[compiler-rt] r272175 - [esan|wset] Reduce flakiness in samples test

Derek Bruening via llvm-commits llvm-commits at lists.llvm.org
Tue Jul 5 18:29:11 PDT 2016


It's not clear though, b/c if it matched sample #3 in array #0, there
should be a sample in array #1 (downsampled every 4).  Basically the
FileCheck output doesn't make sense to me: it would really help to see the
complete stdout.  Is that available anywhere on the bot in a log or
something?

On Tue, Jul 5, 2016 at 9:27 PM, Derek Bruening <bruening at google.com> wrote:

> I believe you mis-interpreted the problem: the MB is the total size, and
> that line matches.  The problem looks like (though FileCheck output is
> confusing) there are not enough samples to have any in array #1, which is a
> tougher problem to solve.
>
> On Tue, Jul 5, 2016 at 8:20 PM, David Majnemer <david.majnemer at gmail.com>
> wrote:
>
>> I took a wild stab in the dark at fixing this in r274587.  Hopefully it
>> is not just papering over an existing bug but its either that or reverting
>> the changes that added the test in the first place.
>>
>> On Tue, Jul 5, 2016 at 3:12 PM, Justin Bogner via llvm-commits <
>> llvm-commits at lists.llvm.org> wrote:
>>
>>> Derek Bruening via llvm-commits <llvm-commits at lists.llvm.org> writes:
>>> > Please provide a pointer to which bots.  It is not easy to find which
>>> ones
>>> > are being referred to at http://lab.llvm.org:8011/waterfall in order
>>> to see
>>> > the precise test output.
>>>
>>> The output is as David wrote, but here's a run that failed like this
>>> today:
>>>
>>>
>>> http://lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-debian-fast/builds/39465
>>>
>>> This seems to be happening fairly frequently right now.
>>>
>>> > On Fri, Jul 1, 2016 at 1:51 PM, Chandler Carruth <chandlerc at gmail.com>
>>> > wrote:
>>> >
>>> >> Any update here? the build bots are flaking because of this test.
>>> >>
>>> >> On Thu, Jun 23, 2016 at 11:22 PM David Majnemer via llvm-commits <
>>> >> llvm-commits at lists.llvm.org> wrote:
>>> >>
>>> >>> Still seems flaky:
>>> >>>
>>> >>>
>>> /home/llvmbb/llvm-build-dir/llvm-clang-lld-x86_64-debian-fast/llvm.src/projects/compiler-rt/test/esan/TestCases/workingset-samples.cpp:25:17:
>>> >>> error: expected string not found in input
>>> >>>  // CHECK-NEXT: =={{[0-9]+}}==# 0: {{[ 0-9]+}} {{KB|Bytes}} ({{[
>>> 0-9]+}}
>>> >>> cache lines)
>>> >>>                 ^
>>> >>> <stdin>:4:1: note: scanning from here
>>> >>> ==21268== Samples array #1 at period 80 ms
>>> >>> ^
>>> >>> <stdin>:11:36: note: possible intended match here
>>> >>> ==21268== EfficiencySanitizer: the total working set size: 32 MB
>>> (524289
>>> >>> cache lines)
>>> >>>
>>> >>> Looks like the regex pattern should be expanded further?
>>> >>>
>>> >>> On Wed, Jun 8, 2016 at 10:35 AM, Derek Bruening via llvm-commits <
>>> >>> llvm-commits at lists.llvm.org> wrote:
>>> >>>
>>> >>>> Author: bruening
>>> >>>> Date: Wed Jun  8 12:35:52 2016
>>> >>>> New Revision: 272175
>>> >>>>
>>> >>>> URL: http://llvm.org/viewvc/llvm-project?rev=272175&view=rev
>>> >>>> Log:
>>> >>>> [esan|wset] Reduce flakiness in samples test
>>> >>>>
>>> >>>> Generalizes the workingset-samples test to pass when a sample has a
>>> >>>> size of 0, which can happen on a loaded machine.
>>> >>>>
>>> >>>> Modified:
>>> >>>>     compiler-rt/trunk/test/esan/TestCases/workingset-samples.cpp
>>> >>>>
>>> >>>> Modified:
>>> compiler-rt/trunk/test/esan/TestCases/workingset-samples.cpp
>>> >>>> URL:
>>> >>>>
>>> http://llvm.org/viewvc/llvm-project/compiler-rt/trunk/test/esan/TestCases/workingset-samples.cpp?rev=272175&r1=272174&r2=272175&view=diff
>>> >>>>
>>> >>>>
>>> ==============================================================================
>>> >>>> --- compiler-rt/trunk/test/esan/TestCases/workingset-samples.cpp
>>> >>>> (original)
>>> >>>> +++ compiler-rt/trunk/test/esan/TestCases/workingset-samples.cpp Wed
>>> >>>> Jun  8 12:35:52 2016
>>> >>>> @@ -22,12 +22,12 @@ int main(int argc, char **argv) {
>>> >>>>    // We only check for a few samples here to reduce the chance of
>>> >>>> flakiness.
>>> >>>>    // CHECK:      =={{[0-9]+}}== Total number of samples: {{[0-9]+}}
>>> >>>>    // CHECK-NEXT: =={{[0-9]+}}== Samples array #0 at period 20 ms
>>> >>>> -  // CHECK-NEXT: =={{[0-9]+}}==#   0: {{[ 0-9]+}} KB ({{[ 0-9]+}}
>>> cache
>>> >>>> lines)
>>> >>>> -  // CHECK-NEXT: =={{[0-9]+}}==#   1: {{[ 0-9]+}} KB ({{[ 0-9]+}}
>>> cache
>>> >>>> lines)
>>> >>>> -  // CHECK-NEXT: =={{[0-9]+}}==#   2: {{[ 0-9]+}} KB ({{[ 0-9]+}}
>>> cache
>>> >>>> lines)
>>> >>>> -  // CHECK-NEXT: =={{[0-9]+}}==#   3: {{[ 0-9]+}} KB ({{[ 0-9]+}}
>>> cache
>>> >>>> lines)
>>> >>>> +  // CHECK-NEXT: =={{[0-9]+}}==#   0: {{[ 0-9]+}} {{KB|Bytes}} ({{[
>>> >>>> 0-9]+}} cache lines)
>>> >>>> +  // CHECK-NEXT: =={{[0-9]+}}==#   1: {{[ 0-9]+}} {{KB|Bytes}} ({{[
>>> >>>> 0-9]+}} cache lines)
>>> >>>> +  // CHECK-NEXT: =={{[0-9]+}}==#   2: {{[ 0-9]+}} {{KB|Bytes}} ({{[
>>> >>>> 0-9]+}} cache lines)
>>> >>>> +  // CHECK-NEXT: =={{[0-9]+}}==#   3: {{[ 0-9]+}} {{KB|Bytes}} ({{[
>>> >>>> 0-9]+}} cache lines)
>>> >>>>    // CHECK:      =={{[0-9]+}}== Samples array #1 at period 80 ms
>>> >>>> -  // CHECK-NEXT: =={{[0-9]+}}==#   0: {{[ 0-9]+}} KB ({{[ 0-9]+}}
>>> cache
>>> >>>> lines)
>>> >>>> +  // CHECK-NEXT: =={{[0-9]+}}==#   0: {{[ 0-9]+}} {{KB|Bytes}} ({{[
>>> >>>> 0-9]+}} cache lines)
>>> >>>>    // CHECK:      =={{[0-9]+}}== Samples array #2 at period 320 ms
>>> >>>>    // CHECK:      =={{[0-9]+}}== Samples array #3 at period 1280 ms
>>> >>>>    // CHECK:      =={{[0-9]+}}== Samples array #4 at period 5120 ms
>>> >>>>
>>> >>>>
>>> >>>> _______________________________________________
>>> >>>> llvm-commits mailing list
>>> >>>> llvm-commits at lists.llvm.org
>>> >>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>> >>>>
>>> >>>
>>> >>> _______________________________________________
>>> >>> llvm-commits mailing list
>>> >>> llvm-commits at lists.llvm.org
>>> >>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>> >>>
>>> >>
>>> > _______________________________________________
>>> > llvm-commits mailing list
>>> > llvm-commits at lists.llvm.org
>>> > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at lists.llvm.org
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160705/fc448ed9/attachment.html>


More information about the llvm-commits mailing list