relationship counselling for debian and rsync

[Martin Pool]

I heard Debian's trying to freeze soon.  I have a few points about
rsync's Debian package.

I'll see what I can do to close some Debian/rsync bugs.  Fortunately
some of the bad ones seem to be already addressed in our regular
tarballs.

It would be pretty cool if I could be automatically cc'd on Debian bug
reports about rsync, if the BTS supports that.

The debian/copyright file for rsync currently says

+COPYRIGHT
+---------
+
+Rsync was written by Andrew Tridgell and Paul Mackerras, and is
+available under the GPL.
+
+Andrew.Tridgell at anu.edu.au
+paulus at cs.anu.edu.au
+

That text is not really representative anymore, because many other
people have contributed code.  I am not quite sure what the etiquette
for changing these things is.  At the moment, I'm just saying that

  rsync was originally written by Andrew Tridgell, and has received
  improvements and contributions from a wide community of users.
  rsync is currently maintained by Martin Pool.

If that's OK with you please use this text.  I will put it in the
upstream docs sometime soon after checking with tridge.

It would be nice to have a THANKS file, but until somebody searches
the CVS and mail archives by hand we don't have the information.  I'm
trying to make sure to give credit in the NEWS in future.  In the
meantime I think it's best just to say "lots of people".

rsync is still licensed under the GPL, or to be pedantic "the GNU
GPL".  

Rather than shipping cvs.log, which as you note is not up to date, I
would suggest you just ship the NEWS and OLDNEWS files, which describe
user-visible changes.  If people want the nasty details there's always
cvs.samba.org.  At the moment the detailed CVS logs are not published.
I guess we could change this but I don't really see the need to bloat
the package.  (Personally I think it counts as bloat for both the diff
and deb.)

I don't understand the Debian test framework as well as I would like.
You've probably seen an effort to build a test suite of our own.  If
it makes sense, I would like for the Debian test stuff to just do
"make installcheck" (or "make check"), and for Debian people to
contribute patches as necessary to make the testsuite work well.  We
need to be able to test on all kinds of platforms.  If the two don't
merge straightaway it's OK, but it would be a good medium term goal.

I strongly prefer to install binaries with debug symbols to aid user
support, and this is now the default for rsync.  Is this against
Debian policy?  "make install-strip" works properly in CVS but not in
2.5.4; until then you might have to patch it yourself, perhaps by
copying from CVS.

As you know, 2.5.4 fixes the zlib bug, which potentially has security
implications.  It has been out for a week and tested by a few people.
I would be inclined to recommend that Debian ship it unless something
large and unexpected happens.

I'll file some more details against particular open bugs.

Let me know if there is anything I can do to help -- Debian rules ;-)

-- 
Martin