[LLVMdev] code-owner sporks

[dag at cray.com]

<dag at cray.com> writes:

> Sean Silva <silvas at purdue.edu> writes:
>
>>> Really, patches get dropped *all the time* to the point where pings are
>>> a regular part of the development process.  That's a huge waste of time
>>> for everyone.
>>
>> It's only a waste of time if your workflow is entirely synchronous
>> with patch review. Most of us have a number of things that we can work
>> on, so letting a patch chill for a while on the list isn't a big deal
>> because we just go on and do something else.
>
> Of course, that's easier with git.  :)

Just to follow up, I agree that people shouldn't be stuck for days
waiting for their patch to be reviewed.  The cost is more subtle than
that.

- I have to *remember* I submitted the patch (not hard, but it is a
  cost).

- I have to save that e-mail from llvm-commits so I can refer to it when
  the inevitable ping is necessary.

- I have to wade through tons and tons of commit e-mails searching for a
  response to my patch (for some reason the mailing list software often
  breaks threading).  This is a more general problem with the current
  review process, not strictly a timeliness issue.

- I have to send a ping e-mail.

- Now I have to look for responses to *two* e-mails.

In short, e-mail is a very poor vehicle for managing this kind of
process.  I hope that the patch queue in testing helps alleviate the
poor response time problem.  Even just an acknowlegement, "hey I got
your patch but it will be a few days before I can review it", would help
a lot.

                           -David