00:00 goonieb: what to try then?
00:00 goonieb: "nvidia_drm.modeset=1"
00:01 gnarface: yea i guess try turning that on and see if it fixes both of them at once
00:02 gnarface: if it doesn't, i'd probably try feeding it some explicit KMS resolution settings too
00:03 goonieb: and declare modules MODULES="nvidia nvidia_modeset nvidia_uvm nvidia_drm"
00:04 gnarface: i'm not sure if that part is necessary or not. it would depend on your distro i guess.
00:04 goonieb: gnarface: will try soon
00:05 goonieb: gnarface: well that would be were we could play the precedence trick
00:06 goonieb: MODULES="nouveau nvidia nvidia_modeset nvidia_uvm nvidia_drm"
00:08 gnarface: hmmm. the bios order might be relevant too
00:12 goonieb: gnarface: switcing pcie?
00:13 pl44c: so when I tell xorg to load intel and nouveau ddx drivers it kills itself when adding a display
00:13 pl44c: https://dpaste.de/squy
00:13 pl44c: xorg log/dmesg/xorg.conf
00:14 karolherbst: pl44c: you need to configure both devices
00:14 pl44c: actually that was the wrong xorg log
00:14 pl44c: karolherbst: what do you mean
00:18 pl44c: I had two device sections and specified which ddx driver to load
00:18 pl44c: if I swap intel with modesetting I don't get a crash
00:19 karolherbst: weird
00:19 karolherbst: might be that the intel driver is broken or something
00:19 karolherbst: anyway, you want to add the BusID to the intel device as well
00:21 pl44c: alright let's see if that fixes it
00:25 gnarface: goonieb: yea, that's what i meant, switching between PCI/PCIE or onboard
00:25 gnarface: goonieb: (or if they're both add-ons to the same bus, just changing their slot order i guess - bios feature dependent)
00:27 pl44c: karolherbst: nope still crashes after specifying busid for both devices
00:27 pl44c: must be something wrong with the intel ddx as it works only when the intel card uses modesetting
00:28 pl44c: no difference if the nvidia chip uses modesetting or not
00:36 pl44c: ok the system shut itself down
00:36 pl44c: I don't think things are working fine at all
00:37 karolherbst: mhh, weird
00:37 karolherbst: for me it does
00:37 pl44c: where would I pull logs from before the system killed itself
00:38 pl44c: so you're able to run intel/nouveau and add a display via xrandr?
00:43 pl44c: can't replicate the system shutdown
00:49 joepublic: on some systems /var/log/kern.log is a persistent companion to dmesg that survives reboots.
00:52 pl44c: https://dpaste.de/Z8N7
00:52 pl44c: this is all that's there at the end before the log starts again at the latest boot up
00:53 karolherbst: pl44c: well, using modesetting with the intel GPU is fine as intel prefers it that way anyway
00:54 pl44c: I also don't notice tearing on the LVDS display
00:54 pl44c: which I think is all that would affect in that regard
00:55 pl44c: GLXVBlank is default on nouveau right?
01:02 pl44c: ok setting that to true did nothing so I assume it is
01:04 pl44c: and various things I've tried in compton make no difference
01:05 pl44c: is fixing tearing just not possible with an optimus laptop when you drive external displays?
01:33 pl44c: figured out what was causing X to die instantly when adding a display it was "NoTrapSignals"
01:33 pl44c: so now rather than crashing the exernal display is just garbled mess
01:41 gnarface: pl44c: i would think that compton with vsync on should eliminate the tearing. maybe we should more carefully define tearing here?
01:41 gnarface: i would be more willing to suspect something was sabotaging the ability to enable vsync
02:04 pl44c: well when I play a video with mpv there's a diagnal line through the video
02:04 pl44c: I'm trying different combinarions of backends for vsync
02:04 gnarface: try the opengl one maybe
02:16 pl44c: tried every possible backend and vsync combo
02:16 pl44c: same lines
02:16 gnarface: i know i've asked this already, but was it only mpv doing that, or does mplayer and vlc misbehave the same way?
02:17 pl44c: firefox does the same when scrolling
02:17 gnarface: hmm. firefox also uses opengl by defaul tnow
02:17 gnarface: *default now
02:17 gnarface: it might be a clue
02:17 pl44c: you have to force enable it but yeah
02:17 gnarface: it might be a clue, but not an answer to my question
02:18 gnarface: the thing is that it's still possible for programs to ignore the request to enable vsync, or to even actively unset it.
02:19 gnarface: so if you can find some programs NOT doing it, that might be a useful clue
02:19 gnarface: which is why i keep suggesting you try any other video player ever
02:19 pl44c: using mplayer would be the same I'd guess
02:19 pl44c: I could try ffplay
02:19 gnarface: it's not safe to assume mpv and mplayer behave the same for general purposes
02:20 gnarface: it's not even safe to assume that mpv has the best defaults or the most reliable behavior of the available video players (not that it's the worst either, but i don't know why people make this particular mistake for mpv more than other players)
02:21 pl44c: same exact lines with vlc
02:21 gnarface: ok, so that's a useful clue
02:22 gnarface: is the line visible regardless of window size, or is it only happening in fullscreen?
02:23 pl44c: the lines varry based on what is displayed
02:23 pl44c: if I loop the same scene the lines are the same or close to it
02:23 gnarface: also useful, but again, not an answer to the question i just asked
02:24 pl44c: yes its the same pattern of line regardless of window size/fullscreen
02:24 gnarface: ok, good to know
02:24 gnarface: you were using i3WM?
02:24 pl44c: yes
02:25 gnarface: ever tried a non-tiling window manager?
02:25 gnarface: just as a test?
02:25 gnarface: i3WM causes ... *lots* of fundamental problems. this would be a new one, but not beyond what i'd expect is possible.
02:26 pl44c: Don't take away my baby I've been using her for years! (yeah sure I'll install openbox)
02:26 gnarface: no no not openbox
02:26 pl44c: so like a compositing wm ?
02:26 gnarface: yea no compostior and non-tiling
02:26 gnarface: blackbox or e16 should be fine
02:26 gnarface: well
02:27 gnarface: e16 in recent builds has a compositor but you can turn it off. it might be off by default
02:27 gnarface: e17 would also work as a test if you disable the compositor, but it's on by default in e17
02:27 pl44c: oh this is a full DE
02:28 pl44c: I'll just use blackbox less to mess with
02:28 gnarface: yea, try blackbox, make sure compton isn't starting over it or whatever
02:28 gnarface: then maybe just run mplayer -vsync [filename]
02:29 gnarface: what i'm trying to figure out here is why nothing you've changed makes a difference. that suggests we're both missing something important.
02:29 gnarface: but the only way i know to proceed is to keep changing new stuff until something matters.
02:32 gnarface: i know i asked this one before too but... you're running all these displays at their native resolution, right?
02:32 pl44c: yeah 3 displays at 1920x1200
02:33 pl44c: 1 over mdp to dp and two over dvi to hdmi
02:33 pl44c: same monitor model
02:33 gnarface: ok, if you're sure. i just have to ask, because the type of tearing you seem to be describing could also be related to low-quality linear scaling algorithms sometimes embedded in the flat panel hardware itself
02:33 gnarface: but i'd only expect that to be a complication in the case where one or more of them was not being run at it's native resolution
02:34 pl44c: I've seen these line patterns before back when I used a 9500GT and nouveau IIRC
02:34 pl44c: I think I managed to fix it way back then on that hardware
02:35 pl44c: actually no I didn't
02:35 pl44c: but the diagnal line goes into a straight line
02:36 pl44c: so it's kinda like -\
02:36 gnarface: top left to bottom right?
02:36 pl44c: or like -\_
02:36 gnarface: yea, i've seen that before with scaling problems
02:36 gnarface: not actually in Xorg though
02:36 gnarface: not really even in Linux
02:37 gnarface: but in an emulator running on a handheld device that used some ... GPL-violating Linux clone
02:37 gnarface: they did mitigate the issue with a driver fix, i think
02:39 pl44c: ok blackbox really does not like multi monitor with default config
02:39 pl44c: anyway now I get tearing on the laptop display which I didn't get before
02:39 pl44c: not using compton
02:40 gnarface: oh, yea i had to heavily customize my xorg.conf but you could in theory just run some xrandr commands to correct it after launch then hit the menu entry to refresh the blackbox config
02:40 gnarface: e16 would have been bigger but easier to wrangle because it has menus for controlling the monitor handling
02:40 gnarface: but
02:40 pl44c: well I saved the xrandr line to enable my displays
02:40 gnarface: making it worse is actually an important clue
02:40 pl44c: so I just ran that
02:41 pl44c: mplayer -vsync didn't help
02:41 pl44c: same tear
02:41 gnarface: so i wonder if just running compton with vsync on is what cured the laptop's display of tearing, or if something else changed in the window manager behavior
02:41 pl44c: the laptop display just didn't tear
02:41 pl44c: on i3
02:41 pl44c: with no compositor
02:42 gnarface: hmmm. strange
02:42 pl44c: the monitors would
02:42 gnarface: yea, something is definitely wrong but i can't even say for sure it's got anything to do with vsync now
02:43 gnarface: the video you're playing is also at the native screen resolution, right?
02:43 pl44c: well when I fullscreen it it just goes to the laptop display
02:43 gnarface: hmm
02:44 gnarface: with mplayer you should try -fs and -zoom maybe?
02:44 gnarface: i'm curious if the software scaler will do something about this
02:44 pl44c: I'd still have to drag it pixel perfect onto the display
02:44 pl44c: blackbox is unideal for testing multi head setups it seems
02:44 pl44c: will xfce be good?
02:45 gnarface: i just don't know what composites by default or not anymore so i listed the 2 ones i knew didn't and the one i knew how to turn it off for
02:46 pl44c: XFCE has one and I know how to turn it off
02:46 pl44c: think it isn't even on by default
02:46 gnarface: ah, worth a try then
02:46 gnarface: i think if you haven't tried e16 or e17 you would like the multi-monitor controls they have though
02:47 gnarface: blackbox just leaves it up to you to define the entire layout in the xorg.conf
02:47 gnarface: (which works fine for nouveau but i've run into trouble with the official drivers failing to parse standard options correctly)
02:47 pl44c: yeah I don't want to specify everything specifically for something I'm just testing
02:54 pl44c: same lines here
02:54 pl44c: just various -\_ lines
02:55 pl44c: sometimes straight lines across
02:55 pl44c: just depends on what's being displayed
02:57 pl44c: they are consistent lines
02:58 gnarface: yea, the more you describe them, the less it sounds like a vsync issue and the more it sounds like a weird scaling issue
03:00 pl44c: I've used these monitors on different systems with AMDGPU and the NVIDIA driver with no issue
03:01 pl44c: the tearing is different from when on this system I was using the NVIDIAdriver
03:02 pl44c: about ready to just get rid of this system it's been nothing but trouble
03:20 imirkin: pl44c: what's the issue?
03:20 imirkin: there's a lot of scrollback, and i don't really feel like reading through it
03:31 gnarface: imirkin: short version is he's seeing a hard diagonal tear when watching video or scrolling firefox. it seems to be immune to vsync settings
03:32 gnarface: imirkin: (he's got a 3-monitor setup with one being the laptop's own display panel. tearing on the *laptop display panel only* seems to be only present in some window managers. however, compositing doesn't appear to be the differentiating factor)
03:33 gnarface: hmmm. and he's using some adapters.
03:33 gnarface: pl44c: and a docking station, right?
03:34 pl44c: yes the docking station
03:34 gnarface: imirkin: pl44c will have to confirm but i think he said he's using a docking station for the second two monitor ports and the adapters are plugged into that
03:34 pl44c: now I tested the laptop lcd again and it does tear
03:34 pl44c: only at the top though
03:34 pl44c: and a straight line
03:34 pl44c: so all WMs are the same
03:35 gnarface: ok, that's an important clarification. that type of detail can really throw off diagnosis.
03:35 gnarface: imirkin: my own theory was it's about vsync at first but now i'm not so sure.
03:35 pl44c: the video I was testing did not move much at the top
03:37 imirkin: triangle is a sign of GL + no vsync
03:38 imirkin: i'd recommend using xf86-video-nouveau and trying to avoid GL for regular desktop-py things
03:38 imirkin: a straight line down the middle is a different kind of issue
03:39 imirkin: i.e. scanout starts with one buffer and completes with another
03:39 imirkin: (also means "no vsync"
03:40 gnarface: hmm. so i guess the question is about why vsync didn't enable for him?
03:40 imirkin: well, with X, there's no vsync
03:41 gnarface: hmmm
03:41 imirkin: you can get it with direct presentation, by using GL for example
03:41 imirkin: but not with "raw" X
03:41 imirkin: which is all effectively front-buffer-rendered
03:50 pl44c: ok so switching to nouveau from modesetting changings the lines to just straight
03:50 pl44c: how exactly do I stop these lines
03:52 pl44c: I tried disabling gpu acceleration in firefox but that didn't help
03:52 imirkin: blink at 60hz
03:53 imirkin: that way you won't notice it
03:53 imirkin: a little more seriously... i think using a compositor is basically the only way to avoid horizontal tearing in X
03:53 imirkin: some DDX's have a "TearFree" option, but nouveau does not
03:53 imirkin: however the implementation of that is basically equivalent to having a compositor built-in
03:54 pl44c: If I try to use intel for whatever reason and add a display I just get garbled mesh
03:54 pl44c: doesn't matter if I'm using nouveau or modesetting
03:54 pl44c: on the nvidia chip
03:55 imirkin: wait, are some of the screens on a different gpu?
03:56 pl44c: the intel chip powers the laptop display and the nvidia chip does all the ports
03:56 imirkin: ah
03:56 imirkin: well there's no vsync with reverse prime
03:57 imirkin: at least i'm like 99% sure of that... airlied would know off-hand
03:57 imirkin: even with compositors and whatever else
03:58 pl44c: so is there no way to just universally solve screen tearing once I enable the nvidia chip for displays
03:58 imirkin: afaik it's an unsolved issue when displays are attached to multiple chips
04:01 gnarface: bummer
04:01 gnarface: can a dual-link dvi port power 2 monitors?
04:01 pl44c: wish I would've known this sooner would have saved me a bunch of time meddling with pointless rubish
04:02 gnarface: i'm sorry, i remember you even mentioned it as a possible issue and i had no idea.
04:02 imirkin: gnarface: no
04:03 imirkin: pl44c: might want to double-check it with people more in the know. hopefully airlied will pipe up about it.
04:03 imirkin: pl44c: perhaps with wayland, it's all doable.
04:03 imirkin: let's ask in #dri-devel
04:04 gnarface: date
04:04 gnarface: woops, wrong window
04:08 pl44c: so the non-free stack can do it?
04:09 pl44c: guess not nevermind
04:18 pl44c: I think tearing on the laptop display went away when I switched to nouveau ddx on the nvidia
04:18 pl44c: I can't get it to tear anymore
04:31 pl44c: yep switching the nvidia chip to nouveau fixed tearing on the laptop lcd
04:32 pl44c: so it's just modesetting/modesetting that causes it to tear
04:32 pl44c: however that works
04:34 pl44c: anyway it's late now
04:35 pl44c: thanks for the help even though the final solution wasn't found