KitsuWhooa | alkisg: I tested a https://www.asus.com/us/Graphics-Cards/A9250TD128M/overview/ and it worked out of the box, so I doubt ATi cards are affected | 05:25 |
---|---|---|
alkisg | KitsuWhooa: nice... do you think those horizontal lines in vbox were related? I haven't tested if pageflip off fixes vbox as well.. | 05:26 |
KitsuWhooa | Those go away by switching ttys, the ones on nvidia hardware don't | 05:27 |
alkisg | And the one in tnt2 causes a segfault... those all might be bad codepaths, forced by a "central" bug somewhere... | 05:38 |
alkisg | E.g. uninitialized resolution variable => may cause either corruption or segfault later on... | 05:38 |
alkisg | The corruption can be fixed by xrandr -s other-resolution; xrandr -s original-resolution | 05:39 |
KitsuWhooa | which is why a gdb backtrace would be useful | 05:39 |
alkisg | Sure; I'll set up the tnt2 system within the week | 05:40 |
KitsuWhooa | Yeah, I was just mentioning it | 05:40 |
KitsuWhooa | Apologies if it sounded a bit harsh | 05:40 |
alkisg | No no I very much appreciate all your input and help in this | 05:41 |
KitsuWhooa | turns out I found a TNT2 | 06:38 |
KitsuWhooa | and also what identifies itself as "NV10GL Quadro" | 06:43 |
KitsuWhooa | 01:00.0 VGA compatible controller: NVIDIA Corporation NV5 [Riva TNT2 Model 64 / Model 64 Pro] (rev 15) | 07:02 |
KitsuWhooa | Yup, X crashes right after the login screen | 07:04 |
alkisg | KitsuWhooa: do you have epoptes working in your ltsp setup? Could you run this at the login screen, before login, and see if it's still crashing or not? xrandr -s 640x480; xrandr -s original-resolution | 07:11 |
KitsuWhooa | I don't have LTSP set up on epoptes, but I do have a shell | 07:11 |
KitsuWhooa | two of them | 07:11 |
alkisg | OK, you'd need to export DISPLAY and XAUTHORITY for xrandr to work | 07:11 |
KitsuWhooa | would I need to rebuild the squashfs image to add epoptes? | 07:12 |
alkisg | Yes | 07:12 |
KitsuWhooa | Any way to rebuild the image without keeping the old one around during build? | 07:12 |
KitsuWhooa | not enough disk space :p | 07:12 |
alkisg | ltsp-update-image --no-backup -c / | 07:13 |
KitsuWhooa | excellent, thank you | 07:13 |
alkisg | ...stop nbd-server first so that it's not in use | 07:13 |
KitsuWhooa | Also, I think I found a mate bug | 07:13 |
KitsuWhooa | https://tasossah.com/s/3d40462bc831.jpg | 07:13 |
KitsuWhooa | I don't even know what I did to make those two network indicators show up | 07:14 |
alkisg | It goes away if you kill/rerun nm-applet, or on next login | 07:14 |
alkisg | It's some race condition somewhere, I haven't pinpointed it | 07:14 |
KitsuWhooa | So, I need both epoptes and epoptes-client on the server? | 07:15 |
alkisg | Yes. apt install epoptes should be enough, it installs epoptes-client as well | 07:15 |
KitsuWhooa | ah | 07:15 |
KitsuWhooa | okay | 07:15 |
KitsuWhooa | Right, after being distracted for a bit | 08:04 |
KitsuWhooa | I can't seem to be able to find the nouveau debug symbol package | 08:04 |
KitsuWhooa | Ah, they're on a separate repo | 08:11 |
KitsuWhooa | alkisg: that did in fact make it not crash | 08:43 |
KitsuWhooa | I have to say though, the TNT2 is really slow at 1680x1050 | 08:45 |
KitsuWhooa | and I got epoptes set up too, really cool | 08:45 |
alkisg | KitsuWhooa: it sounds like it's something in xorg initialization that goes wrong, and gets correctly initialized with xrandr | 09:05 |
KitsuWhooa | Eeeeyup | 09:22 |
KitsuWhooa | I'll try with gdb now and see if I can find anything interesting | 09:22 |
KitsuWhooa | Also, wakeonlan works and it got me more excited than it should have :p | 09:23 |
KitsuWhooa | s/and/to/ | 09:23 |
KitsuWhooa | alkisg: I ran startx, I attached gdb, and then after a while epoptes said my client went offline and killed my remote terminal | 09:41 |
KitsuWhooa | any idea why? | 09:41 |
KitsuWhooa | (X was in a frozen state) | 09:41 |
KitsuWhooa | network never died either | 09:42 |
alkisg | KitsuWhooa: how much ram does that client have? | 09:51 |
alkisg | oom? | 09:51 |
KitsuWhooa | I checked dmesg, not oom | 09:51 |
KitsuWhooa | it has 3x256 DDR1 | 09:51 |
alkisg | Strange | 09:52 |
alkisg | When the client is logged off, do you see it in epoptes as a blue icon? | 09:52 |
alkisg | (fat client) | 09:52 |
KitsuWhooa | yeah | 09:52 |
KitsuWhooa | and I can even spawn a shell | 09:52 |
alkisg | Btw epoptes is supposed to reconnect after a minute of inactivity, did it? | 09:53 |
alkisg | The remote terminal was "root", not "user", right? | 09:54 |
KitsuWhooa | yup | 09:54 |
KitsuWhooa | it did not reconnect, no | 09:54 |
alkisg | (root epoptes-client runs from systemd as root, while user runs from /etc/xdg/autostart as the user) | 09:54 |
KitsuWhooa | and I checked that the network was still functioning | 09:54 |
alkisg | That's rather strange, if you have a vt on the client it would be interesting to rerun epoptes-client as root, make it crash again, and see the stderr there | 09:55 |
KitsuWhooa | I do yeah | 09:55 |
KitsuWhooa | let me try again | 09:55 |
KitsuWhooa | I have two vts and the remote shell | 09:56 |
KitsuWhooa | I ran startx on the vt and attached gdb through the epoptes root shell | 09:56 |
alkisg | (btw to open another vt if needed, it's openvt bash) | 09:57 |
KitsuWhooa | I have two | 09:57 |
KitsuWhooa | but I will, thanks | 09:57 |
alkisg | e.g. if you want to attach gdb from a vt, so that it is more "stable" | 09:57 |
alkisg | export $(/usr/share/epoptes-client/get-display) => gives you access to xorg btw | 09:58 |
KitsuWhooa | I wanted the gdb session remotely so that it's more comfortable for me to work as I will be using my desktop keyboard | 09:58 |
alkisg | (DISPLAY/XAUTHORITY) | 09:58 |
alkisg | Yeah, I know the feeling :D | 09:58 |
alkisg | I don't know why epoptes would do that though | 09:58 |
alkisg | Ah maybe systemctl status epoptes-client would show the stderr | 09:58 |
KitsuWhooa | service vboxadd stop | 09:58 |
KitsuWhooa | ...oops | 09:58 |
alkisg | Or journalctl something, for the previous crash | 09:59 |
KitsuWhooa | I'll check, yeah | 09:59 |
KitsuWhooa | is it python? | 09:59 |
KitsuWhooa | I'm used to seeing segfaults in dmesg, but if it's python it probably won't be a segfault | 09:59 |
alkisg | epoptes-client is mostly shell + socat | 10:00 |
alkisg | A couple of python scripts are ran externally, as child processes | 10:00 |
KitsuWhooa | https://tasossah.com/s/1ecf05d7e48b.jpg | 10:01 |
KitsuWhooa | might need to recompile nouveau without optimisations | 10:01 |
KitsuWhooa | but nouveau_present_flip_exec definitely has to do with double/triple buffering | 10:02 |
KitsuWhooa | alkisg: 13:24:22 ltsp162 epoptes-client[1561]: Could not detect or access the active display | 10:40 |
KitsuWhooa | 13:36:34 ltsp162 epoptes-client[1561]: 2018/09/16 13:36:34 socat[1561] E waitpid(): child 1629 exited on signal 3 | 10:40 |
KitsuWhooa | former makes sense because X is frozen, not sure why socat died 12 minutes later though | 10:41 |
alkisg | Not detecting the display shouldn't crash epoptes though | 10:43 |
KitsuWhooa | alkisg: I traced the segfault down a bit | 16:44 |
KitsuWhooa | It's an X issue | 16:44 |
KitsuWhooa | exaGetPixmapDriverPrivate() returns a 0 pointer | 16:45 |
alkisg | KitsuWhooa: is this enough info for a bug report? | 19:40 |
KitsuWhooa | Possibly, but I'm going to look more into it tomorrow. X 1.19 also has the same issue, so it happened some time before it. | 19:41 |
alkisg | Great! | 19:42 |
KitsuWhooa | I tried disabling acceleration and it still crashed, which is a bit odd | 19:42 |
KitsuWhooa | because the above gets called due to acceleration | 19:42 |
alkisg | NoAccel was the first thing I tried yeah | 19:42 |
KitsuWhooa | I also tried setting AccelMethod or whatever it is to none | 19:42 |
KitsuWhooa | and that didn't help either | 19:42 |
* alkisg has studied vga internals back in 1990, but has no idea about xorg internals :( | 19:43 | |
KitsuWhooa | I don't know much about it either to be honest. | 19:46 |
KitsuWhooa | I will however say that this whole thing helped me uncover two kernel bugs that affect my desktop, that I'll need to look into/report at some point | 19:47 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!