On 2018-02-02 03:45 AM, Pekka Paalanen wrote:
On Thu, 1 Feb 2018 22:10:33 +0000
Daniel Stone <dan...@fooishbar.org> wrote:
On 1 February 2018 at 22:00, Derek Foreman <der...@osg.samsung.com> wrote:
On 2017-12-14 05:40 AM, Pekka Paalanen wrote:
The 'head' member of 'struct weston_output' is going to go unused and
then disappear, so stop using it and find a head from the proper list.
However, this leaves a problem in cms-colord: if you have multiple
monitors driver with the same CRTC, what do you say to the color
management system? The monitors could be different, but all the color
LUTs etc. are in the CRTC and are shared, as is the framebuffer.
Do the simple hack here and just use whatever head happens to be the
first in the list.
I am a complete non-expert in this area, so if I'm making no sense feel free
to tell me to shut up...
I think if someone's going through the effort to properly setup color
management, then we can't use cloned heads off a single CRTC?
We should probably disallow a CRTCs to drive multiple heads if two or more
of those heads have differing color profiles?
If I went to the trouble of calibrating two displays, I would probably be
extremely surprised to see them looking different with clone mode enabled.
Yeah, what Derek said. I think non-homogeneous colour properties when
colour management is enabled, should be cause to unclone. That being
said, it's a niche enough usecase that I'd be comfortable landing it
with documentation that it was a known shortcoming. I'd be super
comfortable landing it if we also had an output configuration tweak we
could use to force non-clone mode, or really even just a simple
out-of-tree patch (or environment variable, a la atomic?) that allows
people to work around it easily enough by disabling clones.
I agree with Derek's suggestion that we should check color properties
and fail the shared-CRTC cloning, and with Daniel that we could just
document this. I'm not sure what the "output configuration tweak"
refers to. Do you mean an option that would just disable shared-CRTC
I can get behind just documenting it.
We have Weston that handles all option parsing, and Weston also does
the decision to use or not use shared-CRTC cloning by how it attaches
heads to outputs. Libweston will not try shared-CRTC cloning unless
explicitly told to, and it will not transparently fall back to
something else. After the complete clone mode series, which will not
include independent-CRTC cloning support, the option to disable
shared-CRTC cloning would be equivalent to a weston.ini that does not
configure cloning at all.
For myself, I've been struggling with the ramifications of shared-CRTC
vs independent-CRTC cloning, and didn't realize until you pointed out to
me out-of-band that independent-CRTC cloning isn't just fallback
operation (and in fact won't work at all without additional work).
However, the weston.ini syntax as is does not differentiate between
shared-CRTC and independent-CRTC cloning. My intention is that Weston
(not libweston) attemps shared-CRTC cloning and falls back to
For the cms-colord plugin to be able to veto shared-CRTC cloning we
would need libweston infrastructure work to either offer plugins a veto
API or make libweston core aware of color profiles. An alternative
would be to have Weston query the color management plugin directly if
heads are compatible for shared-CRTC mode.
Yes, it had occurred to me that this could be particularly painful. :(
The plug-in veto idea, seems like it would be a fairly heavy amount of
overkill if cms-colord would forever be the only plug-in to actually use it.
If I'm understanding all the pieces in play at this point... We'd
really just be giving cms-colord the ability to stop weston from
launching at all, since independent-CRTC cloning doesn't actually work,
and is (quite reasonably!) out of scope for this series and its follow up?
One more idea coming to mind is that the "same-as" directive in
weston.ini could refer to either exclusively shared-CRTC clone mode or
shared-CRTC with fallback to independent-CRTC mode, and we could use a
hypothetical future output layout configuration directives to force
independent-CRTC mode by defining two outputs to show the same area of
Given that, what would you recommend for now?
Can we generate a log message when cms-colord configuration results in
suboptimal display for cloned heads (possibly including some strings to
identify the display/displays that look wrong)?
Or even a warning when cms-colord and clones are both enabled at the
At this point, I think whoever the first person to need clone+cms-colord
simultaneously is going to have to step up and do some heavy lifting.
As long as we're not silently doing something surprising with color
profiles, I think it should be ok? Since clone mode is all new, this
isn't going to cause a regression for anyone.
wayland-devel mailing list