[02:16] pwnguin: just one of the monitors ? not both [02:16] if that's the case, I might move to that driver [02:17] I like the proprietary driver though because of GLX [02:19] bhuey: im fairly certain ive rotated my tablet and the vga output remained the same [02:19] you might be able to do that with nvidia [02:19] do you have instructions or is it pretty much something that you play with when you install the drivers and run the screen configuration utilities ? [02:19] I was able to do that to the entire scene, but not the individual monitor [02:20] and a Google search yielded nothing informative regarding this [02:23] what you need to do is read the very long and boring nvidia manual in /usr/share/doc/nvidia-glx with an eye towards xrand [02:23] r [02:23] hmmm, ok [02:36] there's no info on rotating just the screen but the entire view itself [03:33] bhuey: well you could always try out raof's stuff and see if you like it [04:22] pwnguin: does he hang out here ? [04:22] raof? yes [04:22] Man, it takes a long time to build an ubuntu kernel. [04:22] technically, nouveau is experimental [04:23] so if it fails for you, don't get super angry or expect anything [04:23] Also, if you've got a geforce 8 or 9 card, use a compositing manager, or a bunch of stuff won't work. [04:26] RAOF, are you building with debuild -b, or did you follow the simplified steps? [04:26] the normal debian packaging will build a ton of variants you probably wont care about [04:26] VARIOUS_ENVIRONENT_VARIABLES fakeroot debian/rules binary-mmiotrace [04:26] IE: I'm just building the flavour I'm interested in. [04:26] okay good [04:27] There just happens to be one metric buttload of modules to build. [04:27] I'll probably farm it out to a PPA once I'm reasonably sure it works; having an MMIO trace capable kernel would be useful for a number of people. [04:28] i'm not familiar with MMIO trace, what is that? [04:28] its a reverse engineering tool [04:29] for charting memory mapped I/O [04:29] "Kindly log every memory-mapped IO operation performed by this kernel module, so I can see what it does" [04:29] Ah, i see [04:29] i think you can guess which module this might be applied to [04:29] fglrx :P [04:29] perhaps it starts with n and rhymes with cydia [04:30] pwnguin: i'll wait for him to come on sometime and ask him how to go about doing this [04:30] bhuey: You mean raof? :) [04:30] is there a performance hit in operating such a kernel (by the overhead of logging all that data)? [04:31] btw, I love that you folks have done all of this work for Ubuntu [04:31] RAOF: same character ? [04:31] ok, I can see lower case like I use to :) [04:31] superm1: There's a huge performance hit when you're actually tracing; X startup takes in the order of minutes. [04:31] RAOF: yeah, I'd appreciate some pointers as to how to rotate just one of my screen so that I can use it in portrait mode of sorts. [04:31] RAOF, but the tracing can be enabled and disabled on demand? [04:31] superm1: And apparently even when it's not active there's a couple of MB overhead per CPU. [04:32] I can rotate the entire virtual view but not just one monitor. I'd like to know how to rotate just one monitor [04:32] Yeah. Tracing can be enabled & disabled at will. [04:32] if it is possible [04:32] bhuey: With nouveau? [04:32] RAOF: any driver [04:32] I'm using nvidia right now [04:32] but with that overhead on a per cpu basis, definitely not a good idea to have available on a production kernel [04:32] until nouveau gets full glx support, it's not a viable option for me [04:32] superm1: Right. Otherwise I'd just be asking the kernel team to enable it. [04:33] RAOF, well if it aides upstream development significantly, you can possibly ask the kernel team to set up a PPA that spits out kernels with it enabled like the mainline builds do for upstream kernel releases [04:34] Possibly. Or I can just do what I'm doing, and shove a -mmiotrace flavour in my nouveau PPA. [04:34] If that turns out to be too hard, I'll ask the kernel team. [04:34] right [04:51] RAOF: do any of the drivers to that ? [04:51] bhuey: Oh, sorry. [04:51] bhuey: Yes. Any driver that supports XRandR 1.2 will do that. [04:51] yeah, but for just one screen [04:51] The nvidia driver *doesn't*, and I'm not sure if you can. [04:51] I've got a dual screen setu [04:51] setup [04:52] The nvidia driver doesn't support XRandR 1.2. [04:52] I can rotate it, but only the entire image [04:52] If you can't do it in nvidia-settings, you can't do it. [04:52] ok, that supports multiple screens then right ? [04:52] ok [04:52] but noveau does this ? what about glx support ? [04:52] Nouveau does this. There's no 3d support, however. [04:52] nouveau [04:52] hmmm [04:53] I'll have to suffer with the nvidia driver then [04:53] (That's not entirely accurate: there's 3d support ranging from "can play OpenArena flawlessly" to "glxgears causes kernel panic". The drivers switch from one state to another on a semi-regular basis) [04:54] glx support is critical [04:54] how's progress with that ? do you know ? [04:54] Slow. [04:54] a quick note, bhuey's video card: 6600gt =) [04:54] yeah, it's hard stuff [04:54] pwnguin: thanks :) [04:54] I really appreciate this help btw [04:55] pwnguin: Yeah, I remembered. I think the *current* state of glx on his card is "glxgears will show artefacts the first time you run it, will kill X the second". [04:55] great [04:55] bhuey: it seems like you have a choice mainly, between "smooth 3d graphics" and "rotates individual screens" [04:55] the nvidia driver wins [04:55] then i guess we [04:55] then i guess we're done here [04:55] ok [04:55] until the next release :) [04:55] thanks [05:55] * RAOF wouldn't expect supported 3d with nouveau in Karmic, either. [05:55] Although I would hope it to be the default driver for nvidia cards. [06:26] Well, I hope the PPAs don't mind building the kernel. I wonder how many hours it'll take. [06:28] RAOF, i know when rtg and those guys push to ppas, they have a way to make sure the ppa won't build all the flavours too, so unless you fully hacked up debian/control to only have your tracable flavour, you might want to check about that [06:34] Whoops. I think that'll build all the flavours. [06:36] Oh, well. I now know how to make it _not_ build all the flavours. Next upload won't build everytihng :/ [07:29] * RAOF plays "upload another 90Mb of kernel source to the ppa, this time it'll definitely build only -mmiotrace" [07:33] well at least you got it fixed now rather than 10 hours from now :) [07:43] RAOF: if Fedora is getting nouveau as default in 11, I'd guess Karmic could as well have nouveau as default... it even supports more cards than nv [07:44] it'd be great, though, I like the message that it's delivering to NVIDIA [07:46] I'm pretty sure it will be chosen as default.. [08:27] Mirv: i doubt it'll really make an impact on usage numbers [08:27] glx is basically vital [08:58] o/ [08:59] I registered a UDS session to discuss defaulting to -nouveau [09:09] bryce: do you still need more testing/reproduction data points on bug 359392 ? [09:09] Launchpad bug 359392 in xserver-xorg-video-intel "[i965] X freezes starting on April 3rd" [Critical,Triaged] https://launchpad.net/bugs/359392 [09:50] ttx: couldn't hurt to have your intel_gpu_dump output for reference [09:50] ttx: assuming your symptoms, hw, etc. are a match with that bug. If not, you should collect that info, but report as a separate bug report [09:50] bryce_: ok, will make my best reproducing it. I'm like seb128, I wasn't ever bitten by it. [09:51] * bryce_ --> bed [09:51] night [16:31] bryce: i doubt I'll be in attendence, but i fully support the idea [17:09] bryce, mdz: just to confirm, are you still getting interrupts when the freeze happens? and does the "current sequence" increment in i915_gem_interrupt? [18:15] jbarnes: lemme check, I got another freeze yesterday [18:16] bryce: thanks [18:16] oh... you mean while the system is frozen? I'll have to re-freeze it first. Will get back to you [18:16] yeah while frozen [18:17] get some good ideas on fixes over the weekend? [18:17] unfortunately no :( [18:17] given the gpu dumps you guys gave me I was hoping it was a render accel problem [18:17] but at least some reporters tried that and still saw hangs [18:17] of course that doesn't mean render accel isn't broken, just that it can't be the whole fix [18:32] jbarnes: hm bummer, I was hoping the new tools would make these bugs completely transparent. But I guess it's like having a backtrace - it helps capture the state but not what caused it to get into that state [18:33] yeah for certain kinds of hangs they can really help [18:33] for others they just give us a few more clues and things to eliminate [18:34] the last hang info you posted actually seemed ok... I asked anholt to confirm [18:34] (ok as in we're not failing to track sequence numbers it looks like) [18:34] jbarnes: that was my "classic" hang that occurs after several hours use, as opposed to the freeze that occurs from mdz's script [19:10] heya apw [20:05] bryce: also, if the kernel you're running also has my i915 error interrupt patch you could check dmesg when the hang happens [20:18] * cwillu 's been away for a little while [20:19] should stock jaunty still have lingering hangs? [20:28] cwillu: yeah [23:53] tjaalton: you around?