Re: lmodern fonts and sarge
Atsuhito Kohda <kohda@pm.tokushima-u.ac.jp> wrote:
> Please note that it could be not so easy as you might
> think; for example bugs #207786 and/or #229725 had
> very critical bad side-effects to tetex packages.
>From a quick glance, #207786 was due to pxfonts thinking updmap is
always available when its prerm is run, which is wrong. My package
shouldn't exhibit this problem since it has no prerm and the calls to
update-updmap, mktexlsr and updmap in its postrm are all followed by
"|| true".
As for #229725, I agree it is nasty. Its cause seems to me to be the
status change of /usr/share/texmf/dvips/config from directory to
symbolic link between tetex-* 1 and 2. *If* it is possible to have a new
arabtex version that installs correctly with tetex 1 and 2 (perhaps the
maintainer scripts can check whether /etc/texmf/dvips/updmap exists,
whether /usr/share/texmf/dvips/config is a directory or a symlink...),
then I think the "Conflicts: arabtex (<= version)" solution would be
acceptable: the woody -> sarge transition would upgrade arabtex on
tetex 1 first, and then tetex could be upgraded. For arabtex to work
with tetex 2, it would be necessary that arabtex.map be in
/etc/texmf/dvips/ and not in /usr/share/texmf/dvips/config/ like it is
in 3.10-5. I didn't check if everything could work, though. I am not the
arabtex maintainer and don't have so much time at the moment.
Now, how does it relate to the problem at hand? #207786 should not
happen with lmodern as explained above. And #229725 is IMHO irrelevant
for the following reason: the problem with #229725 is that you need to
upgrade both arabtex and tetex (and they remain separate); tetex cannot
be upgraded while the old arabtex is installed; and I suppose that
arabtex cannot (yet) be upgraded with the old tetex installed. With
lmodern, the problem would be much less nasty as far as I can see. The
first tetex version incorporating lmodern could conflict with lmodern
(<< foo) where lmodern version foo would be a dummy package to make the
transition smooth. Therefore, the sarge -> sarge+1 transition would
cause the upgrade of lmodern to version foo (actually, the removal of
the lmodern fonts from the system) followed by the upgrade of tetex.
Nothing dreadful.
> Also I guess difficulty of updmap might be not Debian
> specific to some extent. It was drastically changed
> its feature between 1.0 and 2.0
I am not blaming the Debian tetex maintainers, if that was your
impression. I understand very well how difficult it can be to manage
smooth upgrades of big software packages like tetex when the
infrastructure changes. Actually, despite its little problems, I find
the /etc/texmf/updmap.d mechanism much better than the flat
/etc/texmf/dvips/updmap file there was in tetex 1...
> It is easy to criticize *now* that the psfonts.map
> is a generated file so should not be under /etc, but
> in tetex 1.0 it was a normal file and when I packaged
> tetex-2.0 for the first time experimentally, I never
> recognized that it was generated file. It is true that
> we gradually recognized its features step by step and
> we, of course, recognized now the problem clearly.
Please calm down. Saying there is a problem in a package is not
equivalent to blaming its maintainer. Should we stop filing bug reports
because it might offense the maintainers? Come on!
> But we should consider a side-effect too if we change a basic behavior
>of postinst etc. in tetex packages because it could cause problems to
>other related packages.
Sure. Breakage is always a problem.
> In RedHat9, there is a Japanese TeX system with tetex
> but it didn't work at all by default and I found a Japanese
> RedHat user explained in his home page that "It is indeed
> very difficult to set up Japanese TeX to work well with
> RedHat. First edit a file like this way and then ...."
>
> In FreeBSD 5.0, I failed even to install Japanese TeX package.
So, similar situation to that experienced by people who want to produce
PDF files with TeX on Debian with a font that looks like Computer Modern
containing one of French, German, Spanish, Italian, Portuguese, Polish,
Norwegian and probably many other european languages that I don't know.
> I believe tetex of Debian can say "Our Priorities are Our
> Users and Free Software"
It is not bad, but it can be better.
If it appears that the consensus is that lmodern should not be included
because it might cause problems to tetex packages, then so be it. Write
it clearly and I'll ask for its removal before it reaches the archive,
no problem. Users will of course need these fonts and they will install
Michael's or my package from a non-Debian source, and run into more
problems during upgrades because the tetex packages wouldn't have to add
a Conflicts for a non official package, but who cares? Or they won't
install it and simply think one cannot produce correct PDF files with
TeX, at least for some of them.
It's OK for me. Installing the packages on my friends' machines and
managing the small breakage that would not occur (since I would add a
Conflicts: with tetex (>= version that adds lmodern) pretty quickly and
upgrade their lmodern package before they transition from sarge to
sarge+1) would not be very difficult.
Well. I would just be a bit bitter, since as you can infer from my
previous mail, I worked very hard to produce a _good_ package really
quickly so that it can be included in sarge. And when I write _good_, I
mean it. I've been extremely careful when writing the maintainer scripts
so that they are idempotent and manage the configuration file
/etc/texmf/updmap.d/10lmodern in a fully Policy-compliant way. And the
mental process of getting them right is thoroughly documented:
for file in lmodern.preinst.in lmodern.postinst lmodern.postrm; do \
wc -l "$file" && printf "Comments: " \
&& grep -v '^[ ]*#' "$file" | wc -l; \
done
56 lmodern.preinst.in
Comments: 33
90 lmodern.postinst
Comments: 51
118 lmodern.postrm
Comments: 54
I didn't start the thread simply claiming I was going to package the
thing. Just asked what could be done to get these fonts in sarge, and
offered to do the work if that was required to achieve this goal.
Perhaps I didn't wait for enough comments. But I must really work on
other things next month and the package had to hit unstable very quicky
to be tested a little bit before sarge, so I started with the first few
positive comments.
--
Florent
Reply to: