Chris Cooper [:coop] (he/him) Reporter
	Description • 9 years ago

The periodic file update builder is defined here:

https://hg.mozilla.org/build/buildbotcustom/file/f227edf918af/misc.py#l2835

	Dustin J. Mitchell [:dustin] (he/him)
	Comment 1 • 8 years ago

Coop, what is this?  How does it fit into the migration?

	Chris Cooper [:coop] (he/him) Reporter
	Comment 2 • 8 years ago

(In reply to Dustin J. Mitchell [:dustin] from comment #1)
> Coop, what is this?  How does it fit into the migration?

It's something that currently runs in buildbot, and as such needs to migrate to TC.

Buildduty is currently working on the cadence of the updates in buildbot (bug 1093196). Once that settles, we can setup a hook to run this in TC.

There are some interesting facets to this particular job though, some of which may need novel TC configuration. The script needs to be able to update in-tree files after running some transforms on data. From that standpoint, it's going to need to be locked down much in the same way as a release builder.

	Dustin J. Mitchell [:dustin] (he/him)
	Comment 3 • 8 years ago

OK, so this is the thing that lands bumps and HSTS updates and so on.

Sounds like this will need a fleet of dedicated scriptworkers, with puppet config, and some in-tree configuration and/or a hook.

	Aki Sasaki (not active)
	Comment 4 • 8 years ago

Rather than a fleet of scriptworkers, I think we were thinking about a pool of scriptworkers with hg secrets + mercurial logic, and an upstream task that generates the patches.  That allows us to solve a number of hg problems, including version bumping at release / merge time.

	Chris Cooper [:coop] (he/him) Reporter
	Comment 5 • 7 years ago

As :philor notes in, https://bugzilla.mozilla.org/show_bug.cgi?id=1361153#c5 and dustin mentions in comment #3, there's foundational work to allow a pool of scriptworkers to push commits back up to hg. 

Dustin: does that piece need to be a separate TC bug, with this bug being a specific use case?

	Dustin J. Mitchell [:dustin] (he/him)
	Comment 6 • 7 years ago

I think that would be a releng bug, just like the signing scriptworker instances were.

	Henrik Skupin [:whimboo][⌚️UTC+1]
	Comment 7 • 7 years ago

One thing to note (I sent it via email first), and which hit us with the merge of 55 to mozilla-beta...

> Over the weekend we got informed about a couple of Tier-2 perma failures
> in our fx-ui test suite, which happen when the beta simulation takes
> place. Personally I find that kinda short notice given that the merge
> from m-c to m-b happened yesterday.
>
> To avoid having massive amount of perma/intermittent failures especially
> on the merge day shouldn't we also ensure to get the Tier-2 failures
> filed as bugs? At least a week ahead of notice for test owners would be
> great.
>
> For now I might have to mark all the affected tests as skipped and
> figure out the proper solution. If there may be issues coming up which
> affects Firefox itself, those may be present for beta 1 at least.
>
> So what do you think of extending the checks to Tier-2? At least for the
> time when we have the soft freeze of mozilla-central?

	bhearsum@mozilla.com (:bhearsum)
	Comment 8 • 6 years ago

I'm pretty sure you did this, Simon?

	Simon Fraser [:sfraser] ⌚️GMT Assignee
	Comment 9 • 6 years ago

I did!