Bug#354517: Include additional checks in update-language
Any reason why this wasn't sent to the bug? I hope it's just a mistake.
Norbert Preining <preining@logic.at> wrote:
> Pro inclusion:
> - it may be useful for some
> - language check were not done up to now
> - doesn't clutter the file since the functions are not too long
>
> Contra inclusion:
> - there are already tests for the .ini file in fmtutil and for missing
> map file in updmap (but language is missing) and the missing test
> should better be included upstream
I fear including the language test into fmtutil would be extremely
difficult. If it's possible at all, then only for LaTeX, or rather for
"everything which we can safely assume to use
TEXMF/tex/generic/config/language.dat", but even then the *hyphen.tex
files are simple input files and can, even if present, include anything
that's perhaps not in fact present, and cause the format creation to
fail.
Of course we won't catch such problems, either. But I think that for
the decision "should this be implemented in fmtutil" it's important
whether it would be a check that really helps in the real world,
complicated as it is, and not just catch a subset of the errors - it
would be a little like pretending everything is fine, while in fact
fmtutil has no way to detect whether it is.
On the other hand, for a Debian utility like update-language (I know,
update-fontlang) it makes sense to make just a simple check - this would
catch errors that are more specific to Debian and its modularized setup:
Buggy packages; people who mix up installation into TEXMFHOME and
TEXMFLOCAL, etc.
> - it clutters the update-fontlang script a bit
>
> Currently there is a update-fontlang.withchecks in the svn repos which
> implements the checks, and is otherwise the same as the real
> update-fontlang.
I've had a look at the functions now (only after writing the three
pro-paragraphs above...) and it seems to me as if there's a fundamental
problem and source of confusion:
+ if [ -z $(kpsewhich -format=tex $hyph) ] ; then
+ cat >&2 <<EOF
+The config file $fn references a file not in the kpathsea database:
+ $hyph
+This may be ok, but it could also be a typing error.
+EOF
+ fi
+ done
This means that if update-* is called before mktexlsr, there will be
that spurious warning. But in many cases it doesn't make sense to move
mktexlsr earlier, because it must be called "in the end" anyway.
Taking things together, I'm unsure, but I tend to vote against the
check.
Regards, Frank
--
Frank Küster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Zürich
Debian Developer (teTeX)
Reply to: