On Sun, Sep 3, 2023 at 6:42 AM Andrew Dunstan <andrew@dunslane.net> wrote:
> I guess the next thing would be to test it on a few more platforms and also to see if we need to expand the coverage
oflibpq for the intended uses.
Nice. It works fine on my FreeBSD battlestation after "sudo pkg
install p5-FFI-Platypus" and adjusting that lib path. I wonder if
there is a nice way to extract those constants from our headers...
It's using https://sourceware.org/libffi/ under the covers (like most
other scripting language FFI things), and that knows calling
conventions for everything we care about including weird OSes and
architectures. It might be a slight pain to build it on systems that
have no package manager, if cpan can't do it for you? I guess AIX
would be the most painful?
(Huh, while contemplating trying that, I just noticed that the GCC
build farm's AIX 7.2 system seems to have given up the ghost a few
weeks ago. I wonder if it'll come back online with the current
release, or if that's the end... There is still the
overloaded-to-the-point-of-being-hard-to-interact-with AIX 7.1 (=EOL)
machine.)
> I confess I'm a little reluctant to impose this burden on buildfarm owners. We should think about some sort of
fallbackin case this isn't supported on some platform, either due to technological barriers or buildfarm owner
reluctance.
I guess you're thinking that it could be done in such a way that it is
automatically used for $node->safe_psql() and various other things if
Platypus is detected, but otherwise forking psql as now, for a
transition period? Then we could nag build farm owners, and
eventually turn off the fallback stuff after N months. After that it
would begin to be possible to use this in more interesting and
advanced ways.