@lpriestley hit this earlier too.

Story is pretty much the same a year later, so I'm just going to update the "today" in the description. :3

• See T5873 for ApplicationTransactions in Conduit (adding members).
• See T3980 for binding to external sources.

Reopen if this crops up again.

wut

Okay, I think we should do this:

Particularly, although it's conceptually simpler, paging/ordering is probably even messier. Paging/ordering is simpler if we have a single total order relation across all events, even if the order isn't exactly correct in some timezones.

An alternative is to only store date information for these events, then do two separate queries for time ranges:

I think this is:

I think the best way to do this is to create a new class to represent the state of a date control, like AphrontFormDateControlValue or something. Currently, AphrontFormDateControl has most of this logic, but that's awkward since it serves as both a Control and a value. This class will have fields like:

From a technical point of view, I think this looks like:

• Write a subclass of HeraldAdapter. I think PhrictionDocumentHeraldAdapter is a fairly reasonable example.
• In HeraldTestConsoleController, you have to add a hacky piece of code to help Herald figure out which adapter to use.
• In your Editor, implement shouldApplyHeraldRules(), buildHeraldAdapter() and didApplyHeraldRules().
• You should now be able to use the test console to test rules.

For outbound mail:

I'm just going to merge this into T1562, I don't expect to really touch this page before that obsoletes it.

Like "meow meow meoww".

She sounded really hungry earlier.

Does clicking "Explain Why" explain why?

At HEAD, the messages are folded individually.

The comments persist even when the code that you're talking about is deleted.

You can vote for multiple options, and @cburroughs has.

As a reviewer, I think bringing "Done" comments forward is quite important: it makes it a lot easier for me to verify that they're really done. This is the #1 benefit I've reaped from this feature so far.

• T7999 suggested something along the lines of the "minimal UI".
• Via IRC, "(btw, the 'moving comments over' feature has been a big hit here.)"
• Via Twitter, our heroism was recognized.

(Including discussion of the possibility of summarizing, hiding, etc., older inlines.)

I'm going to merge this into T7447, which has some ongoing discussion about this feature.

We don't currently plan to implement this. I'm going to merge this into T4207, which discusses one of the major technical issues (subtasks are a graph, not a tree).

One reason to show old inlines even when they are marked "Done" is that the reviewer who added the inline may want to verify that it is actually done when they look at the updated diff.

• Implement supportsSearch() in your Editor and return true;.
• Implement a SearchIndexer, like ManiphestSearchIndexer.
• You can test the indexer for non-crashiness easily with bin/search index E123.
• To test with the actual global search in the UI, you need to:
  • be running the daemons; and
  • restart the daemons after you make any changes to your SearchIndexer.

I hear CSS4 has white-space: nowrap-first-space-but-wrap-the-rest-normally;.

• Add viewPolicy and editPolicy fields to the object.
• Add these columns in .sql files.

Particularly after T3670, these might get a lot longer than they are right now in reasonable cases.

Particularly after T3670, these might get a lot longer than they are right now in reasonable cases.

They could, yeah. We'd probably need a "List all users with badge X" UI to avoid losing functionality, but that seems pretty reasonable.

Yep. I personally intend to create hundreds of badges to award myself once this is built.

Here's a rough sketch of this. This is a ton of work and should be split into like 300 revisions. See T6526#107028 for product details.

If browsers are carrying state across reloads, maybe we just need to initialize JX.Quicksand.id to window.history.state || 0 (and throw away that object patch).

RTL doesn't flip scrollbars for me on that fiddle:

DarkConsole shares the same scrollbar with the rest of the page, but the chat column has its own scrollbar, and needs to scroll independently of the main page content.

Disabling the pretty scrollbar alone wouldn't fix this without also removing the container frame which enables the chat column to exist.

We can possibly fake this with Quickling in the common case once we summon the courage to enable it without the column open.