Plan_IX has quit [Remote host closed the connection]
blue__penquin has joined #wayland
<emersion>
MrCooper: we want to have total control over the client buffer submission, and be able to do things like measuring client render time
reillybrogan has joined #wayland
<MrCooper>
emersion: the compositor can't know when the client started drawing, if that's what you mean
<MrCooper>
I'm not sure what you mean by "total control over the client buffer submission"
<emersion>
i mean that xwayland or mesa doesn't try to do whatever magic under our nose
ecloud has quit [Ping timeout: 480 seconds]
ecloud has joined #wayland
<pq>
you mean like, a client *must* start rendering when it received wl_surface.frame callback?
<pq>
that's already broken, because a client could start on presentation-time feedback
<pq>
on the event, I mean, not even predicting anything
<pq>
if you want to know client render time, I think there is no way around having to define a protocol ext for it.
<pq>
where did this dicussion begin?
<emersion>
no, i don't mean that a client *must* start rendering at anytime
* pq
asks and immediately disappears.
<emersion>
i just mean that i want the middlemen to disappear
sca has quit []
<emersion>
and to not take policy decisions
<emersion>
policy decisions are for the compositor, not for xwayland or mesa
cseader has joined #wayland
cseader has quit [Remote host closed the connection]
alanquayle[m] has joined #wayland
alanquayle[m] has quit [Remote host closed the connection]
<MrCooper>
I want users not to hit severe stuttering due to compositors failing to wait for buffers to become ready; once the main ones do, the workaround can be dropped from Mesa
<swick>
I recently looked at presentation-time again and it does not make any guarantees when the feedback is send. why again do we think it's a good idea to draw on presentation-time feedback?
<swick>
also, the clients can already wait for buffers to become ready and that is totally fine so I don't see the problem with Xwayland doing that either
<swick>
it would be nice if Xwayland could pass through everything to the wayland compositor as soon as possible but that only seems to make sense when compositors wait for buffers and the commits can be scheduled properly
<MrCooper>
swick: Mesa WSI waits for buffers to become ready, not Xwayland
<MrCooper>
Xwayland will pass through buffers directly once it supports presentation-time
<MrCooper>
(right now it waits for a pending frame callback before sending another buffer if the X11 Present client is asking for FIFO behaviour)
<swick>
ah, ok, mesa waiting isn't any different
<swick>
what exactly do you gain from presentation-time?
<MrCooper>
reliable information about which buffer was presented or skipped
AJ_Z0 has quit [Quit: ZNC 1.6.1+deb1~ubuntu12.04.2 - http://znc.in]
AJ_Z0 has joined #wayland
rgallaispou has quit [Read error: Connection reset by peer]
blue__penquin has quit []
Rasi27 has joined #wayland
Rasi27 has quit [autokilled: Suspected spammer. Mail support@oftc.net with questions (2021-06-11 15:48:02)]
Magneand has joined #wayland
Magneand has quit [autokilled: Suspected spammer. Mail support@oftc.net with questions (2021-06-11 15:50:56)]
arthurpbs0 has joined #wayland
arthurpbs0 has quit [autokilled: Suspected spammer. Mail support@oftc.net with questions (2021-06-11 16:04:09)]
danvet has quit [Ping timeout: 480 seconds]
aleasto has joined #wayland
hau63n has joined #wayland
danvet has joined #wayland
khfeng has quit [Ping timeout: 480 seconds]
marathon has quit [Quit: Leaving]
AbbyH has joined #wayland
AbbyH has quit [Remote host closed the connection]
neonking has joined #wayland
neonking has quit [Remote host closed the connection]
neonking has joined #wayland
dagelf1 has joined #wayland
dagelf1 has quit [Remote host closed the connection]
Syliss_ has joined #wayland
Syliss_ has quit [Remote host closed the connection]
robert_mader has quit [Quit: Leaving.]
<Plagman>
seems like there needs to be a knob right now where mesa can understand it doesn't need to do that
<Plagman>
since some compositors already do not need that behaviour (quite the opposite)
columbar1 has joined #wayland
BaloneyGeek has joined #wayland
BaloneyGeek has quit [Remote host closed the connection]
columbarius has quit [Ping timeout: 480 seconds]
SteepPeak[ic] has joined #wayland
SteepPeak[ic] has quit [Remote host closed the connection]
ppascher has quit [Ping timeout: 480 seconds]
<swick>
meh, a few more month and we can just remove it altogether
AJ_Z0 has quit [Remote host closed the connection]
alyssa has joined #wayland
AJ_Z0 has joined #wayland
* alyssa
stares at Weston headless
<alyssa>
* weston stares back
<swick>
MrCooper: IIRC there is nothing in the protocol which guarantees you that the event is sent at a particular time so it is valid to send it a few frames later which is why I dont understand how it can possibly replace the frame callback
<alyssa>
oh weston doesn't initialize egl then ok
shankaru1 has quit []
randy7 has joined #wayland
randy7 has quit [Remote host closed the connection]
pnowack has quit [Quit: pnowack]
danvet has quit [Ping timeout: 481 seconds]
shankaru has joined #wayland
neonking has quit [Remote host closed the connection]
hardening has quit [Ping timeout: 480 seconds]
hau63n has quit [Quit: Leaving]
pac85[m] has joined #wayland
aleasto has quit [Quit: Konversation terminated!]
aleasto has joined #wayland
aleasto has quit [Remote host closed the connection]