[U-Boot] [STATUS] "Quality" of patches / testing.

[Jason]

Andreas,

On Tue, Oct 18, 2011 at 03:36:21PM +0200, Andreas Bießmann wrote:
> Dear Jason,
> 
> Am 18.10.2011 15:05, schrieb Jason:
...
> > I've been mulling this over, and here's my approach:
> > 
> > 1.) modify git to add a hook in format-patch, output is dumped after the
> > 	 '---' and before the diff.  We use this the run checkpatch.pl with our
> > 	 config.  Our script includes a version tag of checkpatch.pl?
> 
> I think of some script that utilizes git rather than changing git.

As long as the end effect puts all of the versions of a patch series into one
thread, and injects the output of checkpatch, a changelog and possibly the
output of MAKEALL; then I'm all for it.

I'd prefer to create a generic solution so that other projects could take
advantage of it and adapt it to their needs.

> > 2.) Add a '--versioning' option to format-patch which will scan the
> > 	 output directory for previous versions of the patch.
> 
> nice, but ...
> 
> > 	a.) Use the Message-Id of the first version as an In-Reply-To
> > 	b.) Migrate patch changelog over from previous version, append '***
> > 		 ADD CHANGELOG ENTRY HERE ***'
> > 	c.) enforce [PATCH 1/7 V#] Subject line format.
> 
> I never hold old patches long, they are on the list, patchwork or in my
> tree.

Hmmm, if I can get a better handle on rfc 2822 [1], particularly section 3.6.4,
then the threading would be maintained even when the old messages are deleted
from your mail queue.

> But feel free to discuss this on git ML. And I doubt this will help us in
> near future (remember the debian users ;)

Possibly, there's no reason why the wrapper script and the git hook can't both
be pursued.  Looks like Simon Glass already has a good start on the wrapper
script.

thx,

Jason.

[1] http://www.faqs.org/rfcs/rfc2822.html