Re: [platform-testers] [Grep-devel] new snapshot available: grep-2.25.101-be82

[Jim Meyering]

On Mon, Sep 26, 2016 at 10:04 AM, Bruce Dubbs <address@hidden> wrote:
> Jim Meyering wrote:
>>
>> On Sun, Sep 25, 2016 at 11:27 PM, Bruce Dubbs <address@hidden>
>> wrote:
>>>>
>>>>     http://meyering.net/grep/grep-2.25.101-be82.tar.xz
>>
>> ...
>>>
>>> Builds clean with gcc-5.3.0.
>
>
>> Thank you for the quick testing and the reports.
>
>
> When I ran the expensive tests I had a failure.
> FAIL long-pattern-perf (exit status: 1)
>
> It appears to be a timing issue.
>
> $ lscpu
> Architecture:          x86_64
> CPU op-mode(s):        32-bit, 64-bit
> Byte Order:            Little Endian
> CPU(s):                4
> On-line CPU(s) list:   0-3
> Thread(s) per core:    1
> Core(s) per socket:    4
> Socket(s):             1
> NUMA node(s):          1
> Vendor ID:             GenuineIntel
> CPU family:            6
> Model:                 94
> Model name:            Intel(R) Core(TM) i5-6500 CPU @ 3.20GHz
> Stepping:              3
> CPU MHz:               3201.000

Thanks for the report.

How did you configure/compile that? This test runs grep with two
regexps. First it times the use of a shorter one, then it times the
use of a regexp whose byte count is 10 times larger than the first
one. The test requires that the latter duration be no longer than 20x
the duration of the first run. In your case, it actually took 80ms
longer than the 400ms that would be 20x.

As mentioned in its comments, that test is sensitive to load and
timing. Can you reproduce that failure consistently when running that
test in isolation? I.e., run this a few times:

  check -C tests TESTS=long-pattern-perf RUN_EXPENSIVE_TESTS=yes

I may end up changing it to run the quick one 5 times, and use the
minimum time to compare against the duration of the longer-running
command also being run 5 times (or until passing).