Thread: buildfarm future

buildfarm future

From
Andrew Dunstan
Date:
I intended to post this anyway as a followup to the conference session, 
but I have been spurred to do it nowby the Pl/J(ava) discussion.

I outlined a number of possible directions for buildfarm, and we had a 
good discussion.

Based on that, my current intention is to make provision for the 
following, and pretty much in this order:

. an optional pgbench run at the end of each buildfarm run, plus 
supporting web page to plot stats
. clean and release web code
. extra core/contrib/PL regression tests that come with the core 
distribution but are not done by make {check, installcheck}
. extra core regression tests that are not in the core distribution 
(maybe download from pgfoundry)
. support for non-core projects - initially for slony-1 (because they 
asked me about it ages ago).


In connection with the last item, Thomas asked me at the conference to 
consider a setup for PL/Java. If, as I suspect, we end up with something 
similar to the status quo, I will certainly do so, but it might be a while.

If anyone with some good perl skill wants to jump in and help they will 
be welcome.

cheers

andrew


Re: buildfarm future

From
Kris Jurka
Date:
Andrew Dunstan wrote:
> 
> I intended to post this anyway as a followup to the conference session, 
> but I have been spurred to do it nowby the Pl/J(ava) discussion.
> 

Another thought I had today was the ability to attach notes to the web 
site.  For example a recent commit broke the 7.3 branch for everybody. 
I took a look at it and sent a patch for it to -patches, but it would be 
nice to flag that somehow in the web interface so that others would know 
it is being worked on and not to waste time on it.  You could attach 
notes at either the branch or machine-branch level and clear them when a 
clean build came through.

Kris Jurka