1127549 - [pushes] pushes.utils._hg_repository_sync should be more resilient to 400 response for too many heads

Status: RESOLVED INCOMPLETE

(Webtools Graveyard :: Elmo, defect)

Description

[Axel Hecht [:Pike]]

In _hg_repository_sync, we're making sure our local clone has all changesets in the push data we're handling.

We're currently just kicking of a plain old hg pull, but that's getting 400 responses if there are too many heads, see bug 927219.

It'd be nice if we could just pull a particular revision, like the last one. That works for the most part, unless there's a single push on multiple branches. gps did one just the other day, so despite the fact that that's rare, it happens.

I wonder if the cleanest way is to iterate backwards through the revisions in the push, and if they don't resolve in the local clone, to pull that revision.

That should result in just one pull in the generic case, but would handle the case of multi-branch pushes, too.

Comment 1

[Gregory Szorc [:gps]]

This is arguably a bug in Mercurial. See bug 1030239.

Comment 2

[Gregory Szorc [:gps]]

This is a server configuration issue. It should get fixed soonish.

Comment 3

[Axel Hecht [:Pike]]

I had to reopen the server bug, and given how my next week looks, let's get this at least on a branch and use it in prod.

Comment 4

[Axel Hecht [:Pike]]

https://github.com/mozilla/elmo/compare/develop...bug-1127549-hg-server-error-on-pull is what I'm going to test in the automation.

Comment 5

[Axel Hecht [:Pike]]

Resolving this, we didn't hit this for a long time, and the patch doesn't apply no more, either.