Jim Meyering wrote:
> Pádraig Brady wrote:
> ...
>>> Odd... that doesn't fail on any of the systems where I tried it:
>>> rawhide, fedora 11, debian unstable.
>> Yep I noticed it triggered on my linux-2.6.22 glibc-2.6-1 box
>> but not on my F11 laptop. I've setup the test in the attached
>> to output more than 16KiB per process which triggers on F11 at least.
> ...
>> +(mkdir tmp && cd tmp && seq 500 | xargs touch)
>> +
>> +# This will output at least 16KiB per process
>> +# which is enough to trigger on Fedora 11 at least
>> +(find tmp tmp -type f | xargs -n500 -P2 md5sum) |
>> +sed -n '/[0-9a-f]\{32\} /!p' |
>> +grep . > /dev/null && fail=1
>
> Thanks. That's better.
> However, it too sometimes fails to fail:
> [here, the grep matched nothing 2 of 200 times]
Wow that's much too infrequent and suggests your
system is very fast and getting xargs to reap
and start another process, perturbs the buffering
enough. I'll add "tmp tmp tmp" to the test.
thanks,
Pádraig.
