Bug#487630: SUCCESS after removal of 00tetex.cnf (was: Re: Bug#487630: Bug 487630 exists also in version 2007.dfsg.2-3 (and version 2007.dfsg.2-4))

[Hilmar Preusse <hille42@web.de>]

On 04.11.08 Omer Zak (w1@zak.co.il) wrote:
> On Mon, 2008-11-03 at 17:42 +0100, Hilmar Preusse wrote:
> > On 03.11.08 Omer Zak (w1@zak.co.il) wrote:

Hi all,

> > > /etc/texmf/fmt.d:
> > > total 28
> > > drwxr-xr-x  2 root root 4096 Nov  3 16:10 .
> > > drwxr-xr-x 21 root root 4096 Nov  3 15:16 ..
> > > -rw-r--r--  1 root root 3704 Sep  2  2004 00tetex.cnf
> > > -rw-r--r--  1 root root 1364 Jul 28 01:53 00tex.cnf
> > > 
> > And here is the old 00tetex.cnf from teTeX times, which causes
> > your failures. I should have seen in earlier.
> > 
> > To which package belongs it?
> 
> According to dpkg -S, 00tetex.cnf belongs to no package.  The
> laptop has been running Debian Testing for few years (Sarge
> Testing, then Etch Testing, and now Lenny Testing), so it is
> probable that one of the versions of the package, which originally
> owned it, was uninstalled (or upgraded, rendering the file
> unneeded) but neglected to remove this file.
> 
Well, the postinst- and preinst scripts of teTeX in Debian are/were
rather a nightmare regarding "which files have to be
removed/fixed/changed to resurrect mistakes made before". You may
have a look at them just for fun. I guess you run once a testing
version of teTeX, which was broken and we didn't make a fix for this
special version.

Your 00tetex.cnf does not have the magic 
"# -_- DebPkgProvidedMaps -_-" snipped, hence is it read when
creating the fmtutil.cnf.

To get a fix I suggest that the preinst script of an important
package (tex-common ?) greps for that snippet through all files in
/etc/texmf/fmt.d/ and renames all files, which do not have it (e.g.
append the extension "unused") and inform the end user about it.
Packages providing files in there should have the snippet (I'm not
sure about this: update-fmtutil knows the DebPkgProvidedMaps
mechanism, but it is not documented in the Debian TeX policy).

Any further ideas?

H.
-- 
sigmentation fault