Bug#613662: unmet dependencies race condition
Le Thursday 17 February 2011 02:04:27 firstname.lastname@example.org, vous avez écrit :
> >>>>> "D'R" == Didier 'OdyX' Raboud <email@example.com> writes:
> D'R> Such temporary glitches are perfectly normal (and transitional) in the
> unstable D'R> distribution. All those packages are built and uploaded,
> they will reach your D'R> mirror really soon now. Hence closing this
> Can't they be sent in a different order so as not to expose the problem?
This is not feasible, because of the nature of unstable. Packages get uploaded
and get auto-built by build daemons after the upload. As building packages can
take much time, it is frequent to have architectures that have many more
packages built than others.
On those 'slow' architectures, unsatisfied dependencies are frequent, but almost
always get solved by _waiting_ (I run 'aptitude update; aptitude safe-upgrade'
daily and I don't even see these issues). Hence you cannot expect developers to
wait for _all_ architectures to be up2date to upload packages that depend on a
given package, that would just slow up things for no benefit.
If you want a system that has no such things, you should try testing (currently
Wheezy) or even stable (currently Squeeze), whose migration scripts are
_designed_ to avoid this. Unsatisfied dependencies in the testing or sable
suites are bugs; in unstable they are not (and cannot be, for the reasons
Didier Raboud, proud Debian Developer.