[llvm-dev] greendragon build noisy due to mmap_stress.cc

[Chris Matthews via llvm-dev]

The lit test directly pipes the output into the check, and does not store it in a temporary file.

Last week, I ran this test in a loop for 8 hours on a machine right after it failed in a build.  It did not reproduce.  Do you have any suggestions for reproducing?  Maybe we need to have the system under more stress before running the test?

> On Jan 20, 2016, at 1:41 PM, Kostya Serebryany <kcc at google.com> wrote:
> 
> 
> 
> On Wed, Jan 20, 2016 at 1:31 PM, Chris Matthews <chris.matthews at apple.com <mailto:chris.matthews at apple.com>> wrote:
> I worded that poorly, the Jenkins check I added will explain to the user that we know this fails sometimes.
> 
>> On Jan 20, 2016, at 1:30 PM, Chris Matthews <chris.matthews at apple.com <mailto:chris.matthews at apple.com>> wrote:
>> 
>> I have added a Jenkins check for this test, which explains why it fails on some builds.
>> 
>> Can we change the test to keep its output?  Will it just be blank anyways?
> 
> 
> The test simply prints everything to stdout. 
> It's the lit driver that hides the output. 
> 
> Maybe someone with a Mac can run this test manually (with exact same build command line)
> a few thousand times? I continue to suspect a kernel bug. 
>  
>> 
>>> On Jan 20, 2016, at 1:23 PM, Kostya Serebryany <kcc at google.com <mailto:kcc at google.com>> wrote:
>>> 
>>> The test fails again: http://lab.llvm.org:8080/green/job/clang-stage1-cmake-RA_check/9739/testReport/junit/ThreadSanitizer/ThreadSanitizer/mmap_stress_cc/ <http://lab.llvm.org:8080/green/job/clang-stage1-cmake-RA_check/9739/testReport/junit/ThreadSanitizer/ThreadSanitizer/mmap_stress_cc/>
>>> But unfortunately lit does not provide the original log, so we still don't know why... 
>>> 
>>> On Wed, Jan 13, 2016 at 1:11 PM, Chris Matthews <chris.matthews at apple.com <mailto:chris.matthews at apple.com>> wrote:
>>> These ran to completion without error. :(
>>> 
>>>> On Jan 12, 2016, at 3:23 PM, Chris Matthews <chris.matthews at apple.com <mailto:chris.matthews at apple.com>> wrote:
>>>> 
>>>> I’m running this on green-dragon-03 which is running OSX 10.10.5.
>>>> 
>>>> /Users/buildslave/jenkins/sharedspace/clang-stage1-cmake-RA_workspace\@2/clang-build/bin/clang++ ../llvm/projects/compiler-rt/test/tsan/mmap_stress.cc -lpthread 
>>>> while true; do for((i=0;i<10;i++)); do ./a.out || echo ===============FAILED================ & done ; wait; done
>>>> 
>>>> Seems like it is getting about 225 runs a minute, so it will take about 8 hours to complete 100k runs.
>>>> 
>>>>> On Jan 12, 2016, at 1:28 PM, Tobias Grosser via llvm-dev <llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>> wrote:
>>>>> 
>>>>> On 01/12/2016 10:26 PM, Kostya Serebryany wrote:
>>>>>> Hi Tobias,
>>>>>> 
>>>>>> What machine is that?
>>>>>> We have seen this and similar tests be flaky on older Linux kernels due
>>>>>> to kernel bug(s).
>>>>>> May I ask you to run the same test (just built with clang) on the same
>>>>>> machine for ~100000 times and see if it ever crashes?
>>>>>> 
>>>>>> clang++ ~/llvm/projects/compiler-rt/test/tsan/mmap_stress.cc -lpthread
>>>>>> while true; do for((i=0;i<10;i++)); do ./a.out || echo
>>>>>> ===============FAILED================ & done ; wait; done
>>>>> 
>>>>> Add Michael again (who might have access to this Machine).
>>>>> 
>>>>> Tobias
>>>>> _______________________________________________
>>>>> LLVM Developers mailing list
>>>>> llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>
>>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev <http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev>
>>>> 
>>> 
>>> 
>> 
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160121/19a34c61/attachment.html>