00:40krh: I have a bind-mounted lib dir from the chromeos host fs inside the android container
00:40krh: I can touch ext4-vendor/lib/egl/foo from host chromeos
00:40krh: and it shows up inside arc++
00:40krh: I did something today!
00:41krh: oh hiu
00:41krh: progress here
00:43imirkin: a day well-spent
00:44krh: who doesn't love fiddling with filesystems and mount options
00:44imirkin: only people who prefer fiddling with nfs
06:10airlied: mareko: so one reason llvmpipe doesn't use your cpu (https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4385)
06:35daniels: robclark: yeah, any help in figuring out what exactly is going on would be very welcome, because it's quite wtf
07:56danvet: sravn, wanted to suggest to also add a link from drm_writeback_connector to these two hooks
07:56danvet: but that one doesn't even have a proper kerneldoc :-/
07:57danvet: maybe add these two if you have time
07:57danvet:looks sternly at pinchartl
07:58danvet: oh drm_writeback.h isn't even included in the template
08:02sravn: danvet, pinchartl: I will try adding drm_writeback.h to kernel-doc and link them as suggested. And see if it the results looks readable..
08:04sravn: danvet, pinchartl: And will try to dig up with something to describe drm_writeback_connector
08:32cwabbott: daniels: is this a known problem: https://gitlab.freedesktop.org/cwabbott0/mesa/-/jobs/2125508
08:42daniels: cwabbott: no, that's a new one on me. fixed it up and will try to figure out why on earth it happened. perhaps you're just special?
08:44cwabbott: I think I opened the MR before the Windows stuff landed, would that have anything to do with it?
08:46danvet: airlied, [PATCH 5/6] drm/amdgpu: add support for exporting VRAM using DMA-buf v3 <- might want to ack that, or entire series
08:46danvet: it's the patch with the page-less sg trick
08:46danvet: agd5f_, ^^ I just acked the thing
08:50daniels: cwabbott: i can't imagine why that particular error would happen tbh, apart from some kind of amazing corruption (seems to be neither local nor remote), or you were super unlucky and cloned when we had not enough disk space
08:51cwabbott: I don't think it's a coincidence, something similar happened with my other MR https://gitlab.freedesktop.org/cwabbott0/mesa/-/jobs/2115632
08:52cwabbott: (the freedreno failures are legit, I'm working on those now)
08:52daniels: yeah, you share the same git clone cache
08:58cwabbott: ah, that explains why it happened again then
09:00cwabbott: daniels: unfortunately this happened now: https://gitlab.freedesktop.org/cwabbott0/mesa/-/jobs/2126332
09:01cwabbott: I just restarted it
09:01daniels: cwabbott: that would'be been me stopping the service to fix it for good, yeah
09:12airlied: danvet: will take a look tomorrow, though your ack is good enough start!
09:12tomba: bbrezillon: how do I get the drm_bus_cfg for the first bridge from the encoder's atomic funcs? I'm trying to figure out how to get the bus_format with Laurent's DRM_BRIDGE_ATTACH_NO_CONNECTOR model.
09:34ernstp: A lot of launchpad builders seem to be stuck...
09:35ernstp: 4 hours here for example: https://launchpad.net/~oibaf/+archive/ubuntu/graphics-drivers/+build/19090757
09:36ernstp: ah there is an actual launchpad channel!
09:52daniels: ernstp: we have no insight into Canonical's infrastructure, sorry
10:18ernstp: daniels: of course, sorry, I really thought I was posting in ubuntu-devel :-D oops
11:50gpoo_: craftyguy: wrt https://gitlab.freedesktop.org/Mesa_CI/mesa_jenkins/-/merge_requests/1319 at Igalia we were using our Intel CI. Is it possible an oversight when removing it?
13:33bbrezillon: tomba: you can't. Actually the plan is to replace encoder implementations by bridge ones
13:34bbrezillon: tomba: that means your encoder driver has to implement the drm_bride interface instead of the encoder one
13:36tomba: bbrezillon: instead of, or together with? don't we still need the encoder entity, even if it doesn't do much?
13:36tomba: or would the framework create and manage a dummy encoder
13:37bbrezillon: we need the encoder entity, not the encoder implementation (AKA encoder_helper_funcs)
13:38bbrezillon: tomba: see https://patchwork.kernel.org/patch/11347209/
13:39tomba: bbrezillon: I see. thanks! I'll check that patch and see if I can get tidss working with that model.
13:40bbrezillon: tomba: the long term plan is to do what pinchartl did for connectors
14:20shadeslayer: ernstp: #launchpad would probably be better, http://launchpad.net/builders indicates huge queus
14:21shadeslayer: ( I guess they're rebuilding the whole archive )
14:54MrCooper: daniels: you, sir, have outdone yourself with https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4373#note_452236
15:21ajax: i wouldn't say the x server is unmaintainable, but otherwise sure.
15:23daniels: ajax: it's possible to maintain, less possible to sensibly build new functionality on top of
15:24ajax: certainly that
15:24ajax: analogy might have been closer for something like xorg 6
15:25ajax: (.7 or .8 or .9)
15:48craftyguy: gpoo: I will re-enable it. the last time the "Igalia" job was triggered in the Intel CI was March 2019, so I disabled it
15:53gpoo: craftyguy: that is strange, I am sure we used this month.
15:53gpoo: craftyguy: thanks!
15:54craftyguy: gpoo, jasuarez: i replied to the issue that jasuarez opened in the mesa_jenkins gitlab, can you take a look? https://gitlab.freedesktop.org/Mesa_CI/mesa_jenkins/-/issues/45
15:55craftyguy: yeah I don't see any jobs that the CI did for the 'igalia' job in recent history, nothing showing on the result site either: https://mesa-ci.01.org/
16:10Hukka: I have a problem with ice lake graphics and mpv/mplayer. Before they got stuck if I didn't immediately fullscreen the output, but nowadays mpv just seems to segfault in iris_dri.so. I don't seem to have debug symbols, so the trace shows just ??. Building mesa with debug symbols seems a bit daunting, but I suppose it's a must to make a decent bug report?
16:10jekstrand: Hukka: It really helps
16:11jekstrand: Hukka: Your distro may have debug symbol packages you can install
16:12Hukka: I'm running latest ubuntu. Actually with oibaf ppa drivers, to test if mesa 20 would have this fixed. I'm now reverting back to mainline to see if it has the debug symbols. There doesn't seem to be a separate -dbg package, at least
16:15dj-death: Hukka: sure you installed libgl1-mesa-dri-dbgsym ?
16:17gpoo: craftyguy: thanks!
16:19Hukka: Hm, I couldn't get apt to see that package by searching for dbg, but http://ppa.launchpad.net/oibaf/graphics-drivers/ubuntu/pool/main/m/mesa/ seems to have that. I'll try again
16:20Hukka: Hmh, it isn't installable
16:24shadeslayer: hey, is there a convenience function to convert a uint32_t to char* ?
16:25shadeslayer: ( in the linux kernel )
16:25imirkin: like ksnprintf?
16:25imirkin: not sure what you mean by "convert"
16:26shadeslayer: imirkin: I want to call debugfs_create_dir such that I get a /sys/kernel/debugfs/bo/<bo handle> entry
16:26imirkin: yeah, one of the sprintf variants sounds good then
16:27imirkin: esp one which handles buffer overflows
16:27shadeslayer: ^^ that's what I'm looking for, I'll take a look at ksnprintf
16:27imirkin: there's some new "c" variant too
16:28imirkin: (kscprintf? something like that, no clue how it's different/better)
16:28Hukka: Ok, so I had to manually enable the main/debug for the ppa too. Progress
16:28ickle: just scnprintf
16:29Lyude: danvet: btw - do you know if the message I sent to Tejun regarding vblank works got put in your spam or not? figured I should make sure since we haven't heard back from them yet
16:29Hukka: Ok, got a stacktrace
16:30Hukka: Anything else I should do before submitting an issue=
16:33shadeslayer: imirkin: ickle: something like so? https://paste.ubuntu.com/p/njnfhtwxsk/
16:35ickle: remember to write in plain old C
16:37danvet: Lyude, didn't end up in my spam at least
16:37Lyude: hm, alright
16:43imirkin: shadeslayer: that works well for handles less than 1000
16:43imirkin: less well for >= 1000
17:39shadeslayer: imirkin: https://paste.ubuntu.com/p/7F4qncZNDB/ < more along these lines maybe?
17:39imirkin: that's going a little overboard
17:39imirkin: how big is handle?
17:39imirkin: u32? u64?
17:40imirkin: just make a char which can handle the max of that
17:41vsyrjala: there's also kasprintf() when you do need the heap alloc
17:42imirkin: these guys thought of everything...
17:44shadeslayer: imirkin: it's a uint32_t
17:44imirkin: ok, so what's the largest uint32_t?
17:45ickle: are we measuring by !monospace string width? pixel coverage?
17:45shadeslayer: 0xFFFFFFFF ?
17:46imirkin: but you're not printing %x
17:46imirkin: you're printing %u
17:46shadeslayer: I mean, sure in decimal it'd be 4294967295
17:47imirkin: so how big does the fixed-size char array need to be?
17:47ickle: ->| |<-
17:47imirkin: what about the null terminator?
17:48shadeslayer: ah of course, 11
17:48imirkin: nice and simple.
17:48imirkin: ickle: you're off by a few spaces it seems
17:48shadeslayer: looking at kasprintf, I feel like it'd work too?
17:48shadeslayer: foo = kasprintf(GFP_KERNEL, "%lu", handle) ?
17:48imirkin: yeah, probably
17:49vsyrjala: +kfree + error handling. complexity adds up
17:50imirkin: you have to kfree the kasprintf?
17:50ickle: yes, like asprintf
17:51imirkin: i thought it was like alloca. i was wrong.
18:23agd5f_: can we get a backmerge of 5.6 into drm-misc-next-fixes?
18:38Pie-jacker875: can I get a quick opinion on if this is a mesa bug or the fault of this one game? lbry://@Pie-jacker875#6/letmeplaythedamngame#7 skip to 1 minute in to see what's going on
18:39imirkin: i think your link is a bit off...
18:39Pie-jacker875: I'm running a RX 5700XT, Fedora 31. the game is Black Mesa.
18:40Pie-jacker875: maybe it should be called white-screen mesa :P
18:45Pie-jacker875: then here's a mirror https://piejacker.net/videos/letmeplaythedamngame.webm
18:47HdkR: Wait, you have a zoom control in that game? They forgot to map it on the gamepad D:
18:48HdkR: lol, what a fun effect. screen just whites out
18:49Pie-jacker875: yeah there's multiple times it will happen in the game too
18:49Pie-jacker875: not just in the test chamber there
18:49Pie-jacker875: I don't know if there's a way to make it go away other than restarting the game
19:05airlied: ajax: currently the X server lacks a maintainer, I would take that to mean it's defacto unmaintainable :-)
19:06ajax: hah, fair enough
19:07ajax: i was thinking maybe my opinion about it was maybe a bit biased
19:07airlied: ajax: well you have been stockholm syndromed by it :-P
20:48krh: !! isn't and operator, it's two operators
20:50jekstrand: You're not wrong
20:50krh: that means I'm right
20:50jekstrand: I'm not sure it does. :P
20:51krh: but !! on a bool is no way better than just casting float to bool
20:52krh: not that either of those are good
20:54jekstrand: Yeah, but != 0.0 is better than both
20:54jekstrand: I can forgive !! or a bool cast on an integer
20:54jekstrand: Oh a float.... Come on. That's just lazy.
20:56imirkin: it's a double
20:56imirkin: the (float) there is just there to mislead you
20:56jekstrand: imirkin: I don't think that makes it any better...
20:56imirkin: the immediates are doubles, so everything gets upgraded to a double
20:56krh: what is !!-0.0f anyway?
20:57jekstrand: krh: -false, obviously
20:57imirkin: the real question is what is !NaN
20:57imirkin: aN? :)
20:58krh: yes, the 'a' stands for 'any' in that case
21:28mattst88: jekstrand: "You're not wrong" <- you should have said "You're not not right" :P
21:32krh: mattst88: https://www.youtube.com/watch?v=a4VvRWTD3Ok
21:32mattst88: I don't know what I'm watching, but I love it
21:37tlwoerner: mattst88: https://en.wikipedia.org/wiki/Pingu
21:38mattst88: thanks :)
21:38krh: noot noot
22:20anholt: tomeu: more lava fails today https://gitlab.freedesktop.org/anholt/mesa/-/jobs/2134509 https://gitlab.freedesktop.org/anholt/mesa/-/jobs/2134508 https://gitlab.freedesktop.org/anholt/mesa/-/jobs/2134507