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

[Joerg Sonnenberger <joerg%bec.de@localhost>]

On Fri, Jun 12, 2020 at 07:59:11PM -0400, Greg Troxel wrote:
> Joerg Sonnenberger <joerg%bec.de@localhost> writes:
> 
> > On Fri, Jun 12, 2020 at 05:04:49PM +0200, Manuel Bouyer wrote:
> >> For rust, it looks like forcing MAKE_JOBS=1 is a working workaround.
> >
> > Not really, it just makes the problems a lot less visible.
> 
> Can you explain, or point to a PR:
> 
>   what is the core problem
> 
>   has this been actually fixed in current (and pulled up to 8 and 9)
> 
> My impression is that it is actually fixed, which is a lot of the
> motivation for not expecting RELEASE to work.

The core issue is that Rust does a number of operation which are UB
under POSIX rules. To be precise, POSIX specifies a number of things
that are valid after fork in a multiple-threaded program and Rust does
quite a few things not on that list. I'm therefore somewhat allergic to
the continous "rust hits bugs in the dynamic linker" comments, which are
simply miscategorisations. We might be pulling up a lot of the locking
improvements for misbehaving multi-threaded programs up to netbsd-9. I'm
not sure about netbsd-8 at this point.

Joerg