Open question is whether to replace loadAllWhere with loadOneWhere under the assumption that there are always either zero or one results.
That would definitely require adding null check though.

In D25807#21516, @valerio.bozzolan wrote:

P.S. try applying D25809 to check if the red unit test becomes green

Change offset to optional as requested by speck

As usual I'm confused about the PHP magic which does not require defining the method PhabricatorRepositoryCommit::getEpoch() in its class or parent classes, but I assume you tested locally that your patch works, plus PhabricatorRepositoryCommit does have some calls already to $this->getEpoch(), so I'll accept - worst case would be the very same behavior as now.

@valerio.bozzolan Thanks, this unblocks a few more things on my list. I admit that I have not looked into the unit test failure here and I assume you understand it and that it is fine to merge this patch. :)

Indeed, we seem to prefer throwing exceptions after if (!($foo instanceof PhutilSortVector)), or use assert_instances_of() on an array, using gettype() and get_class().

Ironically, adding missing variable names to @param lines increases the amount of errors in PHPStan output:

Don't remove some empty line

Note that I have NOT properly tested this locally as I still have issues with my LDAP setup.

Confirming my suspicion on PHP 8.3 (while I'm still struggling to successfully bind locally).
As far as I can tell (well, I'm running into ldap_bind(): Unable to bind to server: Can't contact LDAP server after the patch above while my local LDAP server is running but that's nothing new triggered by the patch above), the patch above seems to work.

Do what Valerio says :)

Thanks! While it does not work for all and anything (e.g. top bar search button), it is still an improvement when it comes to consistency between mouseclick and return/enter button, so I approve as I haven't seen any fallout and could not find collisions/interference with other existing functionality. (For the records, there is similar code existing in webroot/rsrc/externals/javelin/core/Event.js calling JX.Stratcom.listen() this way.)

Do what avivey analyzed (thanks!) and wrote in D25743#20738

@arp: Sure, changing underlying semantics can solve some problems (and create new ones for existing installations). :)

Ctrl-clicking on the button opens the form in a new tab

Hard to tell if it checks all and anything that could ever be tested (ICS files can be a looot of phun), on the other hand I cannot come up with any downsides of merging this either as it could always be expanded or adjusted and it passes currently for me, so I'll accept.

Fix some ambiguous URIs to Phorge commits/tasks; remove two empty lines