02:43HdkR: imirkin_: I also have some blueooth ones. Sony WH-1000XM3. Really need to switch over to LDAC if you want audio to not sound like garbage
02:44imirkin: well, with my desktop, i can't get _any_ audio
02:44imirkin: with my laptop, it's ok
02:44imirkin: and the tv as well
02:45imirkin: i have some dumb usb bluetooth adapter in the desktop, will try another one at some point
02:45imirkin: it's from about 10y ago, i think
02:45imirkin: supports all the way up to bluetooth 2.1 :)
02:46imirkin: laptop's not much newer, but i guess newer-enough
03:00imirkin: [yeah, looks like laptop is bluetooth 3.0]
13:00rhyskidd: Lyude: just wanted to check that you're okay with this series before i merge it: https://github.com/envytools/envytools/pull/189
13:01rhyskidd: thanks for your other reviews
15:11jamincollins: imirkin: found some more details around getting triple display working with the Lenovo P1. After upgrading my kernel I found that I was back to the same corrupt display issue. After some trial and error, I found that it seems I have to step by step move the displays (1 display at a time) or I get the corruption
15:19jamincollins: :shrug: I don't make the rules =)
15:20jamincollins: I just try to figure them out enough to have my stuff work the way I want it to
15:21imirkin_: it's a bit hard to tell whose fault it is
15:21imirkin_: the way it works is...
15:22imirkin_: X is responsible for drawing pictures
15:22imirkin_: those pictures are then passed to the kernel kms driver
15:22imirkin_: and the kms driver is told show *this* subrectangle of the picture on *this* display
15:22imirkin_: so ... if the picture is bad, then what's displayed will be bad as well
15:23imirkin_: or if the kms driver is screwing up the picture/subrectangle/etc settings, the displayed image will also be bad
15:23jamincollins: right, and I suspect that is the case... I suspect that when I move all the displays at once the calculated bounding box is wrong
15:23imirkin_: maybe. that would point to the modesetting driver and/or helpers that it uses.
15:23imirkin_: (which is entirely separate from the kernel kms drivers)
15:58AndrewR: imirkin_, hello! Can you enlighten me on where nouveau (nv50) hooks into va state tracker? Because I think va-api processing (not decoding) actually work better with vl-created buffers ..so, idea is to hack them in depend on entry point ...
16:02imirkin_: XVMC_VL works better because there's no video decode engine involved ;)
16:08imirkin_: AndrewR: just look for XVMC_VL in nv50/*
16:09AndrewR: imirkin_, yeah, nouveau_video_buffer_create() in nouveau_video.c . Still, trying to find right conditions to trigger it ....
16:10imirkin_: the video state trackers trigger it
16:11imirkin_: it's a callback in screen (video_buffer_create or something like that)
16:14AndrewR: imirkin_, but my problem apparently arise when ffmpe via av-api tries to alocate two differently-sized surfaces, for scaling ... not sure how to detect this from hw driver?
16:14AndrewR: **va-api ....
16:14imirkin_: with vdpau, there's a VDPAU_TRACE=1 env var which will show you all vdpau api calls. perhaps libva has something similar?
16:22AndrewR: LIBVA_MESSAGING_LEVEL - https://github.com/intel/libva/blob/master/NEWS (starting from 1.8.3)
16:25AndrewR: imirkin_, ..but it doesn't print much (at least with mesa/gallium - only few lines of info)...
16:31AndrewR: imirkin_, there also apparently VA_TRACE, too ... but it not even creates file for me :/ https://github.com/intel/libva/blob/master/va/va_trace.c
16:32imirkin_: all i know about va-api is how to not use it :)
16:32imirkin_: i'm _really_ good at that though
16:45AndrewR: imirkin_, :} still, any reason(s) from your side for such decision? (apart from bugs)
16:47imirkin_: for a very very very long time, mplayer was the only player application that correctly decoded video files 100% of the time
16:47imirkin_: and mplayer only supported xvmc, and then later, vdpau
16:48RSpliet: (coincidentally, by adding vdpau support, it stopped correctly decoding video files on machines with nouveau :-P)
16:48imirkin_: vdpau on nouveau is broken
16:48imirkin_: and has been since day 1
16:48imirkin_: but i think it has gotten more so with time
16:48imirkin_: as people "improve" the video state trackers
16:49RSpliet: Yeah, the back-end defo needs some TLC
16:49imirkin_: unfortunately it's become a bit of a hangsville
16:49imirkin_: and i try to stay out of there
16:51RSpliet: imirkin_: I suspect that with the majority of laptops shipping NVIDIA only as a secondary GPU, it's become a less important feature for a lot of people
16:52imirkin_: and we don't support anything maxwell+
16:55RSpliet: Not even first gen (FrankenKepler?) Maxwells?
17:00imirkin_: the memory controller is similar to kepler, which helps reclocking
17:00imirkin_: but the video engine is new
17:01imirkin_: apparently the API is actually quite similar
17:01imirkin_: so should be "easy" to support
17:02imirkin_: where by "easy", i mean "so hard that i can't think of anyone who will put up the time to work it out"
17:02imirkin_: as opposed to the difficult things on the todo list :)
17:18HdkR: imirkin_: Volta and turing also have new video decoders so woo? :P
17:19imirkin_: uh huh
17:19HdkR: Obviously iterative. Still three generations to contend with :)
17:21imirkin_: i can't believe i got (re)started with nouveau by doing vp2
17:22imirkin_: it helped that i didn't really have a job at the time
17:38AndrewR: sorry, I think I need some sleep before I continue.
17:39AndrewR: imirkin_, thanks anyway, not sure if my hacks will ever result in anything useful ... but i'll try!
19:45karolherbst: rhyskidd: was that you who wanted to add support for disassembling falcon images inside the vbios?
20:24danvet: did nouveau ddx not yet migrate to gitlab?
20:24danvet: karolherbst, imirkin_ ^^
20:24danvet:can't find it
20:24imirkin_: not that i know of
20:27SolarAquarion: for some reason i can't load wayland session afaik
22:42rhyskidd: karolherbst: that is right
22:42rhyskidd: the current parser has bugs, i partially know what is going wrong -- just need to time to polish it up
22:42karolherbst: what bugs?
22:43karolherbst: it should be able to just disassemble the parts of the iamge
22:43rhyskidd: yep, at the moment we aren't getting the start and length's quite right
22:43karolherbst: ohh, you meant the parsing of the table?
22:44karolherbst: it kind of looks okayish? what's wrong there?
22:45rhyskidd: yes, for example see the output of: ./nvbios/nvbios nvidia_bios/nv137/rhyskidd/vbios.rom | grep -I Falcon
22:45rhyskidd: i'd have to check my notes and wip branches, but i clearly recall there being some silly bugs
22:45karolherbst: but that looks okay? no?
22:46rhyskidd: once that's fixed, should be fairly easy to pipe the blobs through to decode
22:47karolherbst: "desc stored size" and "uncompressed size" are labeled wrongly, but besides that?
22:49karolherbst: rhyskidd: I think the last entry is also garbage
22:49karolherbst: dunno what to do about that one
22:49karolherbst: UCODE 4
22:49rhyskidd: or the whole UCODE 4
22:50karolherbst: PRE_OS and DEVINIT are the ones we might parse for now
22:50karolherbst: and ignore the rest
22:50rhyskidd: yeh, i couldn't see how the dmem, imem and sec portions are physically laid out on that one
22:50karolherbst: allthough 0x02 and 0x03 seem to parse fine as well
22:50karolherbst: something is odd about the last one, skeggsb might now more?
22:51rhyskidd: your interested in this area?
22:51karolherbst: well... I don't know. I had to debug devinit for the runpm bug and maybe there is weirdo stuff going on inside the image
22:51karolherbst: might help tracking down silly bugs
22:52karolherbst: but the devinit stuff is more or less pubilc now, isn't it?
22:53karolherbst: the heck thought it was a good idea to let wine install content type handlers?
22:54karolherbst: xdg-open *.xml opens wine... very nice
22:54mwk: ugh don't remind me
22:55karolherbst: the heck is wrong with people...
22:55mwk: I once got a .js file from a student, and double-clicked it in thunderbird
22:55karolherbst: :D well :D
22:55imirkin_: ran in node? :)
22:55mwk: saw wine starting up and immediately thought "well fuck now I got pwned"
22:55mwk: but it turns out that idiotic thing just bound .js files to wine's NOTEPAD.EXE
22:56karolherbst: yeah, I have no idea what's going on with the xml stuff
22:56karolherbst: it doesn't stop spawning processes or windows
22:57karolherbst: mwk: https://i.imgur.com/zgTJ7Qe.png :)
22:57karolherbst: that's my task bar right now
22:58mwk: what the
22:58karolherbst: yeah, ... no idea either
22:59karolherbst: heh winebrowser
23:00karolherbst: maybe I should open a bug for that
23:00karolherbst: heh.. and chromium is pain as well with xml files?
23:01karolherbst: humanity really is going downhills fast :p
23:18rhyskidd: devinit is more or less public now for anything up to ~Pascal
23:19imirkin_: chrome does ok with xml iirc
23:19imirkin_: unless it's a very large file. then you're just screwed.
23:19karolherbst: not if you load them locally
23:19karolherbst: file:// prevents loading css/xlst files
23:19karolherbst: it's stupid, I know
23:19imirkin_: unless you run with --make-it-work
23:19karolherbst: sure.. but you know, who does that
23:20karolherbst: not gonna close my chromium session due to stupid reasons
23:20imirkin_: (allow-local-file-something iirc)
23:20karolherbst: maybe there is a configuration flag somewhere
23:21karolherbst: doesn't seem so
23:22imirkin_: no, only on startup
23:22karolherbst: well, at least there is /etc/default/chromium
23:22imirkin_: i tend to just run python3 -m http.server in that dir
23:30karolherbst: heh... oh uhm.. mhh, it's late. I should sleep, I actually have something important going on today.. (interviewing an intern candidate)
23:31imirkin_: worth sleeping for :)