17:12 cosurgi: imirkin: it crashed again.
17:15 cosurgi: imirkin: that's an older xserver, so it was compiled with -O2, so debug info isn't that good, but here it is: https://paste.ubuntu.com/p/7JxW9QSHP3/
17:16 cosurgi: imirkin: now I have started this particular xserver with driver compiled with -g -O0
17:16 cosurgi: (gdb) p *bo
17:16 cosurgi: $1 = {device = 0x564c4ecaf880, handle = 16, size = 100663296, flags = 3221225473, offset = 4194304, map = 0x7fd168113000, config = {nv04 = {surf_flags = 254, surf_pitch = 64}, nv50 = {memtype = 254,
17:16 cosurgi: tile_mode = 64}, nvc0 = {memtype = 254, tile_mode = 64}, data = {254, 64, 0, 0, 0, 0, 0, 0}}}
17:18 imirkin: yeah, i mean i think i determined last time that what the code was doing seemed legit
17:19 imirkin: so why it's getting a SIGBUS is a bit unclear
17:19 imirkin: the memory access appears to be within bounds
17:20 cosurgi: I have dmesg and Xorg log too. Do you want to see them?
17:22 imirkin: nope
17:23 cosurgi: ok. If tat seemd legit, can we confirm with some extra checks in code if its 100% legit? :)
17:25 cosurgi: imirkin: btw, this time the only thing I did was move mouse a little bit. And CPUs were all under high load (I was compiling https://yade-dem.org/doc/ in 10 chroots simultaneously)
17:30 cosurgi: imirkin: will get more useful info from gdb running no `-g -O0` ?
17:30 cosurgi:hopes so :)
17:34 imirkin: hopefully!
17:36 cosurgi: s/running no/running *on*/ :)
18:02 imirkin: skeggsb: btw, you were going to look at my 1024 lut change "on monday" ... i suspect last monday? or the monday before that? how'd that go?
21:20 HdkR: cosurgi: Could be a memory alignmnet thing?
21:21 HdkR: `info regs` and `disas $pc,+16` would have given a bit more information if it was an alignment failure
21:31 cosurgi: HdkR: thanks for the suggestion! I have stopped that gdb session. But no worries! I will try this next time. All six of my Xorg instances are connected to gdb. Expect next crash anytime ;) On average it's once per week.
21:32 cosurgi: two of these xsessions are with `-g -O0`, four with `-g -O2` compiler flags.
21:32 HdkR: wow
21:32 imirkin: HdkR: he's not messing around. but also doesn't really know how to use gdb.
21:33 imirkin: HdkR: he has 3x 4k rotated screens connected ... not exactly the garden-variety setup
21:33 cosurgi: Always when I finish my work for the day I switch to VT 12 (ctrl-alt-12), because crashes very often happen when I wake up from `xset dpms force off` in the morning.
21:34 HdkR: imirkin: Even I wouldn't work this hard on the problem :P
21:34 cosurgi: heheh. I'm learning gdb slowly ;) For 15 years I was used to LOG_DEBUG, std::cerr type of debugging. I tried to learn gdb in 1999, but it was too crappy.
21:34 cosurgi: Now I see it has improved a lot!
21:34 cosurgi: but still I don't know it :>
21:36 HdkR: Would be nice if x86 was cool enough to pass through the alignment fault rather than changing it to sigbus
21:36 HdkR: ARM passes it through and it is super nice
21:36 cosurgi: HdkR: imirkin already diagnosed me with multiple personality disorder. Because I use each xsession for different topic of my work :)
21:36 cosurgi: and of course separate one for gaming
21:36 HdkR: That's actually pretty neat
21:37 HdkR: I just lazily mix everything together and it can become a hassle at times
21:38 cosurgi: in fact I have several different user accounts like: devel, science, work, game. And they all belong to group cosurgi. There is no user cosurgi ! Instead /home/cosurgi is a git repository with my documents. And each of the user's dirs, like /home/devel has onlu git with .dotfiles :)
21:39 cosurgi: And I have prepared a small zsh plugin which shows a `git status` for three repos (documents,dotfiles,current git) in the status line: https://gitlab.com/cosurgi/zsh-git-cal-status-cpp :)
21:39 cosurgi: Like on the status line on this screenshot: https://gitlab.com/cosurgi/zsh-git-cal-status-cpp/raw/master/git-status-scr2.png
21:42 HdkR: Neat
21:42 cosurgi: HdkR: I was actually forced to do this split into several xsessions because sometimes when I run calculations, I'm very lazy and I start each one in separate xterm (instead of writing a nice shell script for that). And each of them opens several windows. And pretty often during "high pressure times" (e.g. right before a confrence) I hit a hrdcoded xserver limit of 256 windows.
21:42 HdkR: Whoa, didn't even know that was a limit
21:43 cosurgi: a nice side effect is that now an xserver crash only kills one of my working topics :)
21:43 HdkR: I only have three windows open atm :P
21:44 HdkR: I may get up to...nine when working
21:44 cosurgi: yeah. Once I did recompile Xorg only to increase it to 512 windows. But then I discovered that okular is leaking the xresources: it consumes one slot per refresh after latex recompile. A day or two after working on .tex in gvim and several hundresds of recompilations I have to kill okular :)
21:45 HdkR: huh
21:45 cosurgi: Because opening even a single window is not possible. Not even an xterm. I have to kill okular from somewhere ssh or from an already opened terminal.
21:52 cosurgi: yeah sorry. I don't actively learn gdb.
21:53 HdkR: It's fine, most people don't need to
22:16 cosurgi: ok, goodnight! :)