Bug#521918: pbuilder --build --binary-arch invokes 'build' target

Petr Pudlak deb at pudlak.name
Mon May 11 18:34:54 UTC 2009


On Mon, May 11, 2009 at 03:11:22PM +0200, Filippo Rusconi wrote:
> On Mon, May 11, 2009 at 02:11:18PM +0200, Julien Cristau wrote:
> > 
> > On Mon, May 11, 2009 at 13:46:30 +0200, Lionel Elie Mamane wrote:
> > 
> > > No, policy is very clear on that: if you call the "build" target, you
> > > _must_ satisfy Build-Depends-Indep and Build-Conflicts-Indep:
> > > 
> > And policy is clearly not followed by any actual practice on this point.
> > So that's as much a bug in policy as anything else (#374029).
> > 
> > Cheers,
> > Julien
> 
> Well, but then, why have new packagers trained by studying the Policy?
> 
> Look at my own situation (which must not be a rare one, I suppose):
> I've worked to make a Debian package of the software I develop [0]
> with the idea that the Debian Policy had to be implemented in the
> package making.
> 
> That software recently entered Debian through NEW and almost
> immediately after that I got a FTBFS bug report [2]: pbuilder called
> debian/rules build without installing the required
> 
> Build-Depends-Indep: texlive-latex-extra, texlive-latex-recommended, texlive-fonts-recommended
> 
> which of course failed because pdflatex was not found on the system
> and thus could not build the LaTeX docs of the software.
> 
> It is a pity to have a Debian Policy so well documented, to point
> package-making learners to that Policy and then have non-conformant
> builders.
> 
> In fact, I'd ask what would be the solution to overcome the problem
> (apart from the desirable fixing the builders)?

Hi, I'm glad I (finally) got some response to the problem!

I've had precisely the same problem as Filippo: I prepared my first package, I
spent many hours studying the Policy to follow it precisely, and to my
disappointment, I got a FTBFS bug report immediately after uploading the
package. It took me considerable time to find out what the problem was (thanks
for help!), and moreover, it wasn't apparently my fault, it was faulty
builder(s).

I don't think the problem is so much in the Policy, I think the problem is with
the builders. The builders must provide dependencies according to
debian/control and the target(s) they're calling. Of course, improving the
Policy is OK, but once the guidelines are agreed upon and written there, the
builders must follow it. Not the other way around.  I'm quite surprised that
these problems with the builders haven't been solved already, considering the
number of packages in Debian!

I suggest to create a dummy package that would be as simple as possible and
that would demonstrate the problem. Then test it with various building tools and
fill eventual bug reports. Maybe I could prepare such dummy package in the
following days, if I have time.

    Best regards,
    Petr





More information about the debian-science-maintainers mailing list