[devel] synchronized subtask rebuild & disttags way too strict

[Alexey Tourbin <alexey.tourbin@gmail.com>]

On Thu, Apr 23, 2020 at 10:21 PM Vladimir D. Seleznev
<vseleznv@altlinux.org> wrote:
> > I think we already have this problem: the decision to rebuild a
> > subtask is made locally (or rather remotely), per architecture. So
> > occasionally we've seen unmet dependencies due to disttag mismatch.
> > Has the problem been addressed?
>
> Hmm, I'll think about that.

For mixed-arch packages, with both arch and noarch subpackages, you
need a two-phase subtask rebuild. There will be two interactions with
the remotes.  The first phase is, "tell me if there is no need to
rebuild". You gather the answers for all architectures / from all
remotes. If all architectures report "no need to rebuild", there will
be no phase two. Otherwise, if at least one remote wants to rebuild,
then you ask all the remotes, "rebuild irregardless".

This all could have been avoided if you disttags were not so strict.
It can be argued that repo+task.subtask is just strict enough, at
least from the end user's point of view.  (On the other hand, it can
also be argued that repo+task.subtask.iter.subiter is the only "sure
way" to prevent mixing subpackages from different rebuilds.)