Re: [PATCH] tests: tail-2/pid: use a 3-second timeout, not 1
From:
Jim Meyering
Subject:
Re: [PATCH] tests: tail-2/pid: use a 3-second timeout, not 1
Date:
Thu, 01 Oct 2009 07:43:21 +0200
Pádraig Brady wrote:
> Jim Meyering wrote:
>> FYI, I ran make -j9 check on a fast quad-core system earlier
>> today and saw this failure:
>>
>> FAIL: tail-2/pid (exit: 1)
>> ==========================
>> ...
>> + timeout 1 tail -s.1 -f /dev/null --pid=2147483647
>> + test 124 = 124
>> + fail=1
>>
>> To me, that means it took more than 1 second for the child to start,
>> and so the parent's 1-second timeout expired, causing the failure.
>
> Fair enough. Though considering it's a fast quad-core system,
> latencies of over 900ms are surprising.
When you punish the I/O subsystem as much as a few of the tests
do, it happens more often than I'd like. (tests/cp/link-heap is one
of the biggest culprits)
> Note we seem to be sleeping for this case which isn't helping.
> I'll look at whether we can check the pid before doing a sleep
> for the next release.
>
> I've nothing outstanding that needs to go into 7.7
Thanks.