Bug#266497: apt-get install textex-bin fails
Frank Küster <frank@debian.org> wrote:
> I think the postinst should only *try* to do its work (running mktexlsr,
> update-*, fmtutil), and still exit 0 if it cannot. This will not make
> the packages better, but it will drop the severity of bugs, because the
> installation can proceed without errors.
>
> I doubt, however, that just adding a || true to all those lines makes
> sense, and I fear we have to postpone this to after sarge.
I have looked into this, and in fact it is quite simple. We hardly ever
had reports because the update-* scripts failed[1]. We do some error
handling only for fmtutil and updmap, and I have simply changed this
error handling from
echo "Error message..."
exit 1
to a critical debconf note. This will not solve any bug - any real bug
will still lead to a FTBFS of build-depending packages - and thus it
does not mean hiding things.
I hope that this will lower the severities of bugs in the cases were the
postinst previously failed because of users not accepting changed
conffiles, or because of strange behavior of dpkg/ucf (those cases where
we suspected power failures, disk full etc.).
It won't help if all formats, or the central ones for most users, like
latex, are not built. But at least the cases were only some formats
cannot be built will lead to less severe bugs. And perhaps people will
think about what they did to conffiles when they read the "Possible
causes are.." part of the debconf note.
We will never find out whether a user error was the cause of this
particular bug (except if Thomas remembers that he made this decision),
and I also have committed a more severe warning into
05TeXMF.cnf. Therefore I intend to close that bug with the next upload.
Regards, Frank
[1] Except the cases where ucf failed because of a corrupted
database. But in these cases, we now know what's going on, and can
usually close the bug fast.
--
Frank Küster, Biozentrum der Univ. Basel
Abt. Biophysikalische Chemie
Reply to: