>> Would there be any value in setting up a project on sourceforge to
>> make use of their compile farm? I know that it doesn't cover all
>> platforms, but it would perhaps be a start to mechanical compile and
>> regression testing.
> I haven't looked at the platforms available in the compile farm
> recently, but afaik regression coverage for over half a dozen platforms
> already happens without (extra) effort: Tom Lane has three or more
> platforms, I've got Linux, Bruce has BSDI, Marc has FreeBSD, we have
> some active W32 developers, etc etc.
I run HPUX and Linux/PPC routinely, so that's only two here. Still, we
have reasonable coverage among the core team and a bunch more platforms
used by active pgsql-hackers people. Also, the project does have an
Alpha in-house at hub.org (if Marc ever gets it back into commission
after that failed OS reinstall...)
> What would SF add to this mix?
The current list of machines at cf.sourceforge.net seems to be
lqqqqqqqChoose compile farm server...qqqqqqqk x A. [x86] Linux 2.2 (Debian 2.2)
x x C. [x86] FreeBSD (4.2-stable) x x
x x G. [Alpha] Compaq Tru64 (5.1) x x H. [Alpha] Linux 2.2 (RedHat 7.0)
x x x x L. [Sparc - E240] Linux 2.2 (Debian
2.2) x x M. [Sparc - E240] Sun Solaris (8) x x
x x Exit x
mqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqj
I think I'll go try a build on that Solaris 8 machine, since we've heard
some reports of problems on Solaris. However, I'm not sure that we need
any organized use of their compilefarm. If they made it easy to
*automatically* run build/install/regress test on multiple machines,
I could see the facility being useful (especially so once a few more
platforms are offered). But right now it looks like it's just shell
access to platforms other than your own, which is not going to help us
all that much.
regards, tom lane