Backport is rejected if an older backport is already there

I think this check is wrong anyway. If we want to supersede a version in backports with a version in another pocket, then we should be able to do so. Certainly I agree with John that an upload to backports should be able to supersede an older upload to backports, but I'd go further and say that this check should be removed entirely as it's wrong.

Does anyone on the Soyuz team know why this check was added?

Note further that changes to released distributions are subject to manual approval anyway, and backports are only relevant to released distributions. So forbidding updates from superseding backports because it might have been accidental or something doesn't really make a lot of sense to me.

Celso,

could you take a look at this?

Colin, see the bug #34089 for reference to what we do for BACKPORTS.

Anyway it's wrong because it's not allowing us to superseded version in BACKPORTS pocket itself, this fix wil be easy, but as you've pointed it might be entirely wrong.

What do you think ?

promoting to critical since it's clearly a blocker to dapper LTS

I can see Martin's argument, but it misses the possibility that a package might stop being backportable from edgy to dapper at some point (either because the package changes to make backporting impossible, or because backporting is no longer desirable for some reason). In this case, the version in -backports would no longer be only minimally smaller than unstable, and it might make sense for a newer version to be uploaded to -updates.

It is also conceivable (perhaps not in Ubuntu, but elsewhere - it's a matter of distribution policy) that a distribution might decide that after a version has been tried out in -backports for a while it can be uploaded more "normally" to -updates.

However, I don't feel strongly about this, and we have no immediate use case where we want to supersede -backports with -updates. Allowing -backports to supersede -backports itself is clearly more urgent.

The trivial fix is implemented in my `small-fixes` it allow new versions in -backports. Sorry for the inconvenient, I'll let you know when the code gets in production.

The central issue needs discussion, I guess. Colin, can you get a consensual position from distro-team ?

According to kiko, this is fixed and awaiting rollout, so marking Fix Committed.

Well, in fact, it was landed in RF 4066.

Released then

no, its not released yet.
Soyuz codeline is a RF 3909 plus several related cherrypicks until RF 4052 and the fix for bug #58187 applied locally .

partially released in Soyuz codeline, only relevant bits of RF 4066 (r=kiko)