Re: Problem building x11/gtk2 on older MacOS X

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Sun, 19 Sep 2010 23:09:30 -0400
    From:        John Ruschmeyer <jruschme%gmail.com@localhost>
    Message-ID:  
<AANLkTi=16G3tVCpqMPr+xj0tycUBBs-czXxUSO5K7cYG%mail.gmail.com@localhost>

  | Anybody seen this before? Any suggestions?

This looks frighteningly like the problems caused to gtk2 by the
pixman update - all caused by having alternative versions of the
libraries in the base system and in pkgsrc, where a pkgsrc library
wants to refer to other pkgsrc libraries, but ends up getting older
versions from the base system instead.

See the threads "travails of the pixman upgrade" and "Is pango still broken?"
on pkgsrc users over the past few days.

To confirm this, you could look and see if the undefined symbols you
are seeing exist in /usr/pkg/lib/libXau.so (and libXdmcp.so) and don't
exist in /usr/X11R6/lib/libXau.so (and libXdmcp.so) (assuming /usr/X11R6
is where your Darwin system has its base X11 - if not, I mean wherever the
base system puts its X11 libraries).

If your pkgsrc modular xorg has built enough that you can manage to
survive using only that, a quick fix would be to simply destroy the
base system X11 setup completely - something like

        mv /usr/X11R6 /usr/X11R6-ancient

should do it.   If my guess is correct, after that you'll probably have
no more problems of this kind (but of course, no system supplied X using
applications are likely to work any more).

If that's too radical, you could at least

        cd /usr/X11R6/lib       ; # or wherever it is
        mkdir old-so
        mv *.so old-so/.

That should prevent linking against any of the old libraries by
cc command lines, yet leave them there (in their libxxx.so.N form)
to be referenced by anything already linked against them (assuming that
Darwin works this way much the same as NetBSD does)

Or, you can look at the patch I just posted in my last message under
the "travails of the pixman upgrade" thread (just an hour ago) and see
if you can work out something similar that will work around your problem.

And one last time, all this (all of it) is predicated upon my guess that
what you are seeing is the same basic problem, just with different actors.

kre