00:39 imirkin: karolherbst: need me to write a patch to fix the text/etc issue?
00:39 karolherbst: imirkin: if you got a good idea on how to fix it, because I don't really have one
00:39 imirkin: ok
00:39 imirkin: i don't have a GREAT idea, but i'll try a couple things
00:40 imirkin: (which mostly involve reading a bunch of code to see what will fit best)
00:40 karolherbst: if you need something fast to reproduce, set the initial size to 1 << 14 and the webgl triggers the corruption in a couple of tests
00:40 imirkin: ok cool
00:40 karolherbst: *webgl cts
00:40 imirkin: right
00:42 karolherbst: imirkin: btw, I just send out the patch to fix those invariance issues in case you want to review that
00:42 karolherbst: the one fixing up the glsl pass
00:42 imirkin: yeah, i see it
00:42 imirkin: i wonder if there's another way
00:42 imirkin: hold on
00:44 imirkin: aha, no. there isn't.
00:44 imirkin: i was thinking if there was a way to move this before propagate_invariance. but there is not.
00:57 karolherbst: mhh, that patch should probably get a cc: stable as well
01:08 imirkin: if you like
02:11 airlied: imirkin: oh vaguely, most support for BIOS VGA mode, but not exactly much more than that
02:11 airlied: makes text mode appears on both outputs etc
02:12 karolherbst: okay... 33 deqp GLES3.1 fails
02:14 karolherbst: mhh, dEQP-GLES31.functional.draw_indirect and dEQP-GLES31.functional.image_load_store.3d stuff :/
02:17 karolherbst: draw_indirect sounds interesting, the image load store stuff is our 3D images fail, which is annoying to fix as my patch only works for maxwell, but not for kepler
02:20 imirkin: airlied: the "lock vga" thing?
02:20 imirkin: we lock it, run the vbios, then unlock it
02:21 imirkin: karolherbst: load/store 3d should work fine on maxwell+
02:21 imirkin: indirect draw is probably something dumb, like a missing flush or something
02:21 karolherbst: imirkin: not that special case where you bind a layer as a 2d image
02:21 imirkin: karolherbst: oh right, yea
02:21 imirkin: i've gotta say - that was a BAD idea.
02:21 karolherbst: we can fix it by disabling tiling on the z axis
02:21 karolherbst: but
02:21 karolherbst: that doesn't work for kepler
02:22 imirkin: we can fix it in other ways than that on maxwell+
02:22 karolherbst: imirkin: https://github.com/karolherbst/mesa/commit/8137bafc7a702f72196613750fe1d429a7a1dcd5
02:22 imirkin: e.g. we can do a 3d fetch, using the bound layer as the third coord
02:22 karolherbst: sure, but that fix is quite simple
02:22 karolherbst: mhhh
02:22 karolherbst: right...
02:22 imirkin: and conditionalize it in the code
02:23 karolherbst: sure, but that has an impact on basically every image load/store operation
02:23 karolherbst: well 2D ones
02:23 imirkin: should be esp fun for the case where you want to do an atomic op on a 3d texture bound as a 2d layer in a shader.
02:23 karolherbst: :(
02:24 imirkin: like i said - i disagree with whoever thought this was a good use-case to support.
02:24 karolherbst: anyway, what I did was what I saw nvidia doing
02:24 imirkin: karolherbst: more likely nvidia de-tiles
02:24 imirkin: rather than never-tiles
02:25 karolherbst: right, they probably have some way to detect when they can tile on the z axis and when not
02:25 karolherbst: I didn't spend much time digging through it
02:25 karolherbst: anyway, it didn't seem to work for kepler anyhow
02:25 imirkin: right
02:26 karolherbst: the draw_indirect thing I will probably look into tomorrow
09:36 bn_work: hi, can someone help me get X.Org X Server 1.19.5 started with the nouveau or even the framebuffer driver? I am trying to start X from systemrescue cd but I get an error claiming module "glx|fbdev" is not not found, can anyone offer any help as to how to fix that? here's the output from `/var/log/X.org.0.log` when it attempts to start: https://bpaste.net/show/ed8a61f207cd
09:38 bn_work: ah, I just tried looking at the Troubleshooting section in the topic and I bet it's the `nomodeset` kernel option I had to boot with to prevent hang upon start :|
12:46 karolherbst: bn_work: probably yes :/
12:46 karolherbst: bn_work: the modesetting driver can't be used if you boot with nomodeset for obvious reasons ;)
12:47 bn_work: karolherbst: hi, thanks for the reply. What is the recommended way to get into X if I can only boot rescuecd with `nomodeset`?
12:47 karolherbst: bn_work: what GPU is it?
12:48 bn_work: nVidia GeForce 8600M GT: https://everymac.com/systems/apple/macbook_pro/specs/macbook-pro-core-2-duo-2.4-17-santa-rosa-specs.html
12:49 karolherbst: maybe pmoreau knows something?
12:50 karolherbst: but I would expect it to boot just fine without nomodeset
12:50 karolherbst: bn_work: is there any error you see?
12:50 karolherbst: or does it simply hang?
12:51 karolherbst: bn_work: also booting in uefi mode should be better than bios
12:51 karolherbst: the gentoo live CD kind of suck big times in this regard
12:51 karolherbst: just use some ubuntu one and install gentoo with that (that's what I usually end up doing)
12:51 bn_work: karolherbst: at which point? if I boot with `nomodeset` and start X manually via `startx`, I get the X.org log pasted above. If I boot "normally" (ie: without the above kernel param), it always hangs at loading kernel modules
12:51 karolherbst: bios mode on macbooks is no fun at all
12:52 bn_work: karolherbst: yes, that was the only way I was able to boot the rescue CD, by choosing "EFI boot"
12:52 karolherbst: huh
12:52 karolherbst: then the rescue CD is broken
12:52 karolherbst: fbdev should be able to load in such a case
12:53 bn_work: uh... install Ubuntu? I'm trying to boot into rescue cd to migrate partitions over from my current rotational HD to an SSD, it already has it (an old ver) installed but it's unbootable currently
12:53 karolherbst: bn_work: no, use the ubuntu live image to install gentoo
12:53 karolherbst: ohhh
12:53 karolherbst: I see
12:54 bn_work: hence the systemrescuecd
12:54 karolherbst: bn_work: the thing is, that the live cd is like super old and issues might have been fixed already
12:54 karolherbst: dmesg might tell whats going on
12:55 bn_work: karolherbst: is there anything you were able to glean from the log pasted above? it seems nouveau detects the nvidia types, am I missing a module for it?
12:55 karolherbst: if you boot with nomodeset the nouveau DDX can't be used
12:56 karolherbst: normally you would still have the efifb or something setup _and_ fbdev would be able to take over
12:56 karolherbst: but that's not happen either
12:56 karolherbst: dmesg might be able to tell us what's wrong
12:56 bn_work: the rescuecd isn't that old, it's v5.1.2 from 2017 I believe?
12:56 karolherbst: it's old enough
12:56 bn_work: much newer than this macbookpro3,1 for sure
12:56 karolherbst: sure
12:57 karolherbst: just saying that within this year we could have fixed issues ;)
12:57 karolherbst: normally I'd say it should just work
12:57 karolherbst: but there can be always issues
12:58 bn_work: I'm at the (proper) sysrescue grub prompt again, should I add `nomodeset` or `nouveau.modeset=0`? or something else?
12:58 karolherbst: if you can, try to boot without any flags and check if there are any errors or something. If you have another machine you could try to ssh into that machine
12:59 karolherbst: that would be probably the best approach to figure out what's going on
12:59 bn_work: so the above pasted log tells you nothing about why fbdev isn't loading? is it missing?
13:00 bn_work: https://www.irccloud.com/pastebin/YizE61TB/
13:01 bn_work: karolherbst: I can't, it hangs if I boot without params, I don't think it inits network by default?
13:01 karolherbst: well, it might be that just the display freezes or something and it happily continues booting
13:02 karolherbst: worst case you could try setup netconsole
13:02 karolherbst: thing is, if you boot with nomodeset or nouveau.modeset=0 everything you run into is not a nouveau issue
13:02 karolherbst: because those options essentially disable nouveau fully
13:03 karolherbst: you coult still try to boot this way and paste the dmesg output
13:03 bn_work: I've left it alone for quite some time (5-7 mins?) until the DVD drive stops making noises, it shouldn't take that long to boot, and the verbose output always stops at the same place, caps lock still works though, FWIW
13:03 karolherbst: maybe that will tell us something why fbdev can't take over
13:04 karolherbst: bn_work: anything nouveau related on the verbose output?
13:04 bn_work: is there a way to boot with `nomodeset` and then enable it after the fact so I can at least look at the X.org log?
13:05 karolherbst: you could try to blacklist nouveau and load it via ssh
13:05 karolherbst: no idea how to do that with the gentoo livecd though
13:05 karolherbst: or uhm... you should be able to rmmod nouveau after booting with nouveau.modeset=0
13:05 karolherbst: and modprobe it again without that option
13:05 karolherbst: like "modeprobe nouveau.modeset=1"
13:07 bn_work: ok, booting with `noveau.modeset=0`
13:07 bn_work: is it `modprobe` or `modeprobe`?
13:08 karolherbst: modprobe
13:09 bn_work: ok, it reports `modprobe: FATAL: Module noveau.modeset=1 not found in directory /lib/modules/4.9.60-std512-amd64`
13:09 karolherbst: uhm "modprobe nouveau nouveau.modeset=1", sorry
13:10 karolherbst: uhm
13:10 karolherbst: that's wrong as well
13:10 karolherbst: "modprobe nouveau modeset=1"
13:10 karolherbst: that
13:10 bn_work: hmm, it worked? but no output... return code 0 though
13:11 karolherbst: did dmesg print anything?
13:11 karolherbst: or is there anything usefull in dmesg?
13:11 bn_work: nothing relevant looking after I ran it
13:12 bn_work: should it output anything into it?
13:12 karolherbst: there should be some nouveau messages comming up
13:13 bn_work: when I do `dmesg -H | grep nouveau` all I get are the kernel CLI options I passed at `+0.000000`
13:13 karolherbst: grep -i nouveau
13:13 bn_work: same
13:13 bn_work: I mean it returned 0, so supposedly it loaded successfully, no?
13:13 karolherbst: did you rmmod nouveau before modprobe?
13:14 bn_work: no
13:14 karolherbst: okay, well you need to rmmod nouveau first, then you can load it with different parameters ;)
13:15 karolherbst: and then my hope is that anything turns up in dmesg
13:15 karolherbst: either bad or good
13:16 bn_work: interesting, when I do `modprobe -v nouveau modeset=1` I see it trying to do `insmod /lib/modules/4.9.60.std512-amd64/kernel/drivers/gpu/drm/nouveau/nouveau.ko modeset=0 modeset=1`
13:17 bn_work: I assume the last invocation overrides the first?
13:17 karolherbst: hopefully
13:17 bn_work: yeah, it hung there
13:17 karolherbst: now hopefully dmesg helps us
13:18 bn_work: I could try enabling networking and sshd I guess
13:18 karolherbst: yeah
13:18 bn_work: to get to it
13:52 bn_work: ok, finally have it back up with networking and sshd, oddly it didn't boot with kernel param `nouveau.modeset=0`, so had to boot with just `nomodeset`, about to test modprobe again
13:53 bn_work: I assume `nouveau.modeset=0` ≠ `nomodeset` and slightly different?
13:53 karolherbst: bn_work: I'd expect so
13:54 bn_work: here we go!
13:55 bn_work: karolherbst: https://www.irccloud.com/pastebin/YXamUuB5/modprobe%20nouvea%20modeset%3D1.log
13:56 karolherbst: ohhhh
13:56 karolherbst: it can't read the vbios
13:56 karolherbst: yeah.. that might have been fixed in a newer version of nouveau
13:57 karolherbst: bn_work: I would assume the vbios to be stored in the ACPI tables directly which we might not support in 4.9 yet
13:57 karolherbst: I'd try out a live image with a newer kernel just to verify that
14:00 bn_work: unfortunately newer livecd version removed clonezilla's partclone, which is why I used this version :| ... but seeing how networking at least works now, maybe I can scp it over once booted? Also, need to burn another CD which is a bit hit-or-miss with this old DVD drive 😰
14:00 karolherbst: ahh
14:00 bn_work: what kernel version should it be?
14:00 karolherbst: yeah you could just use scp or something I guess
14:00 karolherbst: bn_work: dunno
14:02 bn_work: why does it fatally hang?
14:02 bn_work: or wait, looks like ssh still works
14:02 bn_work: at least
14:03 bn_work: I guess modprobe goes comatose
14:03 karolherbst: looks like the code is there with 4.9 though :/ odd
14:03 karolherbst: maybe something is really special about your laptop, dunno
14:04 bn_work: hmm I don't even see the modprobe process
14:05 bn_work: well, it is a macbookpro3,1 VS a traditional PC :)
14:05 bn_work: so does this mean I need to boot in BIOS mode? if so, how?
14:05 karolherbst: nouveau.config=NvBiost=ACPI might just work
14:06 karolherbst: uhm
14:06 bn_work: when I do so, I always just get the (emergency?) grub prompt
14:06 karolherbst: nouveau.config=NvBios=ACPI
14:06 bn_work: kernel mode param on boot?
14:06 karolherbst: but I am sure pmoreau might know more about this
14:06 karolherbst: bn_work: yeah
14:20 bn_work: karolherbst: ok, back up with networking and the new kernel mode param, so try same modprobe cmd again?
14:22 bn_work: actually, I wonder if it'd be better if I tried it from a 2nd screen window in the ssh session that way console and ssh still work
14:23 bn_work: hmm, nothing happened, or it's just return code = 0
14:24 bn_work: oops, forgot to rmmod nouveau first again :P
14:26 bn_work: karolherbst: ok, this time getting `bios: ACPI invalid`: https://www.irccloud.com/pastebin/n5UtzxXk/
14:27 karolherbst: mhh, then I am out of ideas
14:28 karolherbst: huh.. wait
14:29 karolherbst: bn_work: it is a 4.9 kernel, right?
14:29 bn_work: I believe so, yes
14:33 karolherbst: bn_work: might try loading nouveau with debug=trace in addition?
14:37 bn_work: hmm, console still hung anyway even though I ran modprobe from a 2nd screen window in an SSH session
14:38 bn_work: karolherbst: `modprobe nouveau modeset=1 debug=trace`?
14:38 karolherbst: yeah
14:41 bn_work: is vbios = VESA BIOS?
14:42 karolherbst: I think so
14:42 karolherbst: uhm
14:42 karolherbst: no
14:42 karolherbst: video bios
14:42 bn_work: so boot up with `nouveau.config=NvBiost=ACPI` again?
14:42 karolherbst: *NvBios
14:43 karolherbst: without the t, but yea
14:43 karolherbst: h
14:43 bn_work: oops, yeah, that's what I mean
14:43 bn_work: (that's what it was booted up with)
14:43 karolherbst: we support a various amount of methods to retrieve the vbios from the GPU, and I would like to know why all of those fail
14:50 bn_work: BTW, while it was booting I found this, looks very similar to what I'm seeing? https://bugs.freedesktop.org/show_bug.cgi?id=35267
14:55 bn_work: karolherbst: https://www.irccloud.com/pastebin/Y0kWPlxC/modprobe%20-v%20nouveau%20modeset%3D1%20debug%3Dtrace.log
14:56 karolherbst: bn_work: might be that bug, yes
14:57 karolherbst: bn_work: what would be interesting to check is whether it works with the newest kernel and whether it works with the nvidia binary driver
14:59 bn_work: how would I do get the latter for this integrated GFX chipset model?
14:59 bn_work: I'm assuming there's some package for the GeForce 8600M GT?
14:59 karolherbst: 404 driver version
14:59 karolherbst: uhm 340 I think
15:00 karolherbst: let me check
15:00 karolherbst: 340.x yes
15:01 karolherbst: bn_work: but the first thing to try out would be a 4.20 or 4.19 kernel
15:02 karolherbst: if that still doesn't work with nouveau, we could use mmiotrace and the 340.x nvidia driver to figure out what we have to do to get this working inside nouveau
15:04 bn_work: would that tell you where the vbios.rom is located? I had the exact same question as the last comment on that ticket, ie: where did the OP get the vbios
15:04 karolherbst: yeah, essentially
15:06 bn_work: ok, let me see what I can do, I will need to burn another CD though, which may or may not burn successfully given *fingers crossed* :)
15:06 bn_work: s/given/given the age of the DVD drive/
15:08 pmoreau: rhyskidd: Same here, and also bumped the RAM from 4 GB to 8 GB. The one more thing I would like to update would be the screen, to get 1080p or maybe 1440p. :-)
15:09 pmoreau: I’ll send you a pointer to the patches.
15:09 karolherbst: pmoreau: ahh, you are here. Do you have any ideas about bn_work issues?
15:09 karolherbst: pmoreau: on his macbook nouveau isn't able to locate the vbios
15:09 pmoreau: (Haven’t gone back through the logs yet)
15:09 karolherbst: pmoreau: don't need to
15:10 karolherbst: it's a 8600M
15:10 pmoreau: I never needed to do anything funky to locate the VBIOS on my laptop.
15:10 karolherbst: yeah... I kind of feared as much
15:10 pmoreau: I have the next generation: 9400M + 9400M GT
15:10 karolherbst: it's a Mid/Late 2007 macbook
15:10 pmoreau: *9600M GT
15:10 karolherbst: so older than yours
15:10 pmoreau: Yeah
15:10 karolherbst: okay, I was mainly wondering if you knoew anything
15:11 karolherbst: *knew
15:11 pmoreau: No, sorry :-/
15:11 bn_work: pmoreau: do you boot up in BIOS mode or EFI?
15:11 imirkin: bn_work: are you booting with EFI or BIOS boot?
15:11 imirkin: iirc the vbios only shows up in a bios boot
15:11 pmoreau: EFI
15:11 imirkin: not in an efi boot
15:11 karolherbst: ohh, interesting
15:11 imirkin: and so you have to pass in the vbios directly in an efi boot
15:11 imirkin: at least on some models
15:11 karolherbst: imirkin: does it work with the nvidia driver?
15:11 imirkin: there's a bug open about it
15:11 imirkin: dunno, unlikely
15:12 imirkin: https://bugs.freedesktop.org/show_bug.cgi?id=91779
15:12 karolherbst: I vaguely remember know that there was an issue like that
15:13 imirkin: bn_work: --^ see that bug
15:13 imirkin: i believe you have the same model
15:13 karolherbst: wondering how mac os x handles that
15:13 bn_work: imirkin: ya
15:14 imirkin: karolherbst: probably has the vbios shipped as part of it
15:14 imirkin: remember -- very few SKU's out there of mac's
15:14 karolherbst: mhhh
15:15 imirkin: it gets it from somewhere for the bios boot
15:15 imirkin: which is meant for windows
15:15 imirkin: perhaps there's a way to retrieve it
15:15 karolherbst: true
15:15 imirkin: but we don't know how
15:16 bn_work: imirkin: yeah, I had the same question as you on your last comment of https://bugs.freedesktop.org/show_bug.cgi?id=35267
15:17 imirkin: bn_work: do a bios boot, grab it from /sys/kernel/debug/dri/0/vbios.rom
15:18 bn_work: imirkin: unfortunately when I do that, I get an (emergency?) grub shell and am a bit lost :|
15:18 bn_work: do I just basically need to type in (from memory) what was in my grub.cfg?
15:19 imirkin: ah ok. i just know how nouveau works, i don't tend to help much with general "operate the computer" type issues
15:20 imirkin: if you have a livecd, you can do the bios boot, it will most likely work
15:20 imirkin: you can use almost any version of nouveau to retrieve the vbios
15:20 bn_work: imirkin: would I be able to obtain it from a VBox raw VMDK that "boots" off the partition? or does VBox give it a "fake" vbios?
15:20 imirkin: you need to do a real bios boot on the mac
15:20 bn_work: imirkin: that's what I meant, when I boot from the livecd :/
15:21 imirkin: virtual emulation software provides virtual devices
15:21 imirkin: which is not what you want
15:22 bn_work: on my mac the livecd boot options show up as "Windows boot" and "EFI boot", if I choose the former, I get something like https://www.tecmint.com/wp-content/uploads/2017/09/Ubuntu-Grub-Console.png If I choose "EFI boot", I get a normal Grub boot selector
15:23 imirkin: sounds like you should be looking for help from whoever supplied the livecd.
15:24 imirkin: but in order to retrieve your vbios, which is necessary for nouveau to work, you need to do a windows boot at least once to grab it.
15:24 imirkin: after that, you can use efi boot
15:24 bn_work: this happens with my normal Linux partition too, it's something specific to booting in BIOS mode on my mac I believe
15:27 bn_work: is there any way to grab that from within OSX?
15:28 imirkin: probably. no clue how though.
15:28 bn_work: OSX is based off (Open?)BSD so is there an analogue to /sys/kernel/debug/dri/0/vbios.rom?
15:28 imirkin: that's a nouveau thing
15:28 imirkin: and the bsd's aren't so big on the fake filesystems like /proc and /sys like linux is
18:08 pmoreau: rhyskidd: https://github.com/l1k/linux/commits/mbp_switcheroo_v5-4.5 and https://github.com/l1k/linux/commits/mbp_switcheroo_runpm_wip seem to be the latest versions of his work.
18:09 pmoreau: Maybe some patches from https://github.com/l1k/linux/commits/switcheroo_devlink_v2 could be used, but I think those are mostly needed when you have the HDA device, which is not the case on those old laptops.
19:06 rhyskidd: imirkin: i'm looking at the HAS_DIRTYTRACKING_DRAWABLE_SRC patch
19:52 rhyskidd: there's a bit more to it
19:58 imirkin: rhyskidd: yeah, i think there is
19:58 imirkin: it looks like dri3 + screen rotation is broken for me now
19:58 imirkin: i suspect it has something to do with that
19:58 imirkin: dri2 works fine though
20:01 sk3z: hello
20:02 rhyskidd: imirkin: you might also want to look at https://patchwork.kernel.org/patch/10564973/
20:03 sk3z: i was searching bugzilla for this issue i had today, but i am not shure how to report there and if... basically i only today noticed that my 1080 is not working with the drivers, having tried a few distros with different des and only distros that boot at all are those loading kde, and even there it is non usable due to heavy glitching and most weird aberrations
20:03 sk3z: basically this make 99% of recent live distros non bootable
20:05 sk3z: i am a very inexperienced user, so i thought it might have been right to reach out and see if it was worth to report this situation, as it would be probably already known?
20:05 sk3z: thank you in advance
20:07 imirkin: rhyskidd: yeah, i saw that
20:07 imirkin: rhyskidd: and a somewhat different patch for amdgpu
20:07 imirkin: i think fixing this properly will reqiure more reading
20:07 imirkin: i'd like to do a release of nouveau over the weekend, so let me know if you plan on looking nowish
20:07 sk3z: this is very recent situation, as previously drivers used to work. i had problems with updating opensuse (kde) having those issues above, and then i tried xfce cinnamon and kde versions of manjaro, and only one booting was that kde one, with weird glitches as above
20:07 imirkin: if not, i probably should
20:08 imirkin: sk3z: my guess is that the previously "working" setup involved zero nouveau
20:08 imirkin: i.e. just using efifb or whatever
20:08 imirkin: sk3z: you may have better luck adding nouveau.noaccel=1 on the kernel cmdline
20:09 sk3z: i will do, i don't know if these distros just now started using nuveau and they did not in the previous weeks
20:09 imirkin: rhyskidd: i'm also getting a crash with rotated screens in https://hastebin.com/azotenokah.shell -- this only happens when i try to log in with an old gdm. not when i just run xinit directly.
20:09 imirkin: sk3z: more likely they updated nouveau to a version that supports your GPU
20:10 sk3z: and since now it supports my gpu, it loads and break everything?
20:11 imirkin: sk3z: exactly
20:11 sk3z: fantastic
20:11 imirkin: thank you for choosing nvidia. we appreciate that you have a choice of graphics card vendors, and you appear to have chosen the wrong one.
20:12 sk3z: appearently, yes
20:12 imirkin: https://www.youtube.com/watch?v=BoekVaXdA3U
20:12 sk3z: i knew what i was going into, still, i thought it was worth to report
20:13 imirkin: sk3z: that board should, generally speaking, work
20:13 imirkin: however without any sort of additional info, there's not a lot that can be said
20:13 sk3z: since many more won't be able to install anything from a live distro
20:13 imirkin: distros put whatever they want into their livecd's, we have no control over it
20:14 imirkin: unfortunately the recent trend of software has been to use GL for all kinds of unnecessary tasks
20:14 imirkin: which further exasperates the situation with less-than-perfect drivers
20:15 imirkin: whereas it used to be that GL was for games, it's now for drawing basic UI components as well
20:16 sk3z: i see
20:17 sk3z: well, so whose problem is this, aside from mine? i am only worried of the many people that might try a bunch of distros and not getting anyone to start at all
20:17 imirkin: rhyskidd: otoh, i'm not sure that dri3 + rotation *ever* worked in nouveau, with any xorg version
20:17 imirkin: sk3z: the distros control all this
20:17 imirkin: if you have a problem with a distro's livecd, report it to the distro
20:19 sk3z: ok, well i guess, you are suggesting me to send a wide email to all distros, right?
20:20 imirkin: just the ones you had problems with
20:20 sk3z: like, hey people at ubuntu, 1080 can't load nuveau, beware! and then same with debian arch gentoo fedora suse and so on
20:20 imirkin: an email like that would, of course, be ignored
20:20 imirkin: since it provides no additional info
20:21 sk3z: so what additional info is needed?
20:21 imirkin: dunno, ask them
20:21 imirkin: or you could start a campaign with all those distros to make nouveau not be used by default
20:21 imirkin: because you had an issue with it in your setup
20:23 sk3z: well it has been an interesting talk
20:23 sk3z: sorry to bother you guys
20:24 sk3z: good bye
21:15 imirkin__: wow, so unusual to have my work-style setup at home. have to get used to 2x rotated screens
21:15 imirkin__: karolherbst: ok, so you wanted me to look at the repeated_clear and text-lifetime issues, right?
21:15 karolherbst: imirkin__: yes
21:16 karolherbst: preferably the text-lifetime issue, because that's kind of blocking webgl right now... allthough I could just increase the text size locally I guess
21:16 imirkin__: noted.
21:16 karolherbst: but, in regards to the chromium situation probably a good thing to fix first :)
21:16 imirkin__: sure
21:17 imirkin__: and unrelated to all this, i was also going to look at why emersion is getting clear images in wayland
21:18 imirkin__: rhyskidd: ideally we can do a xf86-video-nouveau release this weekend, so let me know where you end up with that dirty tracking thing when you get bored with it
21:18 imirkin__: karolherbst: btw, i now have GK208B and G84 plugged in
21:19 emersion: :)
21:20 imirkin__: emersion: i'm going to start with your thing
21:20 imirkin__: coz it should be the simplest :)
21:21 emersion: nice!
21:21 emersion: if you can't debug it like this, i can build a minimal test case
21:22 imirkin__: well, step 1 is to try it via the x11 backend
21:22 imirkin__: if i can repro that, it should be all gravy
21:22 emersion: not sure it'll pick up the same formats via the X11 backend
21:22 imirkin__: i hacked it locally :)
21:24 imirkin__: no such luck
21:25 imirkin__: some day i'll figure out how to make ^W not close a window in hexchat...
21:27 imirkin__: middle click to close tab is configurable. why isn't ^W. gr.
21:34 imirkin__: urgh. this wayland stuff is not designed to be developer-friendly =/
21:36 imirkin__: no way to specify what gpu render node it should take
21:37 karolherbst: imirkin__: afaik there is through udev
21:37 pmoreau: karolherbst: I posted the few comments I had regarding your “nv50/ir: Add lowering helper” patch.
21:37 karolherbst: imirkin__: you mean to select which GPU is used as the "main" one inside the compositor, right?
21:38 karolherbst: imirkin__: apperantly mutter selects based on the existence of the "master-of-seat" udev flag, no idea how that works out for other compositors
21:39 imirkin__: karolherbst: like i said, not developer-friendly
21:39 imirkin__: emersion: ok, reproduced
21:39 karolherbst: yeah.. the lack of a general config file makes things a bit challanging
21:39 pmoreau: Will try to set things up tomorrow to test my clover series on Radeon (to check that patch 13 “clover/llvm: Allow translating from SPIR-V to LLVM IR” works, since I don’t think anyone did), and then plug in a Kepler or Maxwell card to test Connor’s patch.
21:40 karolherbst: pmoreau: cool
21:40 imirkin__: emersion: so ... looks like wlroots picked up all 3 displays, even though they were across GPUs. surprising. so i saw 3 images get sent (all with the same format)
21:41 imirkin__: emersion: is there a way to run a wayland application from outside the wayland environment? is there something like DISPLAY= for X11 that directs the client to the proper place?
21:43 emersion: imirkin__: yes, it's WAYLAND_DISPLAY
21:43 imirkin__: and what do i point it at?
21:43 emersion: yeah, wlroots will load all GPUs, draw on the one you booted with and share buffers via dmabuf
21:44 emersion: the first wayland session will have WAYLAND_DISPLAY=wayland-0
21:44 imirkin__: ok thanks
21:45 emersion: > no way to specify what gpu render node it should take
21:45 emersion: you can specify the DRM node(s) it should use
21:45 imirkin__: emersion: how?
21:45 emersion: WLR_DRM_DEVICES=/dev/dri/card0:/dev/dri/card1
21:46 imirkin__: aha, didn't get that far
21:46 imirkin__: nice
21:46 emersion: (only specifying one works too ofc)
21:46 imirkin__: much better!
21:47 imirkin__: and how do i figure out the correct WAYLAND_DISPLAY to use?
21:47 imirkin__: (0 appears to be "not it")
21:49 imirkin__: oh, i think it's a permissions thing...
21:49 imirkin__: connect(3, {sa_family=AF_UNIX, sun_path="/var/run/user/1000/wayland-0"}, 31) = -1 EACCES (Permission denied)
21:49 imirkin__: even though the socket is there
21:49 emersion: hmm, are you running wlroots with a different user?
21:49 imirkin__: yes
21:49 imirkin__: have to run as root for the main thing
21:49 emersion: hmm, i'm not sure how this can work
21:50 emersion: maybe try running the client as root too?
21:50 imirkin__: yeah, it might be workable, but easier to just run the client as root :)
22:53 imirkin__: emersion: looks like _mesa_format_matches_format_and_type pans out, and we hit a fallback path, which just does a memcpy of the data
22:54 imirkin__: the source buffer is XRGB
22:55 imirkin__: and somehow the rb->Format = bgra8, while src->format = BGRX8
22:55 imirkin__: that'll definitely end in tears =/
23:12 rhyskidd: imirkin__: i'll let you know where i get to
23:12 rhyskidd: agree that it would be good to get an xf86-video-nouveau release out
23:13 rhyskidd: i've also now seen a couple of other quirks that it would be good to fix, like the implicit declaration of wfbScreenInit
23:17 prOMiNd: anyone not sleeping in this late hour?
23:17 prOMiNd: :)
23:18 emersion: imirkin__: hmm, i'm not familiar enough with nouveau/mesa to understand this
23:34 imirkin__: emersion: yeah, i'm not expecting you to be :)
23:34 imirkin__: emersion: working through it slowly
23:36 emersion: got it, good luck!
23:36 imirkin__: but basically mesa thinks that it has a BGRA8 texture while the gallium resource is BGRX8
23:36 imirkin__: hence the broken alpha data
23:42 emersion: makes sense
23:42 prOMiNd: imirkin__, found a field that is different per memory vendor in 0x9a0250
23:43 prOMiNd: any chance what A1 means for samsung and A3 for micron ?
23:44 imirkin__: sorry, i don't really know what you're talking about
23:58 prOMiNd: me neither, just shooting in the dark for some enlightenment