On 03/17/2018 03:32 PM, Tom Lane wrote:
> Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:
>> I happened to be updating our machine running our buildfarm animals, and
>> I noticed something quite strange - the machine was unexpectedly running
>> out of disk space, which is rather suspicious as it's running just the
>> regression tests :-/
>
>> After a bit of investigation, I found this:
>> Right - the plpgsql_control.out output has about 58GB, which is somewhat
>> excessive I guess. The reason is fairly simple:
>
>> NOTICE: 2147483620..2147483647 by 10: i = 2147483620
>> NOTICE: 2147483620..2147483647 by 10: i = 2147483630
>> NOTICE: 2147483620..2147483647 by 10: i = 2147483640
>> NOTICE: 2147483620..2147483647 by 10: i = -2147483646
>> NOTICE: 2147483620..2147483647 by 10: i = -2147483636
>> NOTICE: 2147483620..2147483647 by 10: i = -2147483626
>> ... many more NOTICE messages ...
>
>> Looking at the plpgsql_call.out timestamp, this seems to be running
>> since January 1, which also correlates with the last results reported to
>> pgbuildfarm.
>
> Ouch. That test is in fact new as of 31 Dec, and what this seems to
> prove is that plpgsql's handling of loop-variable overflow doesn't
> work on fulmar.
>
Yeah :-( I'll try restarting the tests on fulmar, to see if it's
reproducible. If yes, I'll see if a newer version of ICC fixes it.
If not, I'll temporarily disable fulmar, so that the two other animals
on the same machine (magpie, treepie) are not stuck behind it. That
should tell us if the issue is limited to ICC, or if there's something
else wrong with that system.
regards
--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
