mach update-css has confusing semantics

[perlun]

I ran the following, after changing some code that caused tests that previously passed to fail:

$ ./mach update-css /tmp/foo

Then I tried with git diff. It didn't show me anything; I assumed the command hadn't worked. It was only when @jdm told me to git show that I realized it had behind-the-scenes creates a new commit with the .ini file changes in.

This is a confusing behavior and IMHO a bad default. If we really want it to automagically commit code (which I think is questionable to begin with), it should at least not just keep silent about it. The git commit output should be shown if we retain the current behavior. (but I'd vote for just leaving the files dirty since that'd be how I'd guess it would work by just looking at it from the outside.)

Looked a bit but I couldn't immediately find where in this code this happens.

[KiChjang]

I recall seeing a similar issue before... @wafflespeanut might know more.

[wafflespeanut]

Yeah, that was update-wpt. We should do something similar to what we did for #9666 (i.e., pass --no-patch argument by default)

Code: python/servo/testing_commands.py