Re: buildfarm's typedefs list has gone completely nutso

Gum

On 2019-07-18 11:27:49 -0400, Tom Lane wrote:
> Andres Freund <andres(at)anarazel(dot)de> writes:
> > Looks like that made the generated typedef lists sane. Any residual
> > complaints?
>
> Before ack'ing this, I've been waiting around for komodoensis to update
> its typedefs list, in the hopes that the last couple of bogus entries
> would go away. It still hasn't, and I just realized from looking at its
> config that you have it set to do so at most twice a week:

I've changed that yesterday to be just min_hours_since=48. I can easily set it
to more frequent, especially after using -Wi makes the whole step a lot
faster - I just though that swamping the buildfarm with the additional
data would be unnecessary? But perhaps it doesn't matter in comparison
to the normal stage logs.

Boosted to min_hours_since=24 for now.

> 'optional_steps' => {
> 'find_typedefs' => {
> 'branches' => [
> 'HEAD'
> ],
> 'min_hours_since' => 25,
> 'dow' => [
> 1,
> 4
> ]
> }
> },
>
> That seems a bit, er, miserly.

Comes from the default suggestion in build-farm.conf.sample:
# find_typedefs => { branches => ['HEAD'], dow => [1,4],
# min_hours_since => 25 },

I'm happy to boost that to something a lot more aggressive.

At least the REL_11_STABLE logs are from after the change to use -Wi
instead of -W, and therefore ought to be correct.

bf(at)andres-postgres-edb-buildfarm-v1:~/src/pgbuildfarm-client-stock$ TZ=UTC ls -l run_build.pl
-rwxr-xr-x 1 bf bf 63701 Jul 17 17:16 run_build.pl

komodoensis 2019-07-17 03:04:01 HEAD 3308
komodoensis 2019-07-17 17:32:02 REL_12_STABLE 3296
komodoensis 2019-07-17 23:18:24 REL_11_STABLE 3204

Greetings,

Andres Freund