[llvm-testresults] buildbot failure in llvm on llvm-gcc-i386-linux-selfhost

[Jan Sjodin]

Since my code does change the way the debug info is emitted it is likely to be 
my change, but i would have expected it to fail in stage 1 in that case. It is 
strange that slightly different debug info should cause a miscompile.

- Jan




________________________________
From: Duncan Sands <baldrick at free.fr>
To: NAKAMURA Takumi <geek4civic at gmail.com>
Cc: llvm-testresults <llvm-testresults at cs.uiuc.edu>; Daniel Dunbar 
<daniel at zuster.org>; Jan Sjodin <jan_sjodin at yahoo.com>; Peter Collingbourne 
<peter at pcc.me.uk>
Sent: Sat, March 12, 2011 7:43:03 AM
Subject: Re: [llvm-testresults] buildbot failure in llvm on 
llvm-gcc-i386-linux-selfhost

On 12/03/11 12:11, NAKAMURA Takumi wrote:
>>  
>>http://google1.osuosl.org:8011/builders/llvm-gcc-i386-linux-selfhost/builds/1540
> 
> It seems it has been failing on since then. Why?
> 
> ...Takumi(fine)

The llvm-x86_64-linux-checks builder is failing in the same way: these tests
are failing in stage2:
  2006-11-30-pubnames.cpp
  2010-08-31-byvalarg.cpp
They do not fail in stage 1, suggesting a miscompile.  The first failures were
here:
http://google1.osuosl.org:8011/builders/llvm-gcc-i386-linux-selfhost/builds/1540
http://google1.osuosl.org:8011/builders/llvm-x86_64-linux-checks/builds/557

The set of commits for the llvm-x86_64-linux-checks builder is a strict superset
of those implicated by the llvm-gcc-i386-linux-selfhost builder.  Eliminating
those involving clang, this leaves two commits.  I think the first one, by
ddunbar, can be eliminated: it was reverting a recent change, so is unlikely
to have introduced a miscompilation, and the original patch has since been
reapplied without fixing the buildbots.  So that leaves jsjodin's change which
I am about to try reverting.

Ciao, Duncan.

Changed by: ddunbar
Changed at: Fri 11 Mar 2011 11:35:18
Revision: 127477
Changed files:

lib/Transforms/Scalar/CodeGenPrepare.cpp
test/CodeGen/ARM/constants.ll
test/CodeGen/CellSPU/jumptable.ll
test/CodeGen/Thumb/dyn-stackalloc.ll
test/CodeGen/Thumb2/2009-08-06-SpDecBug.ll
test/CodeGen/Thumb2/2009-09-28-ITBlockBug.ll
test/CodeGen/Thumb2/2010-11-22-EpilogueBug.ll
test/CodeGen/Thumb2/cross-rc-coalescing-2.ll
test/CodeGen/X86/2008-04-16-ReMatBug.ll
test/CodeGen/X86/MachineSink-PHIUse.ll
test/CodeGen/X86/personality.ll
test/CodeGen/X86/pr3366.ll
test/CodeGen/X86/sext-i1.ll
test/CodeGen/X86/zext-extract_subreg.ll
test/MC/ARM/simple-encoding.ll
test/Transforms/CodeGenPrepare/basic.ll
Comments:
Revert r127459, "Optimize trivial branches in CodeGenPrepare, which often get
created from the", it broke some GCC test suite tests.
Properties:
Changed by: jsjodin
Changed at: Fri 11 Mar 2011 11:41:18
Revision: 127478
Changed files:

include/llvm/MC/MCSection.h
include/llvm/MC/MCSectionELF.h
lib/CodeGen/AsmPrinter/AsmPrinterDwarf.cpp
Comments:
Remove optimization emitting a reference insted of label difference, since it 
can create more relocations. Removed isBaseAddressKnownZero method, because it 
is no longer used.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-testresults/attachments/20110312/183084f8/attachment.html>