Build-dependencies of backports, and searching the list archive is broken

[Frank Küster <frank@debian.org>]

Hi,

I'm sure we already discussed this question, or at least parts of it,
but unfortunately searching does not work on
http://lists.backports.org/lurker/list/backports-users.html.  It
redirects to 

http://lists.backports.org/lurker/search/20380101.000000.00000000@ml:backports%2Dusers,policy.en.html

and then complains 

,----
| Lurker - failed to render page:
| Bad request (20380101.000000.00000000%40ml:backports-users,policy.en.html):
| 
| The given parameter was not of the correct format. A searc request
| must be formatted like:
| search/YYYYMMDD.HHMMSS.hashcode@word,word,word.xml
`----

So, not to the actual question: How can I best ensure that a package
gets as little dependencies on libraries on backports.org as possible,
when at the same time it build-depends on some packages that are not in
sarge, but on bpo?  For example, for tetex-bin I need libpoppler and
debhelper, but want to avoid dependencies on bpo's XOrg.  auctex needs
texinfo from bpo, nothing else.  And so on.

I'm using pbuilder, and I really do *not* want to log into a sarge
pbuilder chroot and install the needed packages from bpo manually before
building the package.

So what should I do?  One option would be to keep a couple of tar.gz
files for chroots that have just exactly the needed packages pinned to a
bpo version.  But this requires a lot of disk space, seems a little like
brute force, and is also error-prone because I can't always be sure that
pbuilder update will actually update the package from bpo, because the
update might require installing an additional backport.

An alternative would be versioned Build-Conflicts against the relevant
xorg packages, but that doesn't look nice, too.

Is there a better alternative?

Regards, Frank
-- 
Frank Küster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Zürich
Debian Developer (teTeX)