On Thu, Dec 01, 2005 at 05:58:43PM +0100, Matthias Klose wrote:
> Kenneth Pronovici writes:
> > Hi!
> > 
> > > Package: pychecker
> > > Version: 0.8.16-1
> > > 
> > > some cruft is left:
> > 
> > I'm not sure I understand what the problem is.  Are these tests failing
> > for you, or something?  They seem to all pass for me in a pbuilder
> > environment.
> 
> these are the results, when run with python2.4.

Er, mind if I ask why have you modified the package to use python2.4
instead of python2.3?

> > Sorry to be so dense, but I want to make sure I'm trying to fix the
> > right thing. :)
> 
> well, even if tests fail, the test results shouldn't change the
> source.

Ok, first, just so we're on the same page: these .results files only
show up if there is a test failure.  They are generally used either for
debugging or (more often) for creating Debian-specific expected results
(test_expected/*-debian).

It looks like the problem is that the clean rule doesn't remove these
leftover files before the next build, and they're somehow ending up in
your source package.  Could that cause what you're seeing?  It doesn't
quite match up with your bug report, though.  I'd expect to see some
"new" .results files, not some changed results files -- since the Debian
source package for 0.8.16-1 doesn't contain any results files at all.
Any idea what explains this?

It seems like the best solution is to just remove the .results files in
the clean rule.  That way, even if there are leftover .results files
from failures that eventually get fixed, there's no chance that they
accidentally get placed into the source package.  Does this sound like
it solves your problem?

As an aside, if you're going to build and distribute a pychecker package
based on python2.4 (is this maybe for Ubuntu?), I'd like to see you
create new Debian-specific expected results files for these tests. That
way, all of the tests continue to pass and the regression test step can
be left unchanged relative to what's in my original debian/rules file.
At least this way, it will continue to be obvious if the behavior of the
package is changing from release to release (i.e. if new tests fail).
Would you mind doing this?

Thanks,

KEN

-- 
Kenneth J. Pronovici <[EMAIL PROTECTED]>

Attachment: signature.asc
Description: Digital signature

Reply via email to