Quite possibly we should just denormalize profileImageURI and calendarStatus + calendarStatusTTL onto PhabricatorUser, since we almost always need those anyway and mixing real object queries with a read-through cache isn't a great v0 feature.

FWIW, I was pretty skeptical of the icons before we had them but I kind of like them now. :o

Yeah, this is really the only part that feels bad to me:

I prefer retaining titles for threads, a lot of mine are pretty meaningful. They're also meaningful when things convert up to email.

That makes sense to me. I think D12215 is generally a good smooth intro to letting users do everything without overcommitting to tons of JS.

(At the same time, I hesitate a bit because having the icons up there be totally different for every comment is also weird.)

At first actual-real-use, this state feels a little weird to me (the "Not Done", specifically). Maybe we should just hide the checkbox element on drafts, now that we aren't letting you pre-check your own comments?

Note that you always receive a notification. Going to Notifications will always show you everything.

In any case, we're very unlikely to ever put time delays on mail.

See also T7004.

We are doing a policy check here, but it's an unusual check and it's not fully effective.

The expectation is that the rule will fail to run when the user can't see the object. The code is consistent with this expectation, looking into it..

It looks like this is probably not a bug then, and D12211 is a fairly reasonable fix.

I narrowed it down a bit more, although this is in Safari and the behavior is generally super inconsistent/crazy across browsers: I see scrollbars in Safari and Firefox for your reduced case, but not Chrome. I see scrollbars in Safari and Chrome for my reduced case, but not Firefox (all on OSX).

See also T7472.

T6670

haha

Here's that screenshot with more context so it's less confusing given my super hip cool localization:

This was the old ligature block:

I'll try to do this and cross-link the Bugzilla ticket here if that's it. Sounds good?

In particular, the .differential-changeset is the thing picking up the scrollbar -- not the .changeset-view-content -- but adding overflow-y: hidden to either element removes the scrollbar.

Er, sorry, I can reproduce it.

I can reproduce this on Firefox 36 on Windows 7.

Yeah, I'm pretty sure I've only ever used this by accident.

I can reproduce the report only in Firefox; Chrome and Safari work fine for me.

Here's how I do it in Safari:

I can't reproduce this in Safari 8.0.4 on OSX.

Objects which are attributed to that user could be re-attributed to a dummy

I assume that was handled by the password envelope

I assume my patch does not change current behaviour for others

I filed bin/remove support in T7694.

I haven't reproduced this myself, although I saw one other similar report so I suspect it will reproduce easily.

We won't be adding this to the UI. Like with most other objects and applications, it is intentional that objects can not be permanently destroyed from the web UI. Among other things, this severely limits the damage a malicious or compromised user can cause.

It is expected that this block will handle the case you describe, note the comment about rebinding with anonymous credentials and the actual rebind lower down.

Here's my v0 plan for this:

The draft state is now communicated very clearly. Presuming this is resolved at HEAD until we see more reports.