The next step here is to find (almost) every instance of %T in the codebase and convert it to %R. In most cases, the conversion is trivial:

I do have a real-life example where the prose diff engine rendered a suboptimal diff. I was able to fix it by changing the maximum length of the edit distance matrix from 128 to 256.

See PHI908 (et al?) which would like a maximum filesize limit for hosted repositories.

Good news: recent versions of sshd can pass the key fingerprint to the AuthorizedKeysCommand with %f: https://bugzilla.mindrot.org/show_bug.cgi?id=2081

OpenSSH 7.9p1 builds cleanly and does not segfault immediately if configured like this:

Yikes on rPe57bfbf421f4. Let me see if I can make this build.

See PHI885, which requests timeouts on git fetch during working copy construction.

Here's the DOM with the tooltip container and timestamp container colored:

A couple of narrow ideas here:

the message is sent from inside applyFinalEffects(), which is still inside the transaction

See PHI970, which identifies an issue with containerPHID not being populated correctly on a subset of buildables.

I'm going to start landing this stuff now. master will start complaining about unsafe queries all over the place (although much less frequently than it was when I first added the warning). Depending on how much complaining still exists on Friday I might make the warning developer-only, but I'm currently hopeful that I can clean up most of it before the next release promotes.

Let's try again in another couple of years, maybe the notification system will select a more appropriate start time for all-day events by then!

See PHI943. The bin/repository thaw --promote operation and/or the version bumping after a write can currently misbehave in the presence of disabled nodes with larger versions. When we increment the version, we should bump it past the largest version of any node, not just any enabled node.