Re: [Pgbuildfarm-members] [Fwd: PGBuildfarm member narwhal Branch HEAD Status changed from OK to InstallCheck failure]

Andrew Dunstan wrote:
> Dave Page wrote:
>> This was another occurance of the strange create index failure on
>> Narwhal - unfortunately, despite having 'keep_error_builds' => 1 in my
>> BF config it seems to have removed the tree so I can't get the dump that
>> Tom wanted.
>>
>> Does anyone know why the keep_error_builds option didn't work in this
>> case? Or have I misinterpreted it's meaning?
>>
> 
> 
> Well, "anyone" will usually be me ;-)

Yeah, I know but I didn't want to presume :-)

> To the best of my knowledge and belief it does work. This is the code it
> should cause to be executed:
> 
>     my $timestr = strftime "%Y-%m-%d-%H:%M:%S", localtime($now);
>     move("$pgsql", "pgsqlkeep.$timestr");
>     move("inst", "instkeep.$timestr")    if (-d "inst") ;
> 
> Please see if you can trap the script stdout/stderr somewhere so that next
> time it fails you can send it to me.

I already log the run output - there's nothing unusual in there.

> How exactly do you have the buildfarm runs automated? Maybe that does a
> little cleaning of its own?

A simple batch script, run from the task scheduler which fires off the
script. It doesn't do any cleanup at all.

Regards, Dave.