 3 binaries is acd921f02c39fa5cf7aa055b6062bfe9025e3781, which I
> don't see in the llvm git tree.

Ugh -- yes, I realize now -- this was my fault.  Apologies, I deviated from the process.  I had a local change to test-release.sh that I committed that caused this problem.  Sorry for the distraction!

-Brian