[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: sbuild in squeeze



Hi Roger,

On Mon, Nov 08, 2010 at 12:10:23PM +0000, Roger Leigh wrote:
> sbuild in squeeze has been neglected for some time, and it contains
> quite a large number of bugs and deficiencies.  This was mainly due
> to the fact that I couldn't devote a huge amount of time to Debian
> during my PhD, and I only found the time to begin addressing this
> properly over the last two weeks.  The version in unstable now fixes
> 34 bugs present in squeeze:

sorry that this issue was neglected so long.  However the changes are
massive, which was one of the reasons we were overwhelmed by it.  As
this point in the freeze we cannot accept such an update anymore, too.
Given that sbuild is still moving in unstable is an indicator for us
to err on the side of caution.

> #380113, #403246, #464376, #545215, #551311, #559533, #567505, #567506,
> #567821, #567822, #569823, #570463, #570463, #571565, #571726, #571806,
> #578810, #579971, #586275, #588035, #588818, #590153, #591187, #591188,
> #591189, #591620, #595863, #595866, #595872, #595934, #598054, #599348,
> #599350, #602318.
> 
> I'm afraid this isn't just minimal bugfixing, it's basically three new
> upstream releases worth of changes.  But, I think it's important to
> have in squeeze for these key improvements:
> 
> • dpkg-source v3 support
> • integration with the schroot version in squeeze (currently it has no
>   namespace support, and the sbuild-* helper programs are broken)
> • correct removal and reinstallation of Build-Conflicts

I think the AFAICS only currently open RC bug in squeeze (#588818) is
only borderline serious, given that apt is the default resolver and it's
then only a nice to have for experimental builds.[*]  We're often
relying on people using newer tools than the ones in stable to develop
packages and the general package building tasks can be accomplished
by the then-stable sbuild.  That said, the patch in the BTS looks ok,
but further checking should ensure that there weren't any bugfixes
applied to it later.

On v3 support: that's definitely needed.  You listed two patches in
<20101115210839.GJ15585@codelibre.net>.  If that's sufficient, that
would be reasonable for a targetted fix for squeeze.

What's needed to be able to use sbuild's functionality with the stable
schroot?

> and these "nice to have" improvements:
> 
> • uses the aptitude build dependency resolver in place of the old, buggy,
>   and unmaintainable internal resolver

I disagree on this one.  I think it should not be changed at this point.

> • direct building from unpacked source trees, á la debuild et al
> • support for running lintian and piuparts

Despite being useful features they look unsuitable for a freeze to me.
Of course there are many other improvements in the current unstable
version.  But the release is not far away, so that should've happened
before/directly after freeze time, I'm sorry.

Kind regards
Philipp Kern

[*] The original issue is indeed serious for the buildd infrastructure.
    Somewhat an unfortunate situation sbuild's in.

Attachment: signature.asc
Description: Digital signature


Reply to: