On 2021/05/15 12:27, Jeremie Courreges-Anglas wrote:
> On Sat, May 15 2021, Marc Espie <es...@nerim.net> wrote:
> > On Sat, May 15, 2021 at 10:45:19AM +0200, Omar Polo wrote:
> >> "openbsd6" comes directly from python' sys.platform. I find that
> >> strange too, but I assume it's for a good reason. Python (and python
> >> porting) is not really my area of expertise.
> >
> > Nope, there's about zero good reason. Oh well, that's python
net-snmp has similar (can't remember if i already preempted it or not),
and i think some other things that will show up once the release
number is bumped.
we can't really expect other projects to know how openbsd versioning
is done, and that approach does sort-of make sense for some other os
(at least sometimes).
> I know I've had to change some ports to use the
> sys.platform.startswith("openbsd") idiom, indeed it's a bit silly.
sys.platform.startswith("openbsd") seems the best thing to do really,
that's how python stdlib and tests are now doing it.
I suggest if we're patching ports that this is how we should patch them
as it will cope with old python versions, old+new openbsd, and will
still cope if python does later switch to just plain 'openbsd'.
> OpenBSD 7.0 ("openbsd7") is coming soon, now is probably the best time
> to push for a change upstream.
agreed this is the sane way to go, and it would probably be simpler
all round if they did it sooner rather than later.
If anyone contacts them and they push back about doing it outside of
a major python version update, point out that they didn't seem too
concerned about that when they changed EXT_SUFFIX for OpenBSD in
minor version 3.8.7, and that any downstream software currently
using "openbsd6" will be broken in a few months time anyway. ;)
