Wayland, Weston and libinput to patchwork?

[Pekka Paalanen]

On Sun, 5 Oct 2014 22:55:36 +1000
Peter Hutterer <peter.hutterer at who-t.net> wrote:

> On Sat, Oct 04, 2014 at 04:58:38PM +0200, Jonas Ådahl wrote:
> > On Wed, Oct 01, 2014 at 11:45:26AM +0300, Pekka Paalanen wrote:
> > > Hi,
> > > 
> > > at least with Wayland and Weston, we have bit of hard time tracking the
> > > patches that need attention. I think I am currently the only one who
> > > actually keeps a backlog, my backlog is not public, and I cannot be a
> > > maintainer 100% all the time, so this won't scale or work for too long.
> > > 
> > > How about we started using http://patchwork.freedesktop.org/ ?
> > > 
> > > Apparently it works fine for Mesa et al., right?
> > > 
> > > I believe it picks up patch emails from the mailing lists automatically
> > > and creates issues, and with a git hook at fd.o repos, a git-push can
> > > automatically close issues.
> > > 
> > > There was also some command line tool for the patchwork database, IIRC.
> > > 
> > > It wouldn't change how we work: patches are good in the mailing list,
> > > inline, we would still do review on the mailing list, etc. We would
> > > just have an automatically maintained list of open patches.
> > > 
> > > For the record, this was the announcement for Mesa:
> > > http://lists.freedesktop.org/archives/mesa-dev/2013-November/049293.html
> > > 
> > > Wayland-devel mailing list gets patches to three different projects:
> > > Wayland, Weston, and libinput. Is this a problem? Sometimes it is hard
> > > for even humans to see which repository a patch is targeting.
> > > 
> > > From a quick chat with tfheen, it seems like patchwork assumes 1:1
> > > between mailing lists and projects. OTOH, it looks like Xorg project in
> > > patchwork gets patches to a myriad of different git repos, and you can
> > > filter search results based on subject.
> > > 
> > > People do already usually use something like "[PATCH weston v7]" to
> > > identify the target, so filtering by subject should mostly work.
> > > 
> > > What do you think?
> > > 
> > > Would libinput want to be in patchwork?
> > 
> > The patch traffic for libinput is fairly low compared to wayland/weston
> > so I believe there is no urgent need, but still if patchwork can help
> > keeping track, maybe we can still try it out. The issue I see is that it
> > would not be as automatic as one could hope if the one who pushes don't
> > want to resubmit the patches on the ML after for example correcting
> > minor review issues. Have I understood it correctly that this would cause
> > patchwork not to close the issue? If that is the case, I think patchwork
> > would cause more burden than benefit right now.
> 
> yeah, that's pretty much what happens. if the patch isn't identical to the
> one submitted on the list, you need manual intervention, and the same
> applies for any changes in patch sets (you'd need to mark the old one as
> obsolete). that results in patchwork quickly becoming a dumping ground if
> no-one steps up to do that janitor work.

Indeed, patchwork does not seem to be as automatic as we'd wish, but at
least it is a public backlog for Wayland and Weston. When a reviewer
finds an open timeslot, he can just look at patchwork to see what's
waiting, rather than searching for unanswered emails on the list.

Daniel has already set up Patchwork as a test, and currently it is
filtering out libinput patches based on the subject line prefix in []
somehow. So libinput patches should not be appearing in Patchwork at
all for now.


Thanks,
pq