Re: bug packages for autobuilders

[Sam Hartman <hartmans@mit.edu>]

>>>>> "Thomas" == Thomas Bushnell, BSG <tb@becket.net> writes:

    Thomas> Sam Hartman <hartmans@debian.org> writes:
    >> This is mainly because you can't file a bug on the
    >> autobuilders.  While I agree that filing a bug on the
    >> autobuilders in this particular case is wrong, I do believe
    >> that like other parts of Debian, you should be able to file
    >> bugs on autobuilders because they sometimes do buggy things.  I
    >> care about making sure that my packages are available on as
    >> many arches as is reasonable, so I'm likely to look at failed
    >> build logs for my packages before the autobuilder maintainers
    >> get around to it.

    Thomas> I don't think that's quite right.

    Thomas> What exactly do the autobuilders do wrong?  If they are
    Thomas> unable to compile a package from source, according to
    Thomas> policy, then the package has a bug.  (If the bug is arch
    Thomas> specific, the maintainer should talk it over with the
    Thomas> porting team, but the autobuilding wouldn't be an
    Thomas> appropriate locus for bug reports.)


O, let's see.  I really wasn't sure how to change my source to fix the
problem that the m68k autobuilder couldn't build it because it
couldn't find the dsc file.  It was much easier to talk it over with
the m68k porting team and have them fix the broken mirror that had an
incomplete copy of the archive than to find source changes that would
fix a configuration problem on the autobuilder.

Another time the m68k autobuilder had a broken toolchain installed.
The standard configure test for a working gcc was failing.  I guess I
could have included a binary copy of gcc in my package source, but
again treating this as an autobuilder problem rather than a can't
build from source problem worked better for all concerned.