| alkisg | KitsuWhooa: do you have any hypothesis on which generations this affects? I'm thinking of disabling pageflip in all of nvidias, but maybe there are some generations unaffected? | 13:55 |
|---|---|---|
| KitsuWhooa | alkisg: no idea unfortunately. I tested mx200, mx400, mx440, fx5XXX and they all had the same issue | 13:58 |
| alkisg | Thanks; I'll try to test with some newer one if I can find | 13:58 |
| KitsuWhooa | My guess is all of them that don't use dri3/glamor | 13:58 |
| alkisg | Ah | 13:58 |
| KitsuWhooa | I think I have a pcie one somewhere, but I can't test at the moment | 13:59 |
| alkisg | That that makes more sense, if all were affected, more people would have noticed | 13:59 |
| KitsuWhooa | I wonder if it's an exa issue | 14:00 |
| KitsuWhooa | NoAccel would've fixed it though if that was the case | 14:00 |
| KitsuWhooa | Also, alkisg, is there any way to not get X to replace its log file on segfault? | 14:00 |
| KitsuWhooa | In ltsp | 14:01 |
| KitsuWhooa | And how would I go about stopping whatever starts x automatically | 14:01 |
| KitsuWhooa | I was hoping it'd be lightdm but it's not | 14:02 |
| alkisg | LTSP is using LDM, our own display manager, | 14:02 |
| KitsuWhooa | Couldn't find any ldm services | 14:02 |
| alkisg | to go to text mode by default, try SCREEN_07=shell, with no other screens active, | 14:02 |
| alkisg | ltsp supports many xorg/screens, so it has its own management service, ltsp-client-core, you don't need that, just disable the ldm screen | 14:03 |
| alkisg | So, SCREEN_07=shell, then run xinit, then mate-session, then check xorg.log | 14:04 |
| KitsuWhooa | Alright, I'll do that later. Thanks | 14:04 |
| alkisg | np, thank you too | 14:04 |
| alkisg | Btw, ldm doesn't have this problem, I don't know if it's because it's not using double buffering? | 14:04 |
| alkisg | Or maybe it's using wwm, and this isn't using compositing... | 14:05 |
| KitsuWhooa | Yeah, I noticed that | 14:05 |
| KitsuWhooa | Probably because it doesn't do any compositing, yeah | 14:05 |
| alkisg | I wonder if other flavors do have that issue... I'm guessing gnome and kde, but not lxde... | 14:05 |
| KitsuWhooa | A bit unrelated, but since I don't have a riva tnt2, would you be able to install the X/nouveau debug symbols (assuming Ubuntu has them), run X under gdb and get a full backtrace? (bt full) | 14:07 |
| KitsuWhooa | That might be a bit easier to track down and fix | 14:07 |
| KitsuWhooa | Assuming it worked originally with the same driver | 14:08 |
| KitsuWhooa | I kinda doubt gnome and kde would be fast enough on such old cards | 14:08 |
| KitsuWhooa | Or even support compositing. These cards are OpenGL 1.X IIRC | 14:08 |
| alkisg | KitsuWhooa: I think it's the same issue that's causing tnt2 to follow some code path that segfaults, and if that original issue is solved, tnt2 won't segfault either, | 14:13 |
| alkisg | and I'm not very accustomed to gdb, and the schools just started and have a lot of needs, so it'll take me a while... | 14:13 |
| alkisg | I'm thinking of pushing a workaround for now, and revisit this later when things settle down, e.g. in a month or so | 14:14 |
| KitsuWhooa | That could also work, yeah | 14:14 |
| KitsuWhooa | I wonder if I can find one of those cards locally. Or maybe I have one and don't know it | 14:14 |
| alkisg | If you think you can fix this, then maybe I can set up a pc here for that, and give you remote root access? | 14:15 |
| KitsuWhooa | I can't promise anything but I can try | 14:15 |
| alkisg | Great. Will do this within the week, maybe wednesday. | 14:16 |
| * alkisg is having some python2to3 related issues to solve atm, more urgent... :/ | 14:16 | |
| KitsuWhooa | If X crashes after the login screen, I might need vnc too | 14:17 |
| alkisg | epoptes-client --connect ip-of-kitsuwhooa => gives both console and vnc :D | 14:18 |
| KitsuWhooa | Somehow python2 is still being used | 14:18 |
| KitsuWhooa | Here's to hoping it eventually gets dropped from enterprise too | 14:18 |
| KitsuWhooa | Excellent. I assume I need to install epoptes on my desktop, but that's not an issue | 14:18 |
| alkisg | Or just xvnc4viewer -listen; or I can give you ssh; or vnc, really not an issue for me, I have public ips etc | 14:19 |
| alkisg | There's also this trick: ssh client; export $(/usr/share/epoptes-client/get-display) => sets up xauthority and display so that you have access to xorg | 14:20 |
| alkisg | We'll find many ways, np there | 14:20 |
| KitsuWhooa | I'm asking so that I can have my end set up by then, but good to hear | 14:20 |
| alkisg | If you happen to have a public IP or port forward etc, then setting up epoptes is much more effective | 14:21 |
| KitsuWhooa | I do, yeah | 14:21 |
| alkisg | Great, let's do it this way then | 14:21 |
| KitsuWhooa | Alright, I'll look up the docs later | 14:21 |
| alkisg | apt install epoptes should be enough | 14:21 |
| KitsuWhooa | I was thinking port-wise | 14:22 |
| alkisg | Port 789 | 14:22 |
| KitsuWhooa | Alright | 14:22 |
| alkisg | And I think 5500 for vnc | 14:23 |
| alkisg | I can "send you" a vm if you want a quick test, when you're ready | 14:23 |
| KitsuWhooa | I'll probably be in a few hours | 14:24 |
| alkisg | ok | 14:24 |
| KitsuWhooa | I'll highlight you and we can go from there | 14:24 |
| alkisg | Great | 14:24 |
| * alkisg guesses that one way to check for an nvidia card would be: if [ -d /sys/module/nouvaue ]; then X_OPTION_01=...pageflip off... | 14:42 | |
| alkisg | *without the typo, of course... | 14:42 |
| alkisg | Whoops, no, it's too early at the point when init-ltsp.d runs, the module isn't yet loaded :( | 14:53 |
| KitsuWhooa | Newer cards that use nouveau would also be picked up by that | 14:54 |
| KitsuWhooa | My suggestion would be to make a blacklist with pci dev ids | 14:54 |
| alkisg | You said something that I understood like "only dri2 cards affected"; do you think that might also affect non nvidia cards that use dri2 instead of 3? Are any of these common for me to check with? I do have lots of cards... | 14:59 |
| alkisg | (correction, lots and lots of *old* cards :D) | 15:01 |
| alkisg | Looking at the pci-ids from https://envytools.readthedocs.io/en/latest/hw/pciid.html, it looks like a lot of work to blacklist the correct ones | 15:04 |
| alkisg | Using ranges wouldn't be easy either as they have non-graphics cards in that pci ids space as well... | 15:05 |
| KitsuWhooa | alkisg: I seem to have epoptes set up | 17:15 |
| KitsuWhooa | <alkisg> You said something that I understood like "only dri2 cards affected"; do you think that might also affect non nvidia cards that use dri2 instead of 3? Are any of these common for me to check with? I do have lots of cards... <-- I'm not sure, we'd need to see what nouveau does | 17:15 |
| alkisg | KitsuWhooa: if you want a test, pm me your ip... | 17:20 |
| alkisg | (to connect a vm to you via epoptes-client) | 17:21 |
| KitsuWhooa | done | 17:21 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!