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

tetex-3.0 and the repackaging of the orig.tar.gz files



Hi all,

in the last months, I have learned a lot about orig.tar.gz files in
Debian (see for example Bug #278524). The bottom line is: Wherever
possible, a package should have a pristine orig.tar.gz file that is
byte-to-byte identical to the file downloaded from upstream.

On the other hand, I have also found out that it is indeed possible for
the teTeX package to use pristine orig.tar.gz files: For tetex-bin, we
just have to remove the symlinks from the unpacked tree before importing
into CVS (cvs-upgrade does that automatically with option -F), and
change debian/rules to not run restore-symlinks -rm in the clean target.

Therefore I would like to do this once we import tetex-3.0 (release or
beta) into the CVS.

With tetex-base, the situation is a little more complicated. There are
no more symlinks in it, so this is not an issue. However, the upstream
tarball is meant to be unpacked within a texmf/ directory, and contains
the context, doc, fonts, tex, web2c, ... directories in the toplevel of
the tar.gz. As I learned now, this is *not* a problem for dpkg-source
(for details, see the patch I submitted to the bug above). Technically,
it is also not a problem for our CVS archive: For versions of 3.0 (or
2.99.x) and higher, we would have no texmf directory, but the context,
doc, fonts, tex, web2c, ... directories on the same level as the debian
subdirectory, while checkouts of 2.0.2{,a,b,c} would still have only
debian and texmf.

There is, however, the problem that we loose our CVS history of the
files in the upstream tarball. For example, the Changelog file will
appear in CVSROOT/tetex-base/ChangeLog as if it was a new file, with no
connection with its old incarnation in
CVSROOT/tetex-base/texmf/ChangeLog - and the same for every file.

I think, however, that this is not really a problem, because we do not
change those files, only copy them to the right places (of course we'd
have to change some patches, but this is below debian/). The history of
those files is just a history of "imported upstream version N.n.nn",
nothing more. The only thing a newcomer would have to learn is that
upstream versions prior to 3.0(-betas) are in the Attic, in
CVSROOT/tetex-base/texmf. 

What do you think?

On a related note, I'm wondering whether I should check in my now
working and installing packages of 2.99.3 on an experimental branch of
the CVS. This would make it easier to communicate about beta development
- maybe this is something especially Hilmar could value. It would also
give me a more regular backup for my work... On the other hand, it would
mean much more commit messages to this list. At least I find myself
commiting much more often to my local CVS, and much less tested changes,
than I try to do with our centralised CVS - and this would nearly
completely be moved to our centralised CVS. I could tag each commit
message with [experimental], of course, so that everybody who's not
interested can easily filter that.

But I think we should upload teTeX-3.0 into experimental rather early
than too late, anyway, so that people can try compiling it on other
architectures, test libkpathsea4, etc.

Regards, Frank
-- 
Frank Küster
Inst. f. Biochemie der Univ. Zürich
Debian Developer



Reply to: