(Wow. I've never clicked on that button in my life, also when working)

By the way I cannot reproduce in Apache 2.4.61 🤔 Debian package 2.4.61-1~deb12u1

Thanks. What happens in older Apache2 versions?

Thanks! Looking at PhabricatorSubscriptionsMuteController that works with PhabricatorTransactions::TYPE_EDGE, this seems good to me.

Setting as "new validations probably need to be managed". Thanks again

Thanks :) Interesting weird transaction.

This happens because if ($continue_on_missing && $error->getIsMissingFieldError()) in https://we.phorge.it/source/phorge/browse/master/src/applications/transactions/editor/PhabricatorApplicationTransactionEditor.php$1248 is true.

Five months later it seems that we survived. Anything left to do?

We might come from different user journeys.
Your users might be aware what the project means, and that the project (and its workboard) is archived, and they do have an initial reason to look at this archived workboard (e.g. default view is not "open tasks" but "all tasks")?
My users clicked an external link to some project tag in something called Phorge, and that project has been archived in the meantime, and that project is set to show the workboard by default, and that is now just empty columns (as the default view of workboards is "open tasks"), and that is confusing and unhelpful.

Skip assigning variable; direct return

Just that oneline

Would anyone be willing to give this another review? :)

Use $caught, not $ex.

• Use $caught, not $ex.
• Include Herald rule monogram in log message for easier finding.

(Hmm should I replace $ex->getMessage() with $caught->getMessage() or not?)

Yep thanks

In D25736#20139, @aklapper wrote:

(Hmm should I replace $ex->getMessage() with $caught->getMessage() or not?)

(Hmm should I replace $ex->getMessage() with $caught->getMessage() or not?)

Correct, the end user will not realize without opening the network tab of their web browser's developer tools.
However, as a downstream admin, I look at statistics/dashboards of HTTP 4xx and HTTP 5xx errors. When the numbers are not very low I'd usually assume something goes wrong and should be investigated. But that is not the case when Phorge throws a "silent" HTTP 400 for no obvious reasons.

Ahh but Transcripts do render an exception when it fails at least. I get a HeraldInvalidConditionException printing Invalid Condition: Condition references a rule which does not exist!, triggered in HeraldAdapter::doesConditionMatch() (only if you reach that condition in the evaluation of course, because https://we.phorge.it/source/phorge/browse/master/src/applications/herald/engine/HeraldEngine.php$437-445 ).
That's called in HeraldEngine in the line $is_match = $adapter->doesConditionMatch (note that class also has a method with the very same name).

So this is "just" something visible from server logs or browser console, right? It is not an error ever visible to user interface.

(Trying to make this task a bit more about the root problem, and less about the proposed solution)

I mean. See this archived Project (archived Milestone), that is linked from our changelogs:

For example, some people then may ask "why my default view was changed?" and they will start manually "rollbacking" this to their desired view (e.g. showing closed tasks, from the Workboard, with a filter "Show all tasks").

Premising that I like the implementation but I think it's not the desired solution.

Additional consideration under control: an external class may have extended AphrontTypeaheadTemplateView and defining a renderToken() there, and making this snippet operative (I'm not saying that it's a good practice, I'm saying that it would work). But this extreme corner case is not the case.

I personally don't see value in sharing a stacktrace in this case. YMMV :)

Seems super nice. Minor comment. If we do this, the stack trace is NOT shared:

Strip block

Uhmm, transcripts at http://phorge.localhost/herald/transcript/33065/ never ever render a Herald rule condition (disabled or not) within another rule, it seems

I'd be worried about DB performance running such a query as our DB has dozens of millions of files... (This is basically SELECT COUNT(*) FROM file_attachment fa WHERE fa.filePHID NOT IN (SELECT f.phid FROM file); in my understanding?)

Hmm I guess now my patch title is too narrow as it logs any Herald exceptions? :D

Per Valerio's previous comments

Marking as "probably something different should be done"

In short, requireFieldImplementation seems correct in their legacy version, and something is not ideal in the consumer, that has a try catch. That try-catch should be edited.

Note that every unmanaged exception is logged. So if you don't see a log, something is already managing this.

Unrelated but @aklapper I'm curious about the situation in WMF. If you can maybe report something like this:

I think the behavior of not setting the timezone is probably correct, actually. GetFormattedDateFromParts below already constructs a date object again with the timezone set. So I think the timezone offset is being applied twice here.

Still happens after backing out rP2295bcda14e71948516752f8fbada6601b9f0bde thus not a recent regression.

I guess I just do not enjoy a good number of HTTP 400 errors in our error logs when literally nothing went wrong at all.