<dnkl>
awesome work with the scene graph branch! Trying it out as my main driver on one of my work machines.
<dnkl>
it'll be interesting to see if the improve damage tracking improves latency on large displays
<ifreund>
dnkl: thanks! There are still a bit more performance work to be done. For example we redundantly ask the layout generator for a new layout even when nothing has changed.
<ifreund>
also the current design of the wlroots scene graph API leads to overdamaging in many cases, e.e.
<ifreund>
e.g. adding a node at 0,0 then moving it to 50,50 in two consecutive calls damages twice
<ifreund>
but yeah, should still be better than master branch in theory :)
<dnkl>
I don't open/close windows that often, nor do I move them around... so shouldn't really be affected by the current limitations :)
<ifreund>
let me know if you hit any issues :)
<dnkl>
as always ;)
<dnkl>
a fullscreen window "jumps" when its tag is focused. Kind of looks like the window has been un-fullscreen:ed when the tag was unfocused. Known issue? intended behavior?
<novakane>
hmm with Firefox it removed the fullscreen if I switch to another tag
<novakane>
although it seems that with foot fullscreen works correctly
<ifreund>
dnkl: intended but may be reverted if its causing issues
<ifreund>
I thought it should be frame perfect but perhaps I'm either putting too much trust in clients or missing something
<dnkl>
only issue is the perceived frame imperfection. Other than that, I haven't noticed any actual problems
<dnkl>
hmm, may not be the clients (foot) fault... but the program running in foot that is slow to refresh after a resize
<dnkl>
yeah, with a screen full of plain "ls output, I'm not seeing any flicker
<dnkl>
but most fullscreen terminal applications are likely to flicker
<ifreund>
would be cool to have a proper look at using that synchronized updates term extension for better frame perfection therr
<ifreund>
but I'd say this is rivers fault for sending a pointless resize
<dnkl>
synchronized updates are unlikely to help in this scenario
<dnkl>
foot can't wait for the start escape before committing, since it doesn't know if the application supports synchronized updates
marmarismarmar has joined #river
<ifreund>
ah, not knowing if the application supports them is indeed a problem :/
marmarismarmar has quit [Ping timeout: 260 seconds]
uncomfy has joined #river
angry_vincent has quit [Ping timeout: 255 seconds]
angry_vincent has joined #river
angry_vincent has quit [Changing host]
angry_vincent has joined #river
uncomfy has quit [Remote host closed the connection]
angry_vincent has quit [Ping timeout: 255 seconds]
angry_vincent has joined #river
angry_vincent has quit [Changing host]
angry_vincent has joined #river
<plumeus>
I'm not that bright on shell scripts so here's a question: is there a way to set env vars inside the River init script so it'll work during the session?
<plumeus>
Example, set something like MOZ_ENABLE_WAYLAND=1 so Firefox defaults to Wayland in River
<plumeus>
but not necessarily outside of River if I wish to have some other X11 WM installed
<angry_vincent>
Yes.
<ifreund>
plumeus: setting stuff inside the init script will only apply to programs started from that init script
<ifreund>
if you want to set something for the entire river session you need to start river with it set
<ifreund>
I personally use a small start_river.sh wrapper script
<novakane>
oh you added that! Does it work well with gtk apps?
dbuckley has joined #river
Misthios has quit [Quit: Misthios]
Misthios has joined #river
Misthios has quit [Client Quit]
Misthios has joined #river
<ifreund>
novakane: works with everything I tested except firefox
<ifreund>
but it turns out it's impossible to get frame perfection while resizing firefox since the entirety of its content is a desync subsurface apparently
<novakane>
damn, that was the problem with Leon PR, hope it is less buggy though
<novakane>
oh, that sucks
<ifreund>
yeah, firefox is kinda shit once again
<ifreund>
I wish they would just stop using gtk and speak the wayland protocol directly
<novakane>
that would be nice, but I don't have too much hope
Szadek has quit [Ping timeout: 252 seconds]
bonic has joined #river
<novakane>
so I tested with Firefox, and while it's not perfect, it's nowhere near the awfulness it was with the previous transaction system
bonic has quit [Quit: leaving]
bonic has joined #river
<ifreund>
yeah
<ifreund>
it just looks kinda funny if you have river draw borders and go fast
<ifreund>
have any of you ever wished that WAYLAND_DEBUG=1 would print array contents?