tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: the nature of pkgsrc support for OS versions (release vs stable)



On Fri, Jun 12, 2020 at 10:52:14AM -0400, Greg Troxel wrote:
> pkgsrc has long had a notion that we "support" the versions of NetBSD
> that are under maintenance, meaning the current and previous releases.
> This notion probably applies to other operating systems, but I'm going
> to just talk abotu NetBSD for now. 
> 
> (I'm going to ignore netbsd-7 in this discussion; that's going away soon
> and makes the ABI issue more complicated.)
> 
> Due to ABI stability rules, we have considered 8.0-RELEASE and
> 8.x-STABLE as more or less the same thing.  This is 99+% true.  Often we
> can make pkgsrc such that packages build on all points along the release
> to the current stable, for each branch, and obviously if that can be
> done reasonably we should do it.
> 
> Recently, we have found bugs in ld.so that break rust builds and these
> have been pulled up to the stable branches.
> 
> In theory, binaries built on a system with these fixes will work on a
> RELEASE system.  No one has so far presented evidence to the contrary.
> 
> Manuel has requested that anything needed to build packages other than
> "use RELEASE with defaults" be documented for bulk builders.  That's a
> reasonable request and I will create a wiki page about it.

thanks 

> 
> So, the big question is:
> 
>   Does pkgsrc need to support *building* on 8.0_RELEASE, even when that
>   version has bugs that are fixed in 8.X_STABLE, and it's not reasonable
>   to work around this?
> 
> I would say no, that while pkgsrc should ideally be buildable on
> RELEASE, if the problem is a bug in RELEASE, then it's entirely
> reasonable to expect people *building packages* to update to a STABLE
> with the fix.

I'd say: workaround the issue in pkgsrc as far as possible, and if really
it can't then get the fix in _STABLE and document it.
But, as far as possible we should also be able to build on _RELEASE.

For rust, it looks like forcing MAKE_JOBS=1 is a working workaround.

BTW I requested that the chroots for netbsd-9 on pkgbuild be updated to
_STABLE when the currently running build complete.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--


Home | Main Index | Thread Index | Old Index