That's sensible to me. Was just an unexpected side effect of improving the arc land workflow for git and thought I'd mention it.

On this issue, it would be great to tune the behavior regarding "Author date" vs "Commit date" for git. With our current CI tool, current build and previous build are determined in "Commit date" order not "Author Date', which can result in confusing links.

👍

I think I run this Test Plan at least daily, minus the unit tests

just...so awful

I missed this little bit when moving to local disk file storage, which has write permissions for the user running the web server (www-data) and then the GarbageCollectorDaemon emitted:

unlink(/srv/phabricator_files/69/aa/178c834ec7e0c1c9badd11389387)

errors, even though I had read this page, and found and set the phd.user setting to www-data on my own.

Awesome. Here's trace output without --dry-run https://secure.phabricator.com/P1246

looks like this will help, test ran with ./bin/files compact --all --dry-run --trace

Can confirm D10333 fixes for me

Also seeing this on a different, private repo. It is on windows.

Wow, nice

In D9375#45, @talshiri wrote:

Awesome! Can you arc patch D9375 locally and see if:

1. rendering looks reasonable

I can test Outlook (Windows) rendering

This sometimes works for me. It seems like sometimes commits get hung importing so revisions are never closed for those commits. Especially seems to be the case with encoding issues.

just to followup: differential.generated-paths works great.

I'm in a similar scenario, t4 templates for codegen in visual studio often have <auto-generated /> somewhere in them. It'd be nice to have this string configurable so we don't need to have comments like this everywhere:

// <auto-generated />
// @generated

Tested and works for me on the changeset that was displaying this issue. It did prompt to mark 2 of the files in the diff as binary, but allowed the revision to be posted.