[buildd-tools-devel] Bug#559659: Bug#559659: sbuild: Targets unstable even when dealing with experimental stuff

[Cyril Brulebois]

Hi Roger.

Roger Leigh <rleigh at codelibre.net> (06/12/2009):
> Currently, the reason why your uploads default to unstable is
> because sbuild defaults to unstable unless you override it.

Yeah, I gathered that… :)

> I think the safest solution here is to simply not set it by default,
> and require the user to always specify it (or set it in their
> .sbuildrc).

Agreed.

> This is pretty much exactly the same chicken-and-egg problem as
> #529281: we don't know the package distribution until we download
> and unpack it, and we need to do this inside the chroot environment,
> but we can't know which one to use.

In my (specific, I agree but still) case, I did specify a chroot… but
that's addressed by your next points.

> - the chroot name is immaterial since one may name them
>   anything you like (and indeed use unstable chroots for
>   experimental building).  If you specify a chroot with
>   -c, we can't infer the distribution name from this.

Alright. I came to a similar conclusion after looking up what
happened. What confused me was that earlier (one year ago I think) I
noticed there were even chroot aliases (like unstable-amd64 for
sid-amd64, etc.) in schroot configuration files, so I had in mind that
there were some kind of dist ←→ chroot mapping.

> - we can, however, infer the chroot from the distribution,
>   though -c can override the default chroot name.

I still have a single chroot right now, so I have no strong feelings
right now. :)

[[[
Speaking from a developer point of view (as opposed to buildd), I
think I'm going to continue doing the following:
$ debuild -S -i -uc -us
$ # lintian gets run
$ sbuild-shell-wrapper ../package_version_source.changes

The shell wrapper would then take care of getting the distribution
from the .changes, and call sbuild on the matching .dsc
]]]

> - we could add a distribution=… or sbuild-distribution
>   parameter to schroot to allow chroot->distribution
>   inference to work robustly, but I'd rather keep sbuild-
>   specific hacks out of schroot unless unavoidable.

That's the kind of parameter I immediately checked (and why I pasted
my conf file). I guess it could be handy but I have no strong feelings
either, your call.

> If this makes sense and you are OK with it, I'll remove the
> distribution=unstable default from sbuild.

I think it would be a nice move. I guess it deserves a quick check on
the buildd side to make sure sbuild is called with the appropriate
parameter, so that we don't suddenly end up with no packages for
unstable this time. ;)

Mraw,
KiBi.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/buildd-tools-devel/attachments/20091206/7fec3ef6/attachment.pgp>