Re: Buildfarm

Tom Lane wrote:

>"Larry Rosenman" <ler(at)lerctr(dot)org> writes:
>
>
>>Since tom seems to be fixing the back branches, I added 7.3 and 7.2 to
>>firefly's set of branches it tries. Unfortunately
>>neither one went green :(.
>>
>>
>
>There's a limit to how much time I'm prepared to put into that endeavor
>;-) and one Saturday afternoon is about it.
>
>Somewhere along here there needs to be a discussion about what our goals
>are. IMHO the back branches are supposed to be *stable* branches; that
>means we only touch them to fix moderately-critical bugs. Fixing
>cosmetic regression failures has never been classed as a critical bug,
>and I don't think that the existence of the buildfarm should cause us to
>start treating them as critical. So, while I was willing to back-port
>one or two minor changes that looked pretty safe, I think we have to be
>very conservative about doing that, especially for branches as far back
>as 7.2 and 7.3.
>
>
>
>

Yeah. My view is that we should try to keep HEAD, and the latest one or
two STABLE branches as clean as possible. Branches whose release long
predates even the existence of buildfarm don't matter as much. I'm
certainly happier now than I was when we diodn't have any buildfarm
members that were clean on 7.2 or 7.3.

Having said that, changes in test result files should possibly be
treated a bit more liberally than changes in production code. I agree
about what stable means - we have a good reputation on that front and we
should protect it.

cheers

andrew