Re: pgbench tap tests & minor fixes.

On 2017-09-11 15:02:21 -0400, Andrew Dunstan wrote:
> 
> 
> On 09/11/2017 01:58 PM, Tom Lane wrote:
> > Andrew Dunstan <andrew.dunstan@2ndquadrant.com> writes:
> >> On 09/08/2017 09:40 AM, Tom Lane wrote:
> >>> Like you, I'm a bit worried about the code for extracting an exit
> >>> status from IPC::Run::run.  We'll have to keep an eye on the buildfarm
> >>> for a bit.  If there's any trouble, I'd be inclined to drop it down
> >>> to just success/fail rather than checking the exact exit code.
> >> bowerbird seems to have been made unhappy.
> > I saw that failure, but it appears to be a server-side crash:
> >
> > 2017-09-10 19:39:03.395 EDT [1100] LOG:  server process (PID 11464) was terminated by exception 0xC0000005
> > 2017-09-10 19:39:03.395 EDT [1100] HINT:  See C include file "ntstatus.h" for a description of the hexadecimal value.
> >
> > Given the lack of any log outputs from process 11464, it's hard to tell
> > what it was doing, but it seems not to be any of the backends running
> > pgbench queries.  So maybe an autovac worker?  I dunno.  Anyway, it's
> > difficult to credit that this commit caused the failure, even if it did
> > happen during the new test case.  I'm inclined to write it off as another
> > one of the random crashes that bowerbird seems prone to.
> >
> > If the failure proves repeatable, then of course we'll need to look
> > more closely.
> >
> > 			
> 
> 
> Hmm, it had several failures and now a success. Will keep an eye on it.

There just was another failure like that
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=bowerbird&dt=2017-09-19%2001%3A42%3A20
I first thought it might be the new recovery tests, or the changes
leading to its addition, but it's a different test and in the middle of
the run.  Even so, I'd have looked towards my commit, except that
there's a number of previous reports that look similar.

Any chance you could get backtraces on these?

Greetings,

Andres Freund