Broader work required than extension/language map update, see T3626.

A better long-term solution than a temporary fix is wanted, per D17104#203911

Stale localisation issue

Should this hook into any of the HUGE/TRULY_CRAZY_HUGE diff detection stuff that got fixed a few months ago instead of defining a new constant here?

One more of these just came in for repo003.

I never remember the name of this ticket, so here are the things I think it's called: phone home mothership

This currently appears to be mooted by T13188.

I think you can cherry-pick those changes safely. I suspect further changes connected to T13211 and/or T10884 in the near future are likely to be difficult to cherry-pick, though, so ideally you should try to upgrade.

To pull the list down (this takes a while since we can only grab 100 results at a time and the API takes a while to respond):

See PHI962, where a GSuite-backed install has seen multiple "550 No Such User" bounces to the same address even though the address exists. It's currently unclear if this is an issue in Mailgun, GSuite, the series of tubes between the two, or elsewhere.

@epriestley : our current phabricator version is c46be2a70b4db72d61e76690ef095384de2c3f91 that was landed on 04/13, will there be any issues if we directly cherry pick the recent two versions without pulling any other changes?
rP51073b972ece: Try to route cluster writes to nodes which won't need to synchronize first.Wed, Oct 17, 3:08 PM
rPbc6c8c0e93a7: Explicitly shuffle nodes before selecting one for cluster sync.

I made another large push, then a series of small pushes. The big push hit secure001. The small pushes then hit:

Big push is:

To "sort of" test this in a real environment, I'm going to do this:

This is slightly more involved than I initially believed, we have a "Default Edit Policy" and "Default View Policy" which should probably be nuked now and PHI933 also has a sub-request for an "Author" object policy.

One idea here: T12961 and T13012 are security issues related to the misuse of -- in argument lists. I think there was one more (less severe issue?) with Git elsewhere, too.

This is entirely reasonable, the only reason it doesn't already exist is that the lease owner may be any of several different types of objects (I think: user, repository, another Drydock object, maybe null?) and we don't have a convenient UI datasource for letting users type to select any one of these various object types.

We can make shuffle() predictable, but should probably implement our own state-maintaining PRNG to do so. If we use srand() / mt_srand() without maintaining our own state:

https://discourse.phabricator-community.org/t/latest-arcanist-incorrecly-parses-comments/2013 reports an issue with this new behavior - probably if the Subscribers is left empty, the parser takes the first line of the comment.

¯\_(ツ)_/¯