[llvm] r279573 - MachineFunction: Introduce NoPHIs property

Wed Aug 24 19:37:22 PDT 2016

I afraid I may need some help now...
http://lab.llvm.org:8011/builders/clang-ppc64be-linux-lnt/builds/7075/steps/ninja%20check%201/logs/FAIL%3A%20AddressSanitizer-Unit%3A%3AAddressSanitizerInterface.ManyThreadsWithStatsStressTest

[ RUN      ] AddressSanitizerInterface.ManyThreadsWithStatsStressTest
==18839==AddressSanitizer CHECK failed:
/home/buildbots/ppc64be-clang-lnt-test/clang-ppc64be-lnt/llvm/projects/compiler-rt/lib/asan/asan_allocator.cc:484
"((m->free_tid)) == ((kInvalidTid))" (0x65, 0xffffff)
    #0 0x10102680
(/home/buildbots/ppc64be-clang-lnt-test/clang-ppc64be-lnt/stage1/projects/compiler-rt/lib/asan/tests/default/Asan-powerpc64-inline-Test+0x10102680)
    #1 0x10123754
(/home/buildbots/ppc64be-clang-lnt-test/clang-ppc64be-lnt/stage1/projects/compiler-rt/lib/asan/tests/default/Asan-powerpc64-inline-Test+0x10123754)

This is most likely due to  r279664, so please verify that by reverting it
locally.

next, do you know why the stack traces on the bot are not symbolized?
is that because llvm-symbolizer is not supported on PPC?
I need the symbolized stack traces to get some clue...

finally, will you be able to make the test smaller to narrow down the cause
of the failure?

If this is the only failure due to r279664 I'd rather disable this test on
PPC than revert r279664.
WDYT?

--kcc

On Wed, Aug 24, 2016 at 2:35 PM, Bill Seurer <seurer at linux.vnet.ibm.com>
wrote:

> On 08/24/16 13:22, Kostya Serebryany wrote:
>
>> Did reverting help?
>> If yes, I will need someone's help to debug this on PPC -- I really need
>> this change back.
>>
>
> Reverting it did help.  In the future if you (or anyone else) needs help
> with debugging a problem on power I can usually help.  Just let me know
> what you need done.
> --
>
> -Bill Seurer
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160824/8f7e75de/attachment.html>