Bug#343545: tetex-bin: config fails; updmap cannot find arabtex.map

[Frank Küster <frank@debian.org>]

severity 343545 important
# not reassigning to arabtex for now because it's the tex maintainers' fault
thanks

Lionel Elie Mamane <lionel@mamane.lu> wrote:

> Package: tetex-bin
> Version: 3.0-12
> Severity: grave
> Justification: renders package unusable

Dear Lionel, 

thanks for reporting that bug; but it's not correctly rated with
"grave".  arabtex.map is part of a separate package, and is only
installed by a subgroup of our users, and it clearly doesn't make the
package unusable for most of them.  

> I'm fairly sure that I haven't manually changed any tex setting on
> this computer. This is an upgrade through aptitude:
>
> [UPGRADE] tetex-base 3.0-10 -> 3.0-11
> [UPGRADE] tetex-bin 3.0-10.1 -> 3.0-12
> [UPGRADE] tetex-doc 3.0-10 -> 3.0-11
> [UPGRADE] tetex-extra 3.0-10 -> 3.0-11
> [UPGRADE] tethereal 0.10.12-6 -> 0.10.13-1
> [UPGRADE] tex-common 0.9 -> 0.13
> [UPGRADE] arabtex 3.11-7 -> 3.11-9
>
>
> I suppose that if arabtex got configured, it would move arabtex.map in
> the right place and permit tetex to be configured:

Yes, you are right.  It's a bug in arabtex, but in fact it's the code
that Norbert and I suggested that causes the problem.  This is what
happens:

- arabtex 3.11-7 has its map file in /etc/texmf/dvips/map/arabtex.map,
  and an updmap.cfg snippet in /etc/texmf/updmap.d/10arabtex.cfg that
  references that map file.  But the magic comment is missing in
  10arabtex.cfg in 3.11-7

- before arabtex is unpacked, the preinst detects that arabtex.map is
  unchanged and removes it

- arabtex and the tetex apckages are unpacked, the new arabtex.map is
  still arabtex.map.dpkg-new, and the new 10arabtex.cfg (with the magic
  comment) is still 10arabtex.cfg.dpkg-new, while the old 10arabtex.cfg
  is still there

- tetex-bin is configured and runs update-updmap and updmap, and now it
  looks for arabtex.map but cannot find it.

So far for the analysis.  I don't have enough time to think of a clean
solution.  Maybe, since the conffile placements change, anyway, we
should do the removals and moves in postinst?

Regards, Frank
-- 
Frank Küster
Inst. f. Biochemie der Univ. Zürich
Debian Developer