[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

[etch] RC bug in tetex-base will break upgrade to lenny

Hi stable release managers,

we request that you consider allowing an update of tetex-base in etch to
fix a bug that could break the etch-to-lenny upgrade under certain
circumstances.  Below is the detailed explanation.

Frans Pop has uncovered a RC bug in tetex-base:

#420390 Purging tetex transition packages removes texlive configuration files

The problem behind the bug is that the postrm was written, before
TeXLive existed, under the assumption that purging tetex-base means not
wanting any TeX system, and it tries to remove "old cruft" from sarge
and earlier from /etc/texmf.

Meanwhile, this "old cruft" is vital to TeX Live.  For those users who
just run "aptitude dist-upgrade", the transitional tetex-base package
built from the texlive-base sources will do the trick: It has an empty
postrm and can be purged without problems.

However, for users who notice the teTeX->TeX Live transition and
deselect tetex-base in the aptitude TUI (or similar), and mark it for
purging (either in the same aptitude run or any time later), for them
the transitional package will not help.  Depending on the order of
unpacking/configuration/removal, things *might* run smooth if the
purging is done at the same time as the upgrade.  But I think it's as
likely that people just deselect it and later find out they can purge
it, and then they would be bitten by the bug.

The new tetex-base package for etch would differ only in the changelog
and the postrm, which would lack these lines:

    for oldconffile in $UNUSED_CONFFILES $FORMER_UCF; do
      newname=`get_newfilename $oldconffile`
      remove_with_backups /etc/texmf/$newname
    for oldconfdir in $UNUSED_CONFDIRS; do
      rmdir /etc/texmf/$oldconfdir 2>/dev/null || true

and (as you prefer) the variable and function declarations used
therein (which make up about 85 of the 300 lines of the postrm).  

Would you agree with such a fix?

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

Reply to: