I believe the technical reason for this is that T94548 was authored by one of the users included in the filter, so they're a "token receiver" for the story, which ends up translating into the story appearing in the feed. However, I agree that this behavior is surprising and think we should change it.

Yeah, I mean sticky. This feels a lot more useful to me if it's sticky, but I'm not sure how "advanced search" and "search all" are different if it's sticky.

Is this a modal dropdown which persists across pages, or do I type my query first and then select an action from the dropdown?

I can get this to work some of the time so it's not 100% broken, it's just really really finicky right now. We'll shore it up as the rest of the Conpherence changes move forward.

(See T7756, T7757 for linking to rooms/threads/messages.)

See T7755 for followup and discussion.

Technically speaking, I expect us to eventually go in this direction:

I haven't seen any mail from this instance get marked as spam, and haven't seen users reporting it, although who knows.

We had this about two years ago and received a lot of negative feedback about it. We don't have any plans to pursue it again.

You can use the global search in the menu bar.

Two quick notes:

Also, I suppose:

Specifically, the Gmail document about this is here:

I filed an issue against Chrome:

Harbormaster can show that a build is running. If a build is running, arc land will warn you that there's still a running build, too. There's no test-specific mechanism on top of this and we probably won't build one (Harbormaster builds have plenty of other status information, and users who want more granular information about status can click on the build).

(It does.)

This is what things should look like when you're set up:

Alright, I think this is minimal:

One of the requirements for this repro case literally appears to be "the browser window must be on the primary laptop monitor, not a secondary display".

We possibly (?) have a sort-of-reliable (?) repro (?) for this issue (?) that I'm working on reducing.

More or less. Broadly, the plan here is:

Let me finish up a couple of things and I'll try to swing by around 4ish, if that's cool?

What does the rest of your afternoon look like?

If I sat down next to you you could repro it reliably in like less than 10 minutes?

I can't repro with a trackpad, mouse, or keyboard, and can't repro in a logged-in or incognito window.

(Focusing the "Responsible Users" area also works fine locally.)

Those instructions don't repro for me, either.

Let me know how that works.

(Let me know if you see this again, I wasn't actually able to reproduce it so I didn't specifically fix it.)

Does it reproduce reliably for you?

I can't reproduce this on Chrome 41.0.2272.118 on Yosemite.

Yeah, but why send users to a page that's just going to tell them stuff is disabled?

I can't reproduce this at HEAD and rP983c6f specifically fixed this, so I'm 99% sure you're just behind that change and updating again will resolve things. That commit was made about 18 hours ago.

You'll always be a winner in your heart.

Nice, we actually hit a race!

Do you have rP983c6f9d4bcb4414b07327620a09a593f09b625a?

Minor followup in T7735.

Although tokenizer/typeahead stuff is churning a bit in connection with T4100 right now so this should probably wait for that to settle down.

I think we wrote this before typeaheads were modularized (T4420) so it wasn't easy to add a composite datasource at the time.

The page shows a prominent warning that the stuff won't work, right?

• T2549 discusses linking multiple accounts from a single provider.
• T6707 discusses the "refresh" / multi-link error.

Thanks for the report and patch!

• D12256 is the "give up" approach.
• D12258 is the "partially back out of it" approach.