Joerg Jaspert writes ("Re: Include git commit id and git tree id in *.changes
files when uploading?"):
> There are multiple places where stuff can be rejected / "get lost".
> Anything that happens in queued can NOT (easily) appear at api, as
> queued has no way to write into the database.
Thanks for the info.
> If someone wants to tackle things and get more status out of the various
> steps:
>
> - queued is for anything *before* it actually reaches the archive. So
> upload queues and their various steps.
Is the time a package can spend in there bounded? If so then it would
be possible to assume that things that were uploaded, but haven't
appeared in the later stages, got lost or overwritten somehow.
(Quite what to do about that is another question. AFAICT with
multiple rapidly successive uploads with increasing version numbers,
only one of them will get past queued? The DELAYED queues are a
wrinkle here too, I think.)
> - Afterwards its in dak, process-upload in the unchecked cronjob.
>
> The latter can (and does) modify projectb, so adding more status info at
> that point is easy. Queued case needs more work and thought.
Are REJECTed uploads recorded in projectb then? (Or could they be?)
If they are then ...
> MR, as always, welcome. queued is perl, process-upload is python.
... one possible MR is one to expose that in the HTTPS API, I think.
Ian.
--
Ian Jackson <[email protected]> These opinions are my own.
Pronouns: they/he. If I emailed you from @fyvzl.net or @evade.org.uk,
that is a private address which bypasses my fierce spamfilter.
