[06:24] <linfeng> hello every X man
[06:25] <linfeng> Hope I'm not in a wrong chanel :)
[06:25] <linfeng> I want to seek help about building xorg-server for ubuntu ?
[06:25] <mlankhorst> apt-get build-dep xorg-server; apt-get source xorg-server; cd xorg-server; dpkg-buildpackage ?
[06:25] <linfeng> My problem is that I want to backport a patch for xorg 1.6.4, and rebuild it for exactly ubuntu 9.10. I google for a long time but can't find any guidance about this.
[06:26] <tjaalton> erf
[06:26] <tjaalton> 9.10 is EOL a long time ago
[06:26] <mlankhorst> and that^
[06:26] <tjaalton> without any security support for 2,5y now
[06:27] <linfeng> yes, I know. I just want to rebuild it, but I can't find the *official* confirue option for u910.
[06:27] <linfeng> Is there any way that I can find the building options?
[06:28] <tjaalton> grab the source like mlankhorst advised, then read debian/rules
[06:28] <tjaalton> and modify, then run dpkg-buildpackage to build a package..
[06:28] <tjaalton> maybe touch debian/changelog too to change the revision
[06:28] <linfeng> Hmm, I have got the source by apt-get souce.
[06:29] <linfeng> Thank you, then I will try to read debian/rules.
[06:30] <Dandel> Any ideas on when xorg edgers will have test packages for mesa that enable radeon uvd through vdpau? ( This includes some patches for GL_NV_vdpau_interop )
[06:32] <Dandel> I also know of some alsa patches that 7.1 surround sound on top of a few other features for the radeon hda codec.
[07:18] <mlankhorst> Dandel: no idea :P
[09:24] <tjaalton> nice, can't log in after latest updates
[09:38] <tjaalton> removed a crash file and it works again..
[13:36] <Prf_Jakob> Hello
[13:36] <Prf_Jakob> It looks like you guys have the wrong pm-quirk for SVGA, yeilding to suspend hanging the device.
[13:37] <Prf_Jakob> is this the right place to get that changed or is there a better channel for that?
[13:37] <Prf_Jakob> http://paste.ubuntu.com/6128545/
[13:39] <tjaalton> probably #ubuntu-devel material, but first thing to do is file a bug against pm-utils :)
[14:34] <bjsnider> tseliot, so what's the deal with nvidia-persistenced? referenced in saucy but no package is there
[14:37] <tseliot> bjsnider: I think it's still stuck in the queue and I've kind of given up on it
[14:38] <bjsnider> tseliot, the release notes suggest that it's inside the nvidia-installer somewhere, although i haven't unpacked it to check. why not just add it to nvidia-graphics-drivers? why a separate package?
[14:39] <tseliot> bjsnider: it's the only way to control what happens and to fix things if there are problems
[14:40] <bjsnider> tseliot, so is there somewhere you can get the code other than the installer, or do you unpack it and grab the code from there, or how does that work?
[14:41] <tseliot> bjsnider: yes, nvidia has a git repository on github
[14:42] <Dandel> tseliot, On that repository there is actually one key entry that probably should be expanded for mir/xmir and other things... the vendor independent gl libs.
[14:43] <Dandel> libglvnd ( https://github.com/NVIDIA/libglvnd ) the GL Vendor-Neutral Dispatch library.
[14:44] <tseliot> Dandel: that's a proposal but I haven't heard much about its adoption on the mailing lists
[14:45] <tjaalton> a skeleton package is on pkg-xorg
[14:45] <tjaalton> didn't build the last time I tried
[14:46] <Dandel> tseliot, I got the libglvnd compiled without much work.
[14:46] <tjaalton> there's a presentation about that tomorrow at xdc
[14:46] <tjaalton> well it's still not that usable atm
[14:46] <tseliot> exactly my point ^
[14:47] <Dandel> Of course it's not usable yet... I Think the package is eventually meant to be properly integrated into mesa
[14:47] <tjaalton> it doesn't need to be a part of mesa
[14:47] <tjaalton> supported by it, sure
[14:47] <tjaalton> dunno
[14:47] <tjaalton> could be wrong
[14:48] <RAOF> Doesn't need to be a part of mesa, and shouldn't be.
[14:48] <RAOF> That's part of the point! :)
[14:48] <tjaalton> right :)
[14:48] <tjaalton> and morning! everything set up for the presentation? :)
[14:49] <Dandel> Then how about another good question... How many ubuntu devs actually have a reasonable list of amd/ati ( let alone nvidia ) graphics?
[14:49] <tjaalton> "list of graphics"?
[14:50] <Dandel> Ya... It's somewhat important because of certain failures that occur when dealing with specific mixed mode system configurations.
[14:50] <bjsnider> there ain't no ati no more
[14:50] <tseliot> do you mean hybrid graphics?
[14:50] <tseliot> (as in PowerXpress)
[14:50] <Dandel> I know... ATI/AMD are one in the same... I just list ATI since some may not know that they are the same company now.
[14:51] <tjaalton> i basically run on 100% intel
[14:51] <bjsnider> same here
[14:51] <bjsnider> all inthell
[14:51] <tjaalton> test nouveau/radeon so they won't break on major upgrades
[14:51] <tseliot> I have intel, nvidia, amd, intel/nvidia and intel/amd (hybrid) here
[14:51] <Dandel> I strictly use amd graphics ( and processors )
[14:52] <bjsnider> that's too bad
[14:52] <tjaalton> oh i have a hybrid intel/nvidia too, but without PM there's no point in running it hybrid
[14:53] <tjaalton> and it took two years for nouveau to get somewhat mature on it
[14:53] <Dandel> Actually I have an AMD/AMD hybrid... only problem is that mesa does not like switching between the two graphics ( I think it's muxless )
[14:53] <tjaalton> next year there's going to be a laptop refresh, probably the same deal again :)
[14:54] <bjsnider> tjaalton, you got a bios switch that turns one of them off so you can exclusively use the other?
[14:54]  * RAOF has an AMD/AMD hybrid in his luggage
[14:54] <Dandel> I actually plan on a desktop upgrade next year with the hardware refreshes.
[14:54] <Dandel> RAOF, amd APU with dedicated graphics? ( like a6-3400m cpu with hd6650M gpu )
[14:54] <RAOF> Indeed.
[14:55] <tjaalton> bjsnider: yup
[14:55] <bjsnider> Dandel, are you forced to use amd?
[14:55] <tjaalton> bjsnider: so I just run it in integrated mode all time
[14:55] <Dandel> Actually I found that there is some serious problems where a dummy driver may need to be written... It's a driver explicitly written to give AMD and nvidia a bone to stop the need for blacklisting nouveau and radeon
[14:55] <RAOF> It was the wonderful SUMO that didn't quite do tiled->untiled copies correctly.
[14:56] <Dandel> bjsnider, no... It's called Intel was always out of budget every time i upgrade.
[14:56] <tseliot> tjaalton: after my work is done, it will be much easier to switch between the two cards (with intel+nvidia). Only a log out will be required
[14:56] <tseliot> at least on 12.04.4
[14:56] <tjaalton> right, but it's still too much :)
[14:57] <Dandel> tseliot, backend to remove need to log out needs to be added.
[14:57] <tjaalton> especially as the thing runs two sessions all the time (mine and my wifes)
[14:57] <bjsnider> a log out would take the same time as a reboot here
[14:57] <tseliot> I think X is the problem there, unless you do what Bumblebee does
[14:57] <Dandel> I know that libglnvd is meant to assist with this especially when coupled with drm prime and the drm prime helper patches nvidia added.
[14:58] <Dandel> I actually have a couple of configurations where I do not have any hardware accelerated video on some video cards due to the blacklisting problems :/
[14:58] <tseliot> tjaalton: then no, it wouldn't help in your case
[14:59] <Dandel> example mixture... Ubuntu 12.04 with Radeon X1900GT ( or Radeon HD4550 ) and an radeon HD5770 ( the latter most card is using current binary driver from amd )
[14:59] <tseliot> I don't think X would let go of the card driving the display, at least not without restarting it
[15:00] <RAOF> That could be fixed; it watches hotplug events, it could handle hotunplug events.
[15:00] <Dandel> tseliot, There is some support for that... it's in Xrandr 1.4 ( maybe 1.5 )
[15:01] <Dandel> I did notice that piglit packages have stopped having daily builds... There is a patch queued up to fix the daily build.
[15:01] <Dandel> The latest break is due to the added implementation of strndup to piglit.
[15:03] <tseliot> Dandel: that's offloading rendering using RandR providers (which is only part of the problem), which works as long as both cards are on
[15:04] <Dandel> I do not think it'll be too much of a problem to expand the power management to allow the lowest power management mode be where the card is off and when it gets used there is a wake up call to the dedicated graphics card.
[15:07] <Dandel> The actual implementation should be easy to implement if you know where to look. All it takes is adding some code to monitor gpu activity level and whether or not the card has displays attached.
[15:08] <RAOF> Well, and you need to ensure that nothing has an active GLX/EGL context on the card.
[15:08] <RAOF> Even if it's not rendering.
[15:08] <RAOF> Because if  you turn the card off, that context is gone gone gone.
[15:09] <RAOF> airled was working on something like that at one point
[15:09] <Dandel> RAOF, gpu activity level includes testing for an active GLX/EGL context.
[15:11] <Dandel> RAOF, I did spot a bug in the graphics drivers packages where dedicated gpus do not get direct contexts on the multiple fglrx packages.
[15:11] <Dandel> it's an linker problem causing it
[15:13] <tseliot> Dandel: do you have a bug number for that?
[15:14] <Dandel> tseliot, I didn't report it since I ran into the bug with an ubuntu derivitive on lts
[15:14] <tseliot> ok
[15:15] <Dandel> I was doing some verification/stability tests and noticed the issue.
[15:16] <Dandel> fglrx-experimental-9 - versoin 9.010 - very buggy release where even glxinfo didn't work right with standard installation procedures. ( but opengl contexts did work )
[15:17] <Dandel> fglrx-experimental-12 - No direct contexts on the 12.100 release.
[15:18] <Dandel> fglrx-12 - also had no direct contexts (12.104 driver release )
[15:18] <Dandel> I forgot to check the fglrx-experimental-13 :/
[15:19] <Dandel> however, the fglrx ( 13.200 ) driver did not have this problem.
[15:32] <RAOF> mlankhorst: Oh, feel free to push that nicely-reviewed linear tiling paranoia patch to xf86-video-ati
[16:03] <bjsnider> Dandel, how dare you suggest fglrx has any bugs
[16:59] <Dandel> bjsnider, fglrx has many bugs... it takes being able to identify the problem to be able to fix it.