[Debian-med-packaging] r7424 - in trunk/packages/velvet/trunk/debian: . patches

[Andreas Tille]

On Thu, Aug 11, 2011 at 03:50:59PM +0100, Tim Booth wrote:
> total 8
> drwx------ 3 tbooth tbooth 4096 2011-08-11 15:07 latest
> drwx------ 2 tbooth tbooth 4096 2011-08-11 15:08 tarballs
> lrwxrwxrwx 1 tbooth tbooth   17 2011-08-11 15:08 velvet_1.1.05.orig.tar.gz -> velvet_1.1.05.tgz
> 
> So I just get a broken link.  Running uupdate works fine but gives me the unmodified tarball.

Ahhh, hmmm.  That link should actually not be there.  You can find the
tarball (original download and repackaged) inside tarballs as it should
be.  The link is probably a leftover from uscan which simply should be
deleted (inside the script).
 
> (I never thought to look in the tarballs directory - I thought this was
> just for working copies??  Maybe it's just the symlink that needs
> fixing.)

No.  After repacking we do not have a symlink any more but a real file.
Using the tarballs directory is somehow "common practice" somewhere
mentioned in the docs.  It is finally not importand where the tarball
ends up - but at least there is no major problem with the
get-orig-source.
 
> I have to say I find your approach of "since we're modifying the
> original tarball anyway we may as well clean it up" rather confusing.
> I'd expect to find clean-ups like this in the rules file rather than in
> get-orig-source, so this is inconsistent with other packages.

Yes, it was a dirty hack.  However, when I decided for this I was *very*
confused about a build which failed (at that time) when building
something we just do not need at all.  I do not insist in this hack -
but the build system should be fixed to not even try to build
MetaVelvet.

> More to
> the point, if they remove the convenience copy of Zlib in a new version
> (which is very likely IMO) then you'll either have to continue modifying
> the source when there is no compliance reason to do so or else put the
> cleanups back into debian/rules as I have.

You cleanup results in the same binary package which is fine if *and*
*only* *if* the build of MetaVelvet will not fail.  We rather should
tweak the Makefiles in a way that the build does not even try to do so
instead of only cleaning up afterwards (as I *assume* what is the case
in your build - not verified).
 
> BTW - This should be trivial to port from CDBS to DH - I'm working on
> that just now.

That would be even better.

Kind regards

     Andreas.

-- 
http://fam-tille.de