DepWaits no longer processed

I noticed this yesterday on many packages in the primary archive - even though the missing dependencies was fulfilled, builds weren't automatically retried.

As for the retrying issue, that's nothing to do with this bug. It's a bug in the scoring algorithm, such that retries are prioritised below test rebuilds. You could ask an admin to rescore them for you.

A huge amount of retry builds entered the primary archive build queue at 12:40 UTC today on all but the two slowest architectures (sparc and hppa) - was that someone manually shoving through the backlog?

There were 2 build records in inconsistent state blocking the `buildd-retry-depwait` script:

https://edge.launchpad.net/~cdekter/+archive/ppa/+build/1032562
https://edge.launchpad.net/~cdekter/+archive/ppa/+build/1032563

(source originally targeted to 'contrib' was deleted before and left 2 depwait builds behind)

First, we should setup a proper cronscript watch for it (we should have been warned if the didn't run each hour). Second we should support this scenario, it's legitimate.

Thanks for fixing that.

On Wed, 2009-05-20 at 22:33 +0000, Celso Providelo wrote:
> There were 2 build records in inconsistent state blocking the `buildd-
> retry-depwait` script:
>
> https://edge.launchpad.net/~cdekter/+archive/ppa/+build/1032562
> https://edge.launchpad.net/~cdekter/+archive/ppa/+build/1032563
>
> (source originally targeted to 'contrib' was deleted before and left 2
> depwait builds behind)

How did that happen? I see it in debian/control, but aren't PPA uploads
automatically overridden to main? The source is in main - is this some
strangeness with the magic option to follow primary archive components?
But I don't see autokey in Debian or Ubuntu.

> First, we should setup a proper cronscript watch for it (we should have
> been warned if the didn't run each hour). Second we should support this
> scenario, it's legitimate.

How is it legitimate? A build with a component not in its archive's
distribution doesn't make sense.

--
William Grant

We try to preserve the original source information as much as we can, and in this case the uploaded source was target to 'contrib', so its SourcePackageRelease was recorded as such.

The problem lives on the code that calculates the Build context component, it first checks for an active publication (pending or published) and if it couldn't be found (which was the case, because it was deleted) it falls back to the SPR.component. Basically, it shouldn't fall back, because the SPR.component doesn't necessary match the build context, by design, the corresponding publishing record component does.

We should look for the latest publishing record in the build context, regardless of its status, and for PPAs it will be 'main', for primary archive builds it will be the whatever component was established for that source (current override).

Looks like it's happening again.
https://edge.launchpad.net/~ubuntu-mozilla-daily/+archive/ppa => no retry after more than 4 hours.

Right, fixed again.

r8544 (devel)