Re: pgsql: Fix some issues with step generation in partition pruning.

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> While I like having HP-UX buildfarm coverage, the status page says
> that gaur is running 10.20, which according to noted reliable source
> Wikipedia, was released in 1996 and went out of support in 2003. So I
> wonder whether testing that version is really a sensible thing to do
> at this point.

Certainly HPUX 10.20 isn't interesting as an OS anymore. I think the
point of this animal is mostly to test on HPPA hardware, and the reason
HPPA is interesting for us is mostly that it has such strange spinlocks,
and that's not OS-dependent.

If I still had support coverage on the machine, I'd upgrade to v11,
but I don't. (Wikipedia claims that HPUX is going EOL altogether
next year, so such a change wouldn't buy much for long anyhow.)

I experimented a year or so ago with installing OpenBSD, which seems
to be the last major open-source OS with HPPA support. I did get it
running, but it takes about twice as long to do "check-world" as
HPUX on the same machine, which does not speak well to the amount of
effort OpenBSD has put into the port. It already takes gaur near
six hours to do one buildfarm run; I don't want it to be twelve.
So I'm planning to keep running gaur as-is until (a) the hardware dies
or (b) we make a policy decision that breaks support for this
hardware.

pademelon (same machine, vendor compiler) is just there to keep
us honest on C89 support until v11 is EOL. There are other ways
to test C89 compliance, perhaps, but as long as this one is handy
I don't feel like putting effort into figuring that out. (Besides,
we have darn little coverage of not-gcc-derived compilers.)

regards, tom lane