I can't reproduce this. Here's what I did:

Those reproduction instructions do not describe a bug. When you change a file and add it, it is expected that it will be part of any diffs or commits.

Please include reproduction instructions when filing a bug report. See Contributing Bug Reports.

Ah, I was probably getting confused by that: so as I now understand, the superproject membership inherits from the subprojects it contains?

See T10003. See T9905 for broader discussion. See the Changelog to find information about changes before you perform an upgrade.

Also, adding / removing members from superproject

You can just leave a comment here.

What happens when they try?

Ah, thanks. I think it's just missing a disabled state -- I wouldn't expect the edit to actually go through when you save (or are you seeing the edit work, too?).

6081e640b154631cfc36b4d554a75638201762cd is not a commit in the upstream. Do you have local patches?

This is intentional.

There's nothing particularly actionable remaining here, although we'll always have more things to cache.

No, you can not reorganize existing projects.

I previously linked you to Support Resources. Please read that document.

This does not describe a bug with Phabricator. See Contributing Bug Reports and Support Resources.

Actually, even GitHub exposes unrecognized email addresses directly in the page source, which is the same behavior we have:

Is this an actual problem you've experienced, or purely hypothetical?

I don't think this represents a real problem. This information is inherently public: pushing commits to a public repository makes this information public and easily retrievable by automated means.

I also might take a crack at fixing "human language diffs", I think I can get something which is worlds better than what we do right now pretty quickly (even if it isn't perfect) and it would be nice to be able to be smarter all over the place about communicating description/title changes and tailoring things for typo fixes, spelling mistakes and other minor edits. I think this is something we can get groundwork in place for and then improve gradually without needing to do any big rewrites or anything.

The logic around what we send to blog subscribers is a little goofy right now, and sends them the entire post content on every update.

Thanks, fixed.

Definitely, let me get you a rough cut.

It may also make sense to fold accents when normalizing strings for ngram indexes.

I think we fixed this elsewhere, yell if it's still an issue.

I tink at least some of this was added in, e.g., D14231.

You can destroy the whole index, then reindex everything.

We don't know how to reproduce this and haven't seen other reports.

I'm not sure how to reproduce this (it didn't reproduce just now), and haven't seen any other instances of it or further reports. It's possible this was fixed earlier this year; we did some work on inbound mail around the time of this report.

(Test.)

We haven't seen requests for this from other installs and I don't particularly favor this mechanism. Prior alerts in this vein like "daemon and web config differ" have created a lot of noise and little value. T7338 should eventually provide more monitoring tools.

I don't think we actually need, this at least for now. We can look at it again later if we get more feedback about users making these sorts of errors, but it seems unlikely (I think?) that a use would type google and expect it to work.

Let's just wait for a stronger use case before building this, I think our URL construction has generally become much stronger since this idea was originally tossed around.

I don't want the error messages to just disappear since I worry that might be confusing, but I've added an option to dismiss errors once you're done with them in D14886:

Couple of thoughts, not necessarily unbeta stuff:

Resolved by D14823.