This bug has been marked as needing help for a while.
Guillem Jover wrote:
> On Thu, 2013-04-25 at 14:51:29 +0200, Andreas Beckmann wrote:
> On 2013-04-25 10:39, Guillem Jover wrote:
> > > I guess the correct solution though, is to change update-fonts-*
> > > and any other script in xfonts-utils calling mkfont* from other
> > > maintainer scripts, to only run if xfonts-utils itself is
> > > configured or being configured, and calling these through
> > > xfonts-utils maintainer scripts too, so that if xfonts-utils and
> > > something else using it like gsfonts-x11 are both unpacked on the
> > > same run, the actual update-fonts-* executions will be delayed
> > > until xfonts-utils is itself configured, at which point the
> > > scripts should be able to run. I don't think this would be much
> > > more difficult to prepare.
> >
> > Sounds like a job for triggers (but there is currently a dpkg bug
> > that causes it to run triggers for packages that are not configured
> > (or their dependencies are not)
>
> Right, sorry, I didn't mention triggers exactly for that reason.
> Although this one should be easier to switch during jessie, as these
> packages do not need to await processing from xfonts-utils, which
> should avoid the cycles that enable that bug.
Well, it's been four years and the Stretch release is coming up. Sounds
like it's a good time to get this fixed.
update-fonts-{scale,alias} only manipulate text files so the only issue
here is update-fonts-dir (which calls mkfontdir, a simple script which
execs-with-flags mkfontscale, which links against libfontenc.so.1, which
links against zlib1g (whew!), which might not be configred when
update-fonts-dir is called.
I've never had to write dpkg triggers before but I'm happy to learn (and
I've glanced over the lengthy design doc).
If I understand the discussion, then I propose:
1) To convert xfonts-utils postinst and postrm update-fonts-dir calls
to dpkg triggers;
2) To add a note in update-fonts-dir's manpage pointing out the issue;
3) Submit a patch to dh_installxfonts in debhelper once part (1) above
is shown to be robust.
Comments?
--
Regards,
Branden
