Bug#402994: use new libpaper hook to track system paper size

[Frank Küster <frank@debian.org>]

Stephen Gildea <stepheng+debian-bts@gildea.com> wrote:

> I was motivated to try to fix the paper-size problem for everyone
> (well, everyone in the USA) when I was helping a friend who was
> totally confused about why his LaTeX document wouldn't print.  

Yes, that's the really bad point.

> It seems to me that having the user configure libpaper, which runs
> texconfig-sys and changes files in /etc, is no different from the user
> directly running texconfig to change files in /etc.  If there is a
> policy violation, this patch doesn't make it worse.  

If the user runs paperconfig and its hooks run texconfig-sys and change
files in /etc/texmf, this is not a policy violation.  But AFAIK the
hooks will also be run by libpaper's postinst - yes, they will, at least
if libpaper has just been installed, in other words if I install both
TeX and libpaper at the same time.  And if that all is supposed to make
sense, also tetex-bin's/texlive-bin's postinst needs to call the hook,
too:  Otherwise a user who installs TeX after libpaper doesn't get dvips
configured for the system paper size.

> Perhaps we can
> separate the two problems of setting default paper size and moving
> automatically-edited files from /etc to /var ?

Yes, that would be the solution.  But since we can't handle all of
config.ps as a configuration file (I fear), we need to separate the
papersize setting out.

Or wait.  We could

- ship our default config.ps in /usr/share/tetex-bin

- in postinst, run the equivalent of the libpaper hook on this file,
  creating a temporary file which differs from the default one only by
  the papersize setting

- use ucf to install the temporary file in /etc/texmf/dvips/

That way, any other admin changes are handled by ucf just like dpkg
would handle conffile changes, but we can always get the papersize from
libpaper.  And if the libpaper hook also uses ucf, later changes to the
default paper size also won't cause bogus conffile prompts.

I really hope I haven't missed anything...

Regards, Frank
-- 
Dr. Frank Küster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Zürich
Debian Developer (teTeX/TeXLive)