00:38 mlankhorst: happens all the time
09:53 karolherbst: imirkin, mupuf: would it be helpfull for the kepler thing to make a list of all piglit tests where I get SCHED_ERROR or something as bad as this?
09:54 imirkin: is it reliable?
09:54 karolherbst: I am not yet sure
09:54 karolherbst: that's whay I try to find out then
09:54 karolherbst: *what
09:54 karolherbst: just asking if its worth the effort before doing it
09:55 mupuf: karolherbst: got some hours to kill tonight, will try again finding this pwm register
09:56 mupuf: before reviewing hakzsam's patches tomorrow
09:56 mupuf: (gonna be a long day!)
09:59 karolherbst: yeah
09:59 karolherbst: nice
09:59 karolherbst: but I one thing I have to admit is done pretty nice in systemd: collections of coredumps :D
10:00 karolherbst: the compressed ones are brocken though :/
10:00 karolherbst: okay, loading nouveau
10:00 karolherbst: maybe the X server stops to crash
10:03 karolherbst: okay
10:04 karolherbst: imirkin, mupuf: Xs fault? https://gist.github.com/karolherbst/725ed4f573e6baf34927
10:05 karolherbst: I think I know whats going on there
10:08 karolherbst: yeah
10:08 karolherbst: X has the strange behaviour to always load the modesetting driver whenever a new gpu is "appearing" in the system
10:18 karolherbst: https://bugs.freedesktop.org/show_bug.cgi?id=91388
10:25 mlankhorst: karolherbst: this is expected, sasdly
10:25 karolherbst: why is that?
10:26 mlankhorst: can't really hotplug pci cards so the code assumes any newly plugged cards are kms only
10:26 karolherbst: :/
10:27 karolherbst: X should just ignore the card
10:27 karolherbst: at least in my case
10:27 karolherbst: as far as I uderstand stuff, there is no real X support needed for the nvidia card with DRI3 offloading anymore
10:28 karolherbst: that's why no DDX is required
10:28 mlankhorst: -noautoaddgpu or something
10:28 karolherbst: doesn't work
10:28 karolherbst: its in the xorg.conf
10:28 karolherbst: Option "AutoAddGPU" "false"
10:28 mlankhorst: ah
10:29 karolherbst: it works as expected when X knows there is this card at launch time
10:29 karolherbst: then no modesetting driver is loaded
10:29 karolherbst: even if the card is powered off in between
10:30 mlankhorst: dno tbh, code is probably your best reference here
10:42 karolherbst: how serious are "GPC1/TPC0/MP trap: MEM_OUT_OF_BOUNDS" ?
10:43 imirkin: usually not serious
10:43 imirkin: is it from piglit?
10:43 imirkin: there are a few array-out-of-bounds accesses in there
10:43 karolherbst: yeah
10:43 karolherbst: I am running with --dmesg :/
10:43 karolherbst: it takes its time now
10:45 karolherbst: whats the best way to tell piglit to skip the current running test?
10:45 karolherbst: or to pause?
10:46 karolherbst: now I got a "nouveau E[ PGRAPH][0000:01:00.0] DATA_ERROR [INVALID_VALUE] ch 2 [0x00bf88f000 ext_transform_f[7309]] subc 0 class 0xa097 mthd 0x2388 data 0x002de080"
10:47 imirkin: yeah
10:47 imirkin: you should look at the runs we did
10:47 imirkin: and report only things that stand out :p
10:47 karolherbst: I see
10:47 imirkin: karolherbst: http://people.freedesktop.org/~imirkin/
10:48 karolherbst: compared to the nvc0 one?
10:48 imirkin: the nvc0 one includes runs on fermi, kepler, and maxwell
10:48 imirkin: the maxwell dmesg fails are spurious
10:59 Karlton: meh, SF is down -.-
11:10 mupuf: I finally moved Arch's packages to AUR4. That is to say the nouveau-fw and envytools packages
11:38 karolherbst: imirkin: [18801/18801] crash: 1, dmesg-fail: 4, dmesg-warn: 2, fail: 202, pass: 13633, skip: 4955, warn: 4
11:38 karolherbst: maybe the full run only has the errors :/
11:39 karolherbst: imirkin: wanna the result file?
11:44 imirkin: that seems about right
11:44 imirkin: i already have a run for kepler, should be largely identical to what you have
11:45 imirkin: click on the nvc0 one -- you should see a GK106 in the comparison
11:45 karolherbst: yeah
11:46 imirkin: and the relevant results files are uploaded if you want to compare
11:46 karolherbst: yeah, downloading
11:50 karolherbst: mhh
11:50 karolherbst: shaders@point-vertex-id fails for me
11:51 karolherbst: also all spec@!opengl 1.1@read-front are failing
11:51 karolherbst: but I think that's not that important actually
11:51 imirkin: probably coz of dri3
11:51 karolherbst: yeah, most likely
11:51 imirkin: the read-front failures are def coz of dri3
11:52 imirkin: hopefully the point-vertex-id does something dodgy as well, causing the same failure
11:52 karolherbst: arb_occlusion_query passes here :)
11:52 imirkin: yeah, i fixed it
11:52 karolherbst: mhh
11:52 imirkin: and the alphatest-nocolor ones
11:52 karolherbst: it reads 0,0,0,0 instead of 1,1,1,1 or the other way around
11:52 karolherbst: yeah, right
11:52 karolherbst: and some crashes less
11:53 karolherbst: spec@ext_image_dma_buf_import@ext_image_dma_buf_import-sample_argb8888 fails
11:53 karolherbst: the xrgb one passes
11:53 imirkin: yeah, i ignore those
11:53 imirkin: falls into the category of things i don't understand
11:53 karolherbst: then some egl stuff
11:54 imirkin: i barely understand the things i claim to understand. dma-buf imports is a whole 'nother barrel of fish
11:54 karolherbst: but there seems nothign serious
11:55 karolherbst: maybe spec@arb_gpu_shader5@execution@sampler_array_indexing@fs-simple-texture-size is important?
11:55 imirkin: uhhhh
11:55 imirkin: that should pass.
11:55 karolherbst: it fails here
11:55 karolherbst: reads only 0
11:55 imirkin: i wonder if that piglit test didn't exist before
11:56 karolherbst: yeah
11:56 karolherbst: Not Run for yours
11:56 karolherbst: I updated piglit today
11:56 karolherbst: there was some new stuff
11:56 imirkin: ok yeah. fails for me too.
11:57 karolherbst: spec@!opengl 1.0@gl-1.0-swapbuffers-behavior DRI3 PRime issue?
11:57 karolherbst: and all spec@!opengl 1.1@max-texture-size have dmesg warns
11:57 karolherbst: [26146.471567] nouveau W[ DRM] skipped size 0
11:58 imirkin: feel free to track it down :)
11:58 imirkin: i skip max-texture-size
11:58 karolherbst: :D
11:58 imirkin: it causes X to die more often than i want.
11:58 karolherbst: I see
11:58 karolherbst: mhhh
11:58 karolherbst: k
11:59 imirkin: looks like i failed at the implementation of textureSize() for dynamic samplers... somehow.
11:59 karolherbst: issue in arb_transform_feedback2, egl_chromium_sync_control, egl_khr_create_context and egl_nok_texture_from_pixmap, but this is most likely on the "not that important" list
11:59 imirkin: oh right, coz it's not handled by the stupid handleTEX stuff. gr.
12:00 imirkin: well, dunno about "not that important", but def "don't care" :)
12:00 karolherbst: :D
12:00 imirkin: the arb_tf2 thing is sort of important, hakzsam is working on fixing it
12:00 karolherbst: k
12:00 karolherbst: I will never run chromium through nouveau anyway, so that chromium extensions => meh
12:00 imirkin: he has patches, but i need to review them. and i think one of them is a little wrong.
12:01 karolherbst: ohh right, have gles1 disabled, so that explain some skips
12:01 karolherbst: but overall my run is better than yours :p
12:02 imirkin: ughhh... how hard would it be for you to load up nvidia and do a trace for me?
12:02 karolherbst: arb_gpu_shader_fp64 without issues :)
12:02 karolherbst: imirkin: let me say it that way: I usually use bumblebee on a daily basis
12:02 karolherbst: traces though....
12:02 imirkin: perhaps i should have phrased it differently
12:02 karolherbst: not that fun to me at all
12:02 imirkin: would you be willing to fire up nvidia and do a mmt trace for me?
12:03 imirkin: (not mmiotrace)
12:03 karolherbst: ahhh
12:03 karolherbst: yeah then
12:03 karolherbst: envytolls?
12:03 karolherbst: *envytools?
12:03 imirkin: no... it's a valgrind thing
12:03 imirkin: valgrind-mmt
12:03 karolherbst: I see
12:03 imirkin: http://nouveau.freedesktop.org/wiki/Valgrind-mmt/
12:03 karolherbst: mhh
12:03 imirkin: there's a 'demmt' tool in envytools which processes the output
12:04 karolherbst: mhh, I don't have the mmt tool :/ so its not installed by default then
12:04 imirkin: no, it's not upstream
12:05 karolherbst: imirkin: what may also interest you: arb_occlusion_query and arb_shading_language_420pack have no fails or skips
12:05 karolherbst: ohh, the latter has four new tests
12:05 imirkin: yeah, coz i fixed the occl query
12:05 karolherbst: yeah :/
12:05 karolherbst: my short memory is sometimes too slow
12:05 imirkin: it used to count blits, but i turn it off now
12:05 karolherbst: *short-time
12:06 imirkin: it's still broken for primitives generated
12:06 imirkin: but since there's no piglit test for that, no problem ;)
12:06 karolherbst: :=
12:06 karolherbst: :)
12:06 imirkin: on the other hand, tons of tests in piglit for texture borders -- must be the most important feature in GL :)
12:06 imirkin: [for the record, texture borders are almost entirely unused, definitely not in anything semi-recent]
12:07 imirkin: anyways, if you don't want to do the trace, that's fine -- i can try to figure it out
12:07 imirkin: there are only *so* many permutations of argument order that they can come up with
12:08 karolherbst: its no problem
12:09 karolherbst: if it does compile
12:09 imirkin: // TODO: indirect resource/sampler index
12:09 imirkin: heh, i guess i missed that one
12:11 karolherbst: imirkin: some spec@egl_khr_create_context@ tests should check against gles1 support before running
12:11 imirkin: ok
12:11 imirkin: send patches :)
12:11 karolherbst: :D
12:12 karolherbst: if I do it, I may need 1 or 2 hours, a already on piglit working dev may need 5 minutes
12:12 karolherbst: but sure, when I find time to look deeper into how piglit works
12:13 imirkin: but then your next patch will go faster
12:13 karolherbst: right
12:13 karolherbst: anyway, valgrind is installed now
12:14 karolherbst: so, what should I do next?
12:14 imirkin: get nvidia going, then do
12:14 imirkin: alias mmt='~/valgrind/bin/valgrind --tool=mmt --mmt-trace-file=/dev/nvidia0 --mmt-trace-file=/dev/nvidiactl --mmt-trace-file=/dev/dri/card0 --mmt-trace-file=/dev/dri/card1 --mmt-trace-nvidia-ioctls --mmt-trace-nouveau-ioctls'
12:15 imirkin: mmt --log-file=foo.mmt bin/shader_runner tests/spec/arb_gpu_shader5/execution/sampler_array_indexing/fs-simple-texture-size.shader_test -auto
12:15 karolherbst: how can I pause piglit and resume later?
12:16 imirkin: oh, well finish your piglit run ;)
12:16 imirkin: it's not urgent
12:16 karolherbst: 10.000 tests to go
12:16 imirkin: wtf are you doing to make them go so slow?
12:16 karolherbst: --dmesg and run all
12:16 karolherbst: its already my second run
12:16 imirkin: all or just gpu.py?
12:16 karolherbst: all
12:16 imirkin: oh, that'll take forever
12:17 imirkin: there are infinity tests
12:17 karolherbst: I know
12:17 karolherbst: I usually SEGV the infinity ones
12:17 karolherbst: but I am sure I get some SCHED_ERRORs there
12:17 karolherbst: and I want to test if its always the same
12:18 imirkin: just run the same test by hand over and over
12:18 karolherbst: yeah, but I have to kow which one it is
12:18 karolherbst: never run with --dmesg before
12:18 karolherbst: at least not all
12:19 imirkin: no real point in doing it
12:19 imirkin: only the gpu.py tests interact with the driver
12:19 karolherbst: mhhh
12:19 imirkin: the rest are just compiler/etc
12:19 karolherbst: maybe the concurrent thing changed something
12:19 karolherbst: then
12:20 imirkin: yeah, concurrent tends to kill things
12:21 karolherbst: but now that you mentioned this mmt thing, couldn't we just simply trace the coolbit stuff ?
12:21 karolherbst: or won't that help us that much
12:22 imirkin: mmt traces the commands submitted by userspace to the kernel driver
12:22 imirkin: i'm not sure that the command "set clock frequency to X" will be extremely useful :p
12:22 karolherbst: :D
12:22 karolherbst: I see
12:22 karolherbst: are you sure about the "--mmt-trace-file=/dev/dri/card0 --mmt-trace-file=/dev/dri/card1" part?
12:22 karolherbst: or just card1 for me?
12:23 imirkin: that's just for nouveau
12:23 imirkin: i have it in the alias so that it always just works
12:23 imirkin: for nvidia it's /dev/nvidia0 and whatnot
12:23 karolherbst: ahh because card0 is intel here
12:23 imirkin: right, but i don't want a trace on top of nouveau
12:23 imirkin: i know exactly what nouveau does
12:23 imirkin: i want to see what nvidia does ;)
12:24 karolherbst: yeah I know
12:24 karolherbst: I was just asking
12:25 karolherbst: "PIGLIT: {"result": "pass" }" :)
12:26 karolherbst: now I have to use demmt on foo.mmt?
12:27 karolherbst: seems about right
12:27 karolherbst: imirkin: https://gist.github.com/karolherbst/4d825d8e823136a20f9a
12:27 imirkin: no, now you send me foo.mmt ;)
12:28 karolherbst: :D
12:28 imirkin: ERROR: Can't detect chipset, you need to use -m option or regenerate trace with newer mmt (> Sep 7 2014)
12:28 imirkin: hmmmm
12:28 imirkin: looks like some sort of valgrind fail
12:28 karolherbst: http://www.filebin.ca/295VCgBTFHFT/foo.mmt.xz
12:29 karolherbst: ohh
12:29 karolherbst: so its useless?
12:30 imirkin: yeah
12:30 imirkin: since you just built it, i wonder what went wrong
12:30 imirkin: what was your mmt line?
12:31 imirkin: or rather, what was the exact command you ran
12:31 karolherbst: alias mmt='../envytools/valgrind/bin/valgrind --tool=mmt --mmt-trace-file=/dev/nvidia0 --mmt-trace-file=/dev/nvidiactl --mmt-trace-file=/dev/dri/card1 --mmt-trace-nvidia-ioctls --mmt-trace-nouveau-ioctls'
12:31 karolherbst: DISPLAY=:8 mmt --log-file=foo.mmt bin/shader_runner tests/spec/arb_gpu_shader5/execution/sampler_array_indexing/fs-simple-texture-size.shader_test -auto
12:32 imirkin: hmmmmmmmm
12:32 imirkin: and you did --prefix=/path/to/envytools/valgrind ?
12:32 imirkin: and then you did "make install"?
12:32 karolherbst: yeah
12:32 karolherbst: envytools/valgrind is full of valgrind stuff
12:33 karolherbst: and the directory wasn't there a hour ago :)
12:33 imirkin: sure
12:33 imirkin: just wanted to ensure you had actually done a "make install"
12:33 karolherbst: otherwise I would get a command not found
12:34 imirkin: well, you could also be executing it from the build location
12:34 imirkin: which is less likely to work
12:34 karolherbst: I was inside piglits directory
12:36 karolherbst: imirkin: how does mmt check for the chipset? maybe the DISPLAY thing is a bit confusing
12:36 karolherbst: or shouldn't it matter?
12:36 imirkin: it's in the ioctl stream
12:36 karolherbst: okay
12:36 imirkin: or should be
12:36 imirkin: it's also having trouble resolving the maps
12:36 karolherbst: maybe driver too new?
12:36 imirkin: yeah could be
12:36 karolherbst: I can go to 304.125 if you want
12:37 karolherbst: :D
12:37 karolherbst: but which version should be fine?
12:37 imirkin: 340.x should work
12:37 karolherbst: k
12:37 karolherbst: 340.76 it is
12:39 karolherbst: mhhh
12:39 karolherbst: yeah right
12:39 karolherbst: kernel
12:40 imirkin: ugh.
12:42 karolherbst: maybe 346.82 works
12:42 karolherbst: should work, because all the other complained about too new kernel
12:45 karolherbst: okay nice
12:46 karolherbst: okay, looks much better
12:46 imirkin: filebin?
12:46 karolherbst: imirkin: http://www.filebin.ca/295afR5lB6e0/foo.mmt.xz
12:47 karolherbst: will check if 349.16 works
12:47 karolherbst: I used 352.21 for the first one
12:47 imirkin: awesome, this is precisely what i needed
12:48 imirkin: naturally the stupid thing sets some unknown bit. ugh.
12:51 karolherbst: imirkin: 349.16 works, too
12:53 karolherbst: and with 352.21 it fails again
12:53 imirkin: makes sense
12:53 imirkin: probably changed something in the 350 series
12:54 imirkin: perhaps joi will want to take a look... the new mmt stuff is his baby
12:55 karolherbst: ......
12:56 karolherbst: I don'T think the number means that much
12:56 karolherbst: if you look at them: 325, ..., 331, 334, 337, 340, 343, 346, 349, 352 ...
12:56 imirkin: i think it does
12:56 imirkin: yeah, and the 330 series changed a ton of stuff
12:57 karolherbst: mhh
12:57 imirkin: as did the 340 series
12:57 karolherbst: but why always +3?
12:57 imirkin: i mean, internally
12:57 imirkin: the others are probably alpha/beta releases, dunno
12:57 karolherbst: mhh there is also a 346, and 352 release version
12:58 karolherbst: but with a higher minor umber
12:58 karolherbst: *number
13:03 karolherbst: imirkin: strangly mmt couldn't fetch anything from nvidia-settings
13:04 imirkin: not strange at all
13:04 imirkin: it just sends commands over NV-GLX
13:04 imirkin: to the X server
13:04 imirkin: which in turn communicates with the blob driver
13:04 imirkin: but all it'll say is things like "set clock to X" or something
13:04 imirkin: it won't have anything interesting in there
13:05 karolherbst: okay
13:06 Wolf480pl: Hello. I have an NVE4 / GK104 (GTX 770M) in an Optimus setup, and I've been experiencing freezes a couple seconds after starting Xorg. Kernel 4.1.2
13:06 imirkin: Wolf480pl: pastebin dmesg and xorg logs, preferably after the freezes happen
13:12 mupuf: imirkin: the latest version of the blob compiles nicely on reator
13:13 mupuf: albeit you need to use the env var IGNORE_CC_MISMATCH=1 now because I updated to gcc 5.2
13:13 imirkin: mupuf: hm ok.
13:13 imirkin: i was trying it with 340.x iirc
13:13 mupuf: I updated the kernel
13:13 mupuf: so I may have fixed everything
13:14 mupuf: I compiled linux 4.1.2, since I do not think OOT nouveau supports1 4.2 yet
13:15 mupuf: anyway, do I need to keep the nv50 in?
13:15 mupuf: it is running quite hot and it is quite noisy
13:15 imirkin: mupuf: well, the latest blob version to support nv50 is 340.x
13:15 imirkin: i was hoping to trace it =/
13:16 mupuf: oh
13:16 mupuf: let's try that then
13:18 mupuf: imirkin: https://projects.archlinux.org/svntogit/packages.git/tree/trunk/PKGBUILD?h=packages/nvidia-340xx <-- this is how you are supposed to compile it
13:18 Wolf480pl: imirkin, it will be hard for me to get logs after a freeze, because once a freeze occurrs, there's no reaction to pings or magic sysrqs, but I'll try.
13:18 mupuf: the patches are here: https://projects.archlinux.org/svntogit/packages.git/tree/trunk?h=packages/nvidia-340xx
13:19 imirkin: Wolf480pl: ok, well that sucks. can you try getting netconsole up and running?
13:19 imirkin: mupuf: could i convince you to get that going?
13:19 Wolf480pl: I can try. Does it work well over wifi?
13:19 mupuf: yes, already doing it
13:19 imirkin: Wolf480pl: it works better over wired :)
13:19 Wolf480pl: ok
13:19 imirkin: it's udp, so there's no retry or anything
13:20 imirkin: also wifi can involve various insanity to send, which isn't great when you're oops'ing
13:20 Wolf480pl: btw. I noticed that if I set runpm=0 there's a different freeze, which happens after exitting Xorg, and can be recovered from by a SAK (sysrq-K)
13:22 imirkin: can you pastebin a boot with that?
13:22 imirkin: i wonder if you're getting the HUB_INIT timeout issue
13:22 mupuf: imirkin: done, it is /root/blob/NVIDIA-Linux-x86_64-340.76-no-compat32
13:22 Wolf480pl: yeah, I'm getting something like that
13:22 imirkin: mupuf: awesome
13:22 mupuf: kernel/nvidia.ko is there
13:23 imirkin: i'm fighting with txq right now, but i'll take a look at it later today
13:23 mupuf: ok! Shutting down the machine in the mean time
13:26 imirkin: ok, well found the bug, but naturally it doesn't work. gr.
13:26 imirkin: something on fermi is different. OF COURSE. why wouldn't it be.
13:33 Wolf480pl: imirkin, here's my dmesg log: http://pastebin.com/ExaFHC1V and Xorg log: http://pastebin.com/W1TDTvc7
13:34 imirkin: [ 45.302195] nouveau E[ PGRAPH][0000:01:00.0] HUB_INIT timed out
13:34 Wolf480pl: yeah
13:34 imirkin: Wolf480pl: with runpm=0, you can try http://cgit.freedesktop.org/~darktama/nouveau/log/?h=hack-gk106m
13:35 imirkin: Wolf480pl: note that this is an out-of-tree module, it should build just fine against your 4.1.x tree though
13:36 Wolf480pl: how do I build it?
13:37 Wolf480pl: just make?
13:42 imirkin: Wolf480pl: cd drm; make
13:42 imirkin: karolherbst: mind seeing if this fixes the texture size thing? http://paste.debian.net/283686/
13:51 karolherbst: imirkin: run the tests like with mmt then?
13:52 karolherbst: or better all
13:52 imirkin: no need for mmt
13:52 imirkin: but yeah, run just that one test
13:52 karolherbst: k
13:55 Wolf480pl: imirkin, could you please confirm the commit sha1 of that branch is 64b75967fe13499f12876ec10a31d9141e828714 ?
13:56 imirkin: yes
13:56 Wolf480pl: ok, thanks
14:04 Wolf480pl: do I then put the drm/nouveau/nouveau.ko in the extramodules dir, or somewhere else?
14:05 Wolf480pl: or do I just insmod it?
14:07 imirkin: just insmod
14:07 Wolf480pl: do I have to first rmmod the old one?
14:11 Wolf480pl: ok, I did `rmmod nouveau; insmod nouveau.ko` and it immediately printed HUB_INIT timed out
14:12 imirkin: can you try it without loading the "old" nouveau first?
14:12 Wolf480pl: yeah, I can blacklist it, and then load the new one
14:15 Wolf480pl: weird... insmod: ERROR: could not insert module nouveau.ko: Unknown symbol in module
14:15 karolherbst: ttm
14:15 imirkin: you need to insert its deps
14:15 imirkin: modinfo nouveau should tell you what they are
14:16 Wolf480pl: oh, right
14:16 karolherbst: imirkin: pass
14:16 imirkin: karolherbst: yay!
14:16 imirkin: ok, i'll push... probably won't work on maxwell but... wtvr
14:17 karolherbst: will run gpu.py piglit though
14:18 karolherbst: imirkin: shaders@glsl-fs-lots-of-tex fail sounds strange
14:20 karolherbst: colors in piglit are in rgb format?
14:20 karolherbst: mhh, less red in their
14:20 karolherbst: than expected
14:20 karolherbst: *there
14:24 Wolf480pl: imirkin, with this kernel module and runpm=0 I could start Xorg and exit it, w/o a freeze. Should I now try without runpm=0 or, should I try xrandr --setprovideroffloadsink ?
14:25 Wolf480pl: (because so far Xorg was running on Intel as primary GPU)
14:25 imirkin: Wolf480pl: it seems like runpm=0 is required with that hack branch
14:25 imirkin: Wolf480pl: setting the offload sink/etc stuff should work though
14:25 imirkin: karolherbst: that failure is expected.
14:25 karolherbst: ohh
14:25 imirkin: i have a local patch to fix up the test
14:28 Wolf480pl: `DRI_PRIME=1 glxinfo` says "OpenGL vendor string: nouveau", but DRI_PRIME=1 glxgears is all black
14:29 Wolf480pl: oh, I might need a compositing WM, is that right?
14:30 imirkin: yeah
14:31 Wolf480pl: anyway, thanks, see you later
14:51 karolherbst: imirkin: glean-fbo passes now
14:52 karolherbst: I have 3 more passes since the last run
14:53 imirkin: and i pushed the txq change
14:53 karolherbst: yeah
14:53 karolherbst: don't know when I last installed mesa though
14:53 imirkin: well, i pushed it like 5 mins ago :)
14:54 karolherbst: okay, 2 hours
14:54 karolherbst: allthough wait
14:55 karolherbst: last merge was at 7/17/2015, 11:22:50 PM GMT+2:00 DST
14:55 karolherbst: so since then and now 3 more tests passes
14:56 karolherbst: obviously fs-simple-texture-size passes now
14:58 karolherbst: mhh
14:58 karolherbst: imirkin: spec@egl_chromium_sync_control@conformance
14:58 karolherbst: looks like something random
14:58 imirkin: yeah, that's in the "don't care" category of things
14:58 imirkin: anything with egl, really
15:04 karolherbst: mupuf: how is it going?
15:15 karolherbst: right somebody wanted me to check fear2 with nouveau
15:15 karolherbst: not here anymore ./