Hello Peter,
>> The abort is by a client, but the code seems to only check the first
>> client of a thread. ISTM that if the second or later client abort it may
>> not be detected? Probably an intermediate aggregation at the thread level
>> is needed, or maybe a global variable, or as errors are counted somewhere,
>> it may be enough just to check that the count is non zero?
>
> fixed
With an aggregation. Fine with me.
>> The patch removes a check that there was an output and that no
>> transactions where processed. ISTM it should be kept. If a different exit
>> status is chosen on abort, that would allow to keep it easily.
>
> fixed
Ok.
>> Probably there should be some documentation changes as well?
>
> done
Ok.
Patch applies cleanly, compiles, global & local "make check" are ok.
Doc build is okay.
I noticed "for (int i"… pgbench is welcome to C99:-)
No further remarks. I turned the patch as ready on the CF app.
Attached a file I used for some manual tests.
--
Fabien.