1303556 - ffxbld pushes to central and aurora and beta fail gecko-decision with "HTTPError: 400 Client Error: Bad Request for url: http://taskcluster/queue/v1/task/..."

Status: RESOLVED FIXED

(Taskcluster :: General, defect)

Description

[Phil Ringnalda (:philor)]

ffxbld isn't quite unique (my vague memory is that it came up in some bug recently that there are four such Mercurial accounts), but quite rare in that his Mercurial account is exactly that, "ffxbld", not an email address.

He does two sorts of pushes, pushing release tags to mozilla-esr45, mozilla-release, and mozilla-beta, and pushing updates to the blocklist, HSTS and HPKP lists to mozilla-esr45, mozilla-aurora and mozilla-central.

gecko-decision on his pushes to esr45/release/beta still works, but somewhere between July 30th and September 3rd on aurora (https://treeherder.mozilla.org/#/jobs?repo=mozilla-aurora&author=ffxbld&filter-searchStr=8b92d81cffc3f97b03bcb1cb01220b58b731c962&fromchange=fb90485f11a590834aaa80481d61f8a4c11368e2) and between June 4th and August 27th on central (https://treeherder.mozilla.org/#/jobs?repo=mozilla-central&author=ffxbld&filter-searchStr=6751f6b4d53bef7733d3063aa3f72b0832dbde74&fromchange=3e8ee3599a67edd971770af4982ad4b0fe77f073) they started failing, "HTTPError: 400 Client Error: Bad Request for url: http://taskcluster/queue/v1/task/..."

On aurora, that failure apparently comes after dependencies are triggered, so we still run builds and tests on the push, but on central, we don't, so we have a security-related automated push carefully timed to happen when nobody will be around to watch it, pushed directly to mozilla-central, which doesn't run multiple tier-1 builds and their tests.

Comment 1

[Ryan VanderMeulen [:RyanVM]]

This seems like a pretty critical issue to me.

Comment 2

[Phil Ringnalda (:philor)]

Apparently my guess that it was his lack of email address was wrong (or, that's been broken in-tree after Gecko 49), since seabld's pushes to mozilla-release don't burn the gecko-decision task.

Comment 3

[Dustin J. Mitchell [:dustin] (he/him)]

[task 2016-10-01T12:55:42.304659Z] * data.metadata.owner should match format "email" 

so, your guess is correct.

I'm assuming it's hard to change that user to have the same form as EVERY OTHER USERNAME.  Jonas, how would you feel about relaxing the format requirement for the owner?  The other option is to check the format of `owner` in the decision task and substitute [e-mail] if it doesn't match.

Comment 4

[Jonas Finnemann Jensen (:jonasfj)]

@dustin,
We should just create a dummy email for bots like ffxbld
(I assume ffxbld is a bot).

For github we already do this, and github advices a pattern like:
  <username>@users.noreply.github.com

I'm sure you can come up with a different email address pattern for bots.
We loose a lot of value if we just allow random strings as owner.

Comment 5

[Dustin J. Mitchell [:dustin] (he/him)]

Attachment: Bug 1303556: ensure owner is always an email address;

Review commit: https://reviewboard.mozilla.org/r/82896/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/82896/

Comment 6

[Jonas Finnemann Jensen (:jonasfj)]

Comment on attachment 8797296 [details]
Bug 1303556: ensure owner is always an email address;

https://reviewboard.mozilla.org/r/82896/#review81784

Looks like a decent solution to me.

Comment 7

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/d8a271c72525
ensure owner is always an email address; r=jonasfj

Comment 8

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/d8a271c72525

Comment 9

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/14d67f59ae64

Comment 10

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/releases/mozilla-beta/rev/9f64850761cb6c55b53b379bfd125140e00033e9