[llvm-commits] [llvm] r134907 - /llvm/trunk/utils/TableGen/

[David A. Greene]

greened at obbligato.org (David A. Greene) writes:

> I'm all for code review, especially on complex patches.  That's why I
> sent this one for review ahead of time.  But the submitter needs to get
> at least some indication someone is looking at it in a timely manner.
> Otherwise the submitter is completely in the dark.

I think what might work is the following:

- Within 16 working hours after a review request is sent, the relevant
  code owner either sends a review or sends an acknowledgement that the
  request was received.  Either type of response should indicate that
  the responder is the code owner (so the sumitter knows who has final
  say).

  [I realize that in this case I did not wait 24 working hours.  That
   was a screw-up on my part and I apologize.]

- If the reviewer needs more time to do the review, he or she will
  provide an estimated time to complete the review.  Extensions to that
  timeframe should be communicated as soon as possible once they are
  known.  The idea here is to keep the submitter in the loop.

- If, after 24 hours, the submitter has not received any
  acknowledgement, he or she sends a single ping to call attention to
  the review request.

- Following the ping, the submitter should wait an additional eight
  working hours.  If there is no response forthcoming, the submitter may
  commit the patch which is then subject to post-commit review.

- A ping should almost never be necessary.

This will at least give a predictable timeframe for the submitter.  As
it is now, people who submit code are lost as to when things might move
forward.  That makes scheduling and planning very difficult.

Does this sound reasonable?

                                 -Dave