Re: openssl-1.0-only packages?

[Thomas Klausner <tk%giga.or.at@localhost>]

On Mon, Feb 19, 2018 at 09:23:03AM -0500, Greg Troxel wrote:
> 
> Thomas Klausner <tk%giga.or.at@localhost> writes:
> 
> > On Mon, Feb 19, 2018 at 09:10:50AM -0500, Greg Troxel wrote:
> >> 
> >> Thomas Klausner <tk%giga.or.at@localhost> writes:
> >> 
> >> I think moving pkgsrc openssl to 1.1 is out of the question before
> >> 2018Q1.
> >> 
> >> I suspect we're going to need the sub-prefix 1.0 as we update to 1.1, in
> >> order to avoid breaking half of pkgsrc.   Despite the troubles, it seems
> >> worthwhile as part of making the most of a bad situation.
> >> 
> >> Are you suggesting some way for packages to build against pkgsrc 1.0
> >> while system is at 1.1, as a first step, before pkgsrc is updated?
> >
> > I'm not really sure about the solution, but I'm thinking of something
> > like adding an openssl-1.0 package with its own prefix, switching the
> > packages that need that version to it, and updating openssl to 1.1.
> 
> On netbsd-current, does pkgsrc use the native 1.1 exclusively now?

Yes.

> If
> so, we should get a sense of how bad things are from a bulk build on
> current.

There were some postings about this on pkgsrc-users. Some old versions
of popular languages are not fixed, but the latest versions all work
with 1.1. (python 3.4 (2.7 and 3.5+ are fine), ruby 2.2 and 2.3 (2.4+
are fine), php-5.6 (7+ are fine)).

> For things like python 2.7 with many dependencies, I suspect it may be
> better to patch them for 1.1.  But we can't fix everything.

python 2.7 is fine.

> Another question is how much various upstreams will cope given 6 months
> of waiting.  I realize NetBSD-current is an immediate issue, but it's
> not clear to me that faster updating to 1.1 in pkgsrc is optimal.

For python 3.4 and php-5.6, there are clear won't-fix statements from
upstream. 1.1 has been out quite a while, I think 6 more months won't
give us many more fixes.
 Thomas