Re: Patch commit requirements

[Geoff Keating <geoffk at geoffk dot org>]

> Date: Fri, 22 Jun 2001 21:22:58 +0100 (BST)
> From: "Joseph S. Myers" <jsm28@cam.ac.uk>
> X-X-Sender:  <jsm28@kern.srcf.societies.cam.ac.uk>
> cc: Geoff Keating <geoffk@redhat.com>, "gcc@gcc.gnu.org" <gcc@gcc.gnu.org>,
>         "jh@suse.cz" <jh@suse.cz>, "amylaar@redhat.com" <amylaar@redhat.com>,
>         "law@cygnus.com" <law@cygnus.com>
> 
> On Fri, 22 Jun 2001, Mark Mitchell wrote:
> 
> > > I'm considering simply reverting such patches in the future, as soon
> > > as I get the report of the build failure.  Does anyone think this
> > > unreasonable?
> > 
> > I certainly don't.  In fact, I have a proposal before the SC that
> > deals with this issue as part of the way of doing future releases.  
> > So, it might make sense for that debate to conclude (hopefully, soon),
> > and then we can have an Official Policy.
> 
> Could this policy allow reversion of patches that add undocumented
> options, target macros, etc., as well?

That really doesn't sound like a good idea.

Mark cut out an important part of my message.  I would only revert a
patch when:

1. It causes a build failure, _AND_

2. The submittor has not said, in the message to gcc-patches, which
   platform he tested it on.

The reason being that in that case, it's quite possible the patch
never worked on any platform, and the submittor failed to test it or
a mistake was made while committing it.  

The rules for committing patches say that it is mandatory to say on
what platform the patch was tested, so people who follow the rules
will never be affected by this.

I wouldn't recommend reverting patches that just cause bugs, unless
they're so severe as to prevent anyone else from working on GCC until
the problem is fixed.  I'm not even recommending reverting patches
just because they completely break some platform, although I can see
why some people might like that idea.

-- 
- Geoffrey Keating <geoffk@geoffk.org>