Re: review process

[Daniel Berlin <dberlin at dberlin dot org>]

On Monday, March 31, 2003, at 02:46  AM, Gerald Pfeifer wrote:

> [ gcc-patches -> gcc; please omit the former when replying! ]
>
> On Mon, 31 Mar 2003, Fergus Henderson wrote:
> > > > 	- define a new category of "self-approve" developers?
> > Well, the difference is that you're only allowed to approve your own
> > patches, not someone else's patches.  The intent here is to create
> > a new position which is less trusted than a normal "maintainer",
> > and whose approval powers are thus significantly limited --
> > limited to only approving their own patches, only after these patches
> > have been reviewed by another human, and only after a one-week delay.
>
> So, wouldn't it be more natural, in some way, to have maintainers that
> may approve all patches _but_ their own?  (I'm not sure how well 
> reviews
> by arbitrary third parties would work, both procedurally and related to
> quality.)
>
> Managing droped patches using Bugzilla sounds like a good idea, and in
> fact I have suggested it at least twice in the past.

I can *almost* automate detecting missed patches after a day, too, 
since patch submission is generally in one of a small number of formats.
It's actually detecting approval that is the tricky part right now.

If we could
1. Be okay with the idea that any patch not submitted in a reasonable 
format (IE changelog + patch in message, or changelog in message + 
attached file with patch), and with [PATCH] in the subject line, 
wouldn't be tracked.
2. Be okay with the idea that: patches approved by people without 
reasonable mailers (reasonable being those that can tell us whether a 
message is a reply to another message), when it's not easily detectable 
otherwise, will be assumed to not be approved.
3. Standardize the text + possible placement of approval messages.

then I'm pretty sure i could auto-track missed patches.

But it's worthless if we can't detect already-approved patches, thus 
the need to come up with a format for approval messages.

Obviously, i don't need something in some strict format, I just need to 
be able to determine that a given message containing patch was approved 
(or approved with changes, etc).  Thus, simply saying that approval 
messages should contain a line starting with "approved".
> My message at http://gcc.gnu.org/ml/gcc/2003-03/msg01529.html 
> essentially
> has all three suggestions which came up in this thread, BTW. :->
>
> Gerald
>
> --
> Gerald "Jerry"   pfeifer at dbai dot tuwien dot ac dot at   
> http://www.pfeifer.com/gerald/