[PATCH weston v6 00/73] Head-based output configuration API a.k.a clone mode infrastructure

[Pekka Paalanen]

On Fri, 13 Apr 2018 17:16:14 +0300
Pekka Paalanen <ppaalanen at gmail.com> wrote:

> On Thu, 12 Apr 2018 12:09:26 +0200
> Daniel Stone <daniel at fooishbar.org> wrote:
> 
> > Hi Pekka,
> > 
> > On 16 February 2018 at 15:56, Pekka Paalanen <ppaalanen at gmail.com> wrote:  
> > > here is the v6 of the shared-CRTC clone mode series. Since v5, quite
> > > many patches have been extracted from this series, sent out and merged
> > > upstream. However, now the series is bigger than ever, because here I am
> > > posting the complete series, including the full DRM-backend migration
> > > and DRM shared-CRTC clone mode implementation, thanks to having the
> > > basic atomic modesetting landed upstream.    
> > 
> > I reviewed clonemode-7-part1 from
> > https://gitlab.collabora.com/pq/weston, which is slightly revised (&
> > significantly truncated) compared to this series. Here's a braindump:
> > 
> > When adding head_list to a weston_output, we can't send
> > wp_presentation_feedback's sync_output event to the same resource for
> > multiple wl_outputs:
> > https://gitlab.collabora.com/pq/weston/commit/d0f5afff9213b81c329a98cdc7593a0101f953d5#note_7491  
> 
> Fixed.
> 
> > Should be a pretty easy fix (don't iterate through multiple heads once
> > you've sent an event to at least one head's resource), but I don't
> > know what the long-term fix is. Maybe there isn't a good one.  
> 
> I think the fix there is very much sufficient.
> 
> > I think this is also a small and easily-fixable bug:
> > https://gitlab.collabora.com/pq/weston/commit/55206c5e8453722caed18d7fc6e3d9ed0966e623#note_7496  
> 
> Unfortunately, it's complicated and not a bug. :-P
> Explanation in Gitlab.
> 
> > The only real interface change I'd like to see is using a listener
> > list rather than a single callback:
> > https://gitlab.collabora.com/pq/weston/commit/ebb74da3c89d90810f8209340da33faeea3169bd#note_7493  
> 
> Done.
> 
> > Here are a couple of throwaway comments and nitpicks; feel free to ignore them:
> > https://gitlab.collabora.com/pq/weston/commit/199fda53546128afb6c0553749dfa85bd00799a4#note_7492
> > https://gitlab.collabora.com/pq/weston/commit/c618aa3dc9c2046e5540bbe444375f79e2e2c71f#note_7494
> > https://gitlab.collabora.com/pq/weston/commit/6a20dd4e16ae75431149c19e42b52eaf2e179b8c#note_7495  
> 
> Yeah.
> 
> > The rest looks good to me! I think it'd be good to merge 'libweston:
> > cancel idle_repaint on output destroy' ASAP in particular. The fbdev
> > and RDP patches are quite lazily Acked-by me, as I lost the will to
> > live whilst trying to review them. The rest of the series, with at
> > least the first three addressed, is Reviewed-by me.  
> 
> Your R-b and A-b should be in place, the revised series is at
> https://gitlab.collabora.com/pq/weston/commits/clonemode-8-part1

If one looked close enough at that branch, you'd see the commit
"libweston: new head-based output management API" duplicated. I've
fixed that.

> I ran through my test procedure from the v6 cover-letter, and
> everything worked as expected there, except for the DRM-backend tests.
> 
> drmModeAtomicCommit() fails with -EBUSY after the "start with two
> outputs, unplug and re-plug one" test. It happens already on upstream
> master, so it's not introduced by this series. The panel has seconds
> showing, and I am unplugging and replugging a DisplayPort monitor.
> 
> The zero outputs test fails on hotplug which looks like my bad in this
> series, so I need to fix that before we land this. The test works fine
> on upstream master.

Found the cause of the zero output hotplug failure, and it is because
of how the intermediate code works: libweston core is completely
migrated but as long as the DRM-backend fakes it, there's a crasher if
you have configured an output to be "off" in weston.ini and hotplugs
happen. The problem will disappear when the frontend stops creating an
output just to be able to turn it off, or the DRM-backend starts
properly creating outputs as the head-based output API expects. Anyway,
I've added a workaround in commit "weston: migrate DRM to head-based
output API", and now the test works.

The first 39 patches of the clone mode series, as present in the branch
https://gitlab.collabora.com/pq/weston/commits/clonemode-9-part1
are now pushed into master:
   61c4a730..42c0e148  master -> master


Thanks,
pq
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <https://lists.freedesktop.org/archives/wayland-devel/attachments/20180417/ec8def14/attachment.sig>