I just got another internal support request about this, so I figured I should probably raise a feature request here. I semi-frequently get internal support requests within my company asking why builds on code reviews have failed. We put a nice big warning on them:
We wrote this warning when people used to have outdated versions of Arcanist that didn't support pushing to staging repositories at all.
The issue now is often that users don't have the SSH agent configured correctly, because they clone over HTTPS (and the staging repository is SSH). As per Q264, I'm looking for a way to make "arc diff" fail if pushing to the staging repository fails. This is because if pushing to the staging repository fails, then the build for the diff will always fail because it can't get a working copy. Having to track down why the build is failing at this step costs more time than if "arc diff" just failed to push to the staging repository in the first place.
The other outcome of this is that developers get used to seeing the build fail on their diffs and ignore it (because none of the "arc diff"s they do result in a working build). I don't get any visibility over when developers are doing this, but it has the consequence that integration tests won't be running for those developers when they're submitting code.