[Qemu-devel] Re: Fwd: Proposal: Improving patch tracking and review using Rietveld

[Paolo Bonzini <pbonzini@redhat.com>]

On 01/27/2011 11:23 AM, Paolo Bonzini wrote:
> On 01/27/2011 11:19 AM, Stefan Hajnoczi wrote:
>> On Thu, Jan 27, 2011 at 08:55:30AM +0100, Paolo Bonzini wrote:
>>> Forwarding this from the GCC mailing list. Since patchwork isn't
>>> more than a mail archive the way it's implemented in QEMU, this may
>>> be a more interesting possibility.
>>
>> What features are you looking for beyond archiving?
>
> Well, you pretty much nailed it.
>
>> It would be nice to have a dashboard of currently unapplied patches
>> (with old versions of patch series ignored).
>
> and this:
>
>> BTW the email integration sounds good and is critical.
>
> :)
>
> The only tricky point is whether email integration includes ignoring
> older versions of patch series. I'll ask on gcc@gcc.gnu.org how/whether
> that works.

Here's the outcome of my conversation with the GCC developer who 
proposed it.

- the tool would require adoption of a special tool for patch submitter. 
  The tool is a self-contained Python script that could be included in 
the QEMU repository (upload.py).

- reviewers (and submitters discussing the issue) can either reply 
directly to email, or use the web tool.  Diego said that either way can 
be used, but the web tool is actually pretty addictive.  It produces 
basically the equivalent of the "inline comment" mails we use, and has 
decent keyboard bindings (that said, creating a comment always requires 
a double click).


That said, it looks like the integration with git is (still?) a bit too 
rough to be usable.  In particular, you can more or less track a patch 
series but not the commit messages of each series.

I created two issues in the tracker as examples:
- http://code.google.com/p/rietveld/issues/detail?id=267
- http://code.google.com/p/rietveld/issues/detail?id=268

http://code.google.com/p/rietveld/issues/detail?id=262 is also of 
interest, even though it refers to Mercurial.

It's possible that if these are fixed, the remaining problems can be 
worked around by hacking upload.py or wrapping a custom script around 
it, that would be more similar to git-send-email in appearance and behavior.

Paolo