Re: pkgsrc picking up wrong fontconfig

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

On Sat, Aug 31, 2019 at 07:23:22AM -0400, Greg Troxel wrote:
> Thomas Klausner <tk%giga.or.at@localhost> writes:
> 
> >> When foo depends on bar, and foo's bl3 has bar, then anything that
> >> depends on foo (via bl3) has to end up with the *same bar* that foo
> >> used.  Lacking a mechanism to do exactly that, we have to have a safe
> >> approximation.
> >> 
> >> Really the whole notion of native vs pkgsrc X libs, other than a great
> >> native/modular divide, is not really sound.  If native isn't fully ok,
> >> so that we can build everything with only native, we have to either give
> >> up on native entirely or give up on the modules that are semi-ok.
> >
> > The longer I think about this, the more it reminds me of the gcc
> > issue. Basically it should be a global setting, but it's decided on a
> > per-package level.
> 
> Yes, that's what I was trying to say...

Sorry for being slow :)

> > For pango, the workaround seems easier than for gcc - we can just make
> > everyone depend on the latest pango (and latest fontconfig). Add the
> > fontconfig + freetype2 version patterns from the Makefile to the
> > buildlink3.mk file and bump the API (I think, since we don't want it
> > to be overridable) depends, then bump all users.
> 
> I don't think pango is the problem; it's just fontconfig.   And pango
> requiring new fontconfig  means every other program that uses fontconfig
> that could possibly also be linked with pango (or something that could be
> linked with something that could be linked with pango, transitive
> closure) has to use the new fontconfig.
> 
> So what I want to do is set the BUILDLINK_API_DEPENDS in fontconfig/bl3
> to 2.11.91.  That means native fontconfig on -8 will never be used by
> pkgsrc, I think.
> 
> (There may be a similar freetype2 issue.)

Yes, I think that's probably the best solution.
Will need a recursive bump though.
 Thomas