15:16 karolherbst: imirkin: it seems like nv30 regressed :O https://gitlab.freedesktop.org/mesa/mesa/-/issues/5061
15:16 karolherbst: but last time I tried it was fine...
18:40 reinist12: I delt with the building system on cocoapods ruby not swift version and things started to compile for my phones on IOs, those days i anyways do not have nvidia card available, only at home on my desktop, but cracking the firmware signatures for PM should not be hard, buck or bezel i later perhaps migrate to, but they required java which i did not bother building for IOs, no package was available. I am dealing with sw programming on apps nowdays for
18:40 reinist12: mobile.
18:40 reinist12: And some other areas for optimization and data mining like bitcoin. the nvidia pm firmware is not something i would still fancy to be touched though, but it can be provided anyhow still. There is a lot better method than ever using this.
18:40 reinist12: I think i am doing my development all with android studio in the future my own.
18:43 reinist12: As the ARM was bought by NVIDIA, there are new socs with nvidia cards, which i may reach to soon.
18:43 reinist12: And i may be dealing with one of those, seems like quadro or such families of integrated cards on mediatek and other socs are sometimes available.
18:44 karolherbst: imirkin: actually.. let me check on my nv30 card again
18:47 karolherbst: imirkin: what's weird.. that my desktop can even display the UEFI splash on it :D
18:48 karolherbst: imirkin: uhhh.. gnome has artifacts :/
18:48 karolherbst: Mesa 21.1.4
18:48 karolherbst: ehhhh
18:49 karolherbst: shit
18:49 karolherbst: that deserves a video
18:56 reinist12: so my days are quite fully occupied cause i manage all my own in my own company, there is no time to other stupid things at all anymore. like arguing or whatever which i would not care about, cheers.
18:58 karolherbst: imirkin: https://drive.google.com/file/d/1XrCObCHNlOxmZnq1mEsY9EyUtZrWHyuo/view?usp=sharing :)
18:59 karolherbst: ehhh..
18:59 karolherbst: wait
18:59 karolherbst: now it should work
19:01 karolherbst: well...
19:01 karolherbst: long story short: it's busted
19:08 Lyude: is there any context to what reinist12 said? I only ask because if there isn't, what just happened is extremely funny lol
19:08 karolherbst: Lyude: I think you know who that person is
19:09 Lyude: OH it's them. huh
19:09 karolherbst: well.. I think
19:09 Lyude: that's a lot more coherent then I expected
19:09 Lyude: yeah I wasn't totally sure
19:09 karolherbst: might be another one :O
19:09 Lyude: I'm going to assume it was another one just because the idea of someone joining, bragging about data analysis and bitcoin and then saying they're too important for this, is very funny to me lol
19:09 karolherbst: ohhhh wait...
19:10 karolherbst: we had this weirdo $hitcoin person at some point
19:10 Lyude: lol, strange
19:10 karolherbst: but that person was selling modded drivers/vbios/whatever to get 10% more mining speed or so
19:11 karolherbst: asked tons of questions
19:12 karolherbst: but I think that person actually used insider information as this was the same one giving us a "signing tool" for modded vbios :O
19:12 karolherbst: via a website
19:12 Lyude: ooooooooooooh. that's, extremely sketchy
19:13 karolherbst: :D really
19:13 karolherbst: it was even under the official nvidia domain or something
19:13 Lyude: sounds like at some point someone is probably going to be losing their NDA privileges with nvidia
19:13 karolherbst: anyway.. at some point nvidia detected it
19:13 karolherbst: yeah.. it was years ago
19:13 karolherbst: like 2 or 3 or so
19:17 Lyude: yeah I vaguely remember it now actually
19:25 reinist12: I am pretending to answer the last time :), it's not funny to waste time, and getting into cases where no coordination works, things are delayed in teams where, along comes better efficiency when it is still really possible , i did not brag but it needs to be used the chanche until that chanche is still there, later all that would be allready gone, so i check out this point, i missed the part of bragging and or having mentioned my importance though.
19:25 Lyude: who even are you
19:26 reinist12: Cause i am always very low profile and probably will be in the future, cause this is the best for me.
19:26 reinist12: Mart of course.
19:26 Lyude: who
19:26 Lyude: do you work for nvidia or something
19:26 Lyude: OH
19:26 Lyude: ok, thanks for coming mart
19:34 Lyude: hm
19:39 Lyude: you know, something just occurred to me. we have mode +ntM set, which means that the only way he could have been talking is if he registered his nickname
19:39 Lyude: this may be useful :)
19:39 karolherbst: well
19:40 karolherbst: I think mart also has like 100 VPN accounts
19:41 Lyude: well, that's the thing. the only way you can talk in this channel with +M set is if you register an account, and I'm doing akills on him which means I ban via his nickserv registration and not just his host or nickname (that's not the bans you saw just being added, but chanserv will keep adding more for anything that matches his nickname reg)
19:44 karolherbst: yeah...
19:44 karolherbst: let's see how that turns out
19:45 Lyude: mhm, this gives me some fun ideas as well
19:45 karolherbst: but I think it's just something we always have to deal with :/
19:45 karolherbst: but honestly, as long as the threats are kept to a minimum I am not caring all that much
19:46 Lyude: yeah, it doesn't bother me much either. i just used to help run an IRC network, so I've seen all sorts of tricks get used to make it really hard for folks to spam
19:47 Lyude: so if I get bored on some weekend I might look into some of the ideas I've got
19:59 Lyude: hey sarnold !
19:59 sarnold: hey Lyude :)
20:00 ccr: wasn't this channel (or at least some of the other fdo channels) similarly +M (or whatever the equivalent mode was) on freenode .. didn't really stop him. I think at least #intel-gfx was.
20:00 Lyude: ccr: yeah - I noticed he's been registering his nicks
20:00 ccr: though admittably the user registration process on OFTC is different
20:01 Lyude: which is how he's been getting past this
20:01 Lyude: might also definitely explain why there's quite a bit of time between when he comes back for the most part
20:01 Lyude: "might also definitely" is a word salad oops but you know what I mean
20:02 ccr: :)
20:02 ccr: possibly probably!
20:02 Lyude: sarnold: anyway I need to get back to focusing on work but I'll still be around, feel free to highlight me if there's any more info you need
20:29 raket: Is there something wrong with mesa-21.1.4 ? i'm getting strange error like "[xcb] Unknown sequence number while processing queue [xcb] Most likely this is a multi-threaded client and XInitThreads has not been called [xcb] Aborting, sorry about that. ezquake-linux-x86_64: xcb_io.c:269: poll_for_event: Assertion `!xcb_xlib_threads_sequence_lost' failed. Received signal 6, exiting... DOUBLE SIGNAL FAULT:
20:29 raket: Received signal 11, exiting.. ^CDOUBLE SIGNAL FAULT: Received signal 2, exiting..." reverting to mesa-20.3.4 fixes it, seems to be nouveau only
20:36 Lyude: o.O
20:36 Lyude: that's real strange
20:37 Lyude: actually, my guess is that we got some mt threaded stuff in and did something wrong. karolherbst ^
20:37 Lyude: sounds like the fix shouldn't be too difficult, assuming it is just a missing XInitThreads call
20:37 raket: https://www.linuxquestions.org/questions/slackware-14/problem-with-x-mesa-21-1-4-x86_64-1-txz-4175697476/
20:38 raket: It really doesn't matter, but would be cool if it works =), (reverting to old mesa isn't really a big deal)
20:38 karolherbst: Lyude: we didn't :)
20:38 karolherbst: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10752
20:38 Lyude: damn, bug goes back to strange
20:38 Lyude: raket: I mean we should definitely fix it lol
20:38 Lyude: raket: think you could bisect it?
20:39 raket: Lyude: what's bisect?
20:39 Lyude: raket: git bisect, it's a feature of git that lets you test a couple of commits from the development histotry for something to try reproducing a bug, then you give it feedback and it figures out what commit introduced the bug
20:40 ccr: e.g. method of tracking down which change in code "caused" the issue to manifest
20:40 raket: from mesa? :)
20:41 Lyude: raket: yeah. if you don't feel comfortable doing that that's probably fine though, this bug looks like it's a lot easier to track down then your 2k issue
20:41 Lyude: raket: assuming easyquake is just some quake-like game I can get for free?
20:41 raket: Lyude: git clone https://github.com/ezquake/ezquake-source
20:41 ccr: .. you need the Quake 1 datafiles tho, I assume
20:42 raket: nquake.com :D
20:42 ccr: raket, what's your NVidia GPU you're hitting this on?
20:42 raket: ccr: NV120
20:43 raket: and yeah, for the record i won 4 tournaments in a row with nouveau+nv120, reclocked -bah- with mounted fans :D
20:43 ccr: I see.
20:43 Lyude: raket: I guess lemme know if you need me to do the bisect, although it might take me a day or two to get too (luckily my plate is finally clearing up for the time being…)
20:43 Lyude: raket: oh nice!
20:44 raket: Lyude: i could do the bisect myself, but i need instruction how to do it. ciscon: wake up, let's track it down!
20:44 ccr: closest I have to that is a GTX 750 aka NV117/GM107
20:44 Lyude: raket: do you know how to build mesa from source?
20:46 raket: Lyude: slackware comes with a build file, i could try that and jut git clone mesa
20:47 Lyude: raket: ahh cool that should work. that also saves me having to explain how to convince applications to load a different mesa build
20:48 raket: how do i do the biselct stuff
20:48 raket: (point me to the manual, i'll rtfm)
20:48 Lyude: oh cool - https://git-scm.com/docs/git-bisect
20:48 Lyude: thanks for helping btw! I also asked about this in dri-devel to see if anyone's already working on this issue
20:49 raket: helping! it needs to be fixed! i don't want ugly llvm-11.x so-files laying around in /opt/lib64 and using old mesa files just for the sake of old mesa :)
20:50 Lyude: hehe
21:02 Lyude: raket: btw - would also definitely recommend reporting this on the gitlab for mesa
21:37 raket: Lyude: problem is that mesa-20.3.5 won't compile with llvm 12.0.1, is there a way to use gdb to track it down?
21:37 raket: or should i revert to llvm-11.x?
21:37 Lyude: raket: I would go to an older version of llvm, the llvm api changes a ton so that's not too surprising
21:39 raket: Lyude: i tried all the mesa on https://slackware.uk/cumulative/slackware64-current/slackware64/x/ back to 21.1.0 and they all seem to have the problem. 20.3.4 does not... wonder if the llvm 11.x with mesa 20.3.5 works...
21:39 raket: *installs llvm 11.x*
21:39 Lyude: raket: any luck with the bisect yet?
21:39 Lyude: or is that what this is from
21:40 Lyude: actually - I just also realized that z (as in X.Y.Z) releases would diverge from the main branch, so you might actually want to bisect between 21.1 and 20.3 instead of 21.1.0 and 20.3.4
21:41 raket: yeah.. i will try with llvm-11.x and try to compile 20.3.5 and see if it works
21:41 Lyude: otherwise git might get confused and give you the wrong bisect result
21:42 raket: if the error is between 20.3.4 and 20.3.5 (i guess it is) we'll catch it
21:45 Lyude: raket: mhm - still good to bisect between those two afterwards so we can figure out exactly what commit did it
21:46 raket: Lyude: ofcourse i will, however, if 20.3.4 works and 20.3.5 it's easier to find it out - 21.x seems to need llvm 12.x
21:48 raket: (21.x won't compile at all with llvm-11.1)
21:50 ccr: nouveau does not need llvm, so if you limit the drivers to be built, you'd have faster build and not need to fiddle with the llvm dependency
21:51 raket: which DRI_DRIVERS and GALLIUM_DRIVERS should be built then? (-Ddri-drivers=$DRI_DRIVER -Dgallium-drivers=$GALLIUM_DRIVERS)
21:51 raket: just nouveau?
21:52 ccr: well, unless you need something else
21:53 ccr: nouveau is a gallium driver so -Ddri-drivers= -Dgallium-drivers=nouveau
21:54 ccr: depending on Meson version you may need to use -Ddri-drivers=[] to signify empty list
21:58 raket: ccr: oh.. interesting, could you post a "meson setup \ --opt1=blabla \ --opt2=blabla \ --optlast=blabla" to build it?
21:58 RSpliet: raket: that bug report mentions that it also happens with amdgpu. That points towards core mesa/gallium no?
21:59 raket: RSpliet: if you want a i can connect a rx550 to double-check if that's true, just tell me and i will
21:59 RSpliet: raket: just thinking of all the ways we could be barking up the wrong tree
22:00 ccr: raket, well .. this is my custom build script I use https://tnsp.org/~ccr/mesabuild.sh but it also enables few other drivers which I use/test
22:00 RSpliet: Don't take my word on it without double-checking, but I don't recall nouveau using llvm for anything except SPIR-V and maybe maybe OpenCL?
22:01 ccr: yeah, I don't use/enable llvm in my builds
22:04 raket: ccr: so dri-drivers doesn't need nouveau?
22:05 RSpliet: And if you say you can't reproduce on Fedora... I don't think they do much downstream patching at all these days, it's just regular upstream maybe with some code axed out that may be contentious in their eyes legal-wise.
22:08 RSpliet: In fact, I don't see any patches in the srpm, so presumably it's just vanilla upstream
22:18 Lyude: mesa-dri-drivers contains the nouveau dri driver. also what RSpliet said - we typically limit downstream patches to either fixes that are pending upstream already and just haven't landed in a release, something that upstream doesn't think would be useful for them, or _very_ exceptional cases of upstream being difficult. I've only ever seen like, one example of the last one though.
22:33 ccr: raket, no. there exists the very old "nouveau" that was not based on gallium, and is a mesa/dri driver, but you probably don't want that.
22:39 RSpliet: that's called nouveau_vieux, it's part of upstream mesa and packaged by most distros, and it works only for cards from before NV50/2006
23:04 karolherbst: RSpliet: nope
23:04 karolherbst: before nv30 :)
23:04 karolherbst: we have a nv30 driver in gallium
23:04 karolherbst: or can vieux handle those as well?
23:15 ccr:has no idea, but vieux only seems to have source files named up to nv20