[10:06] ricotz: I guess your segfault isnt nouveau specific - http://lists.freedesktop.org/archives/mesa-dev/2010-May/000865.html [10:22] Sarvatt, oh, thanks, this sounds promising, i will try this patch later [10:25] i just told airlied about you having it on nouveau too and the good and bad commits and passed along your backtrace in #dri-devel, might want to join there in case they discuss the problem :D [10:26] i dont see the problem on i915g, seems to be limited to nouveau and radeon [10:37] ohh i'm wrong [10:38] texture_from_pixmap in mesa/demos segfaults the same way on i915g [10:45] http://pastebin.com/Z3UKydSM [12:19] jcristau: do you have a script for converting things to xtrace .proto's or did you do dri2 by hand? [12:55] Sarvatt: did it by hand [12:55] from dri2proto.{h,txt} [12:56] ah ok, wasn't sure if i was missing out on some magic script before looking into updating some of the protos :) [13:49] Sarvatt: yeah all the gallium drivers share the same dri code. [15:57] hmm, doesn't glproto need to be updated following http://cgit.freedesktop.org/xorg/proto/glproto/commit/?id=f2b9a6a29edf930f30eade7a0abe40a6d3c4962b ? [15:58] * Sarvatt is updating some things in xtrace's glx.proto [15:58] err sorry I meant http://cgit.freedesktop.org/mesa/mesa/commit/?h=7.8&id=42ea25cb4ecae09b5cc011a95d42ba7f0645dde3 [15:58] GLX_BLIT_COMPLETE_INTEL was changed to GLX_COPY_COMPLETE_INTEL but its still BLIT in glproto [16:03] what the heck am I talking about, no glxext.h in glproto! [16:05] #define GLX_BLIT_COMPLETE_INTEL 0x8181 in glxtokens.h though [16:34] well sent the patch for it to the lists at any rate, it's COPY instead of BLIT on the official spec - http://www.opengl.org/registry/specs/INTEL/swap_event.txt [16:35] wonder what that screws up [16:36] clutter is using BLIT... === JanC_ is now known as JanC [17:03] oh hell, need to add 65k placeholders to put all glx opcodes in xtrace?! [17:05] wanted to put the GLX_EXT_texture_from_pixmap ones in there at least but those are 1k in even :( [17:06] guess i'll have to add logic to it to recognize something like SKIP 1000 in the lists [17:10] hi [17:12] heyo Bernardo [17:12] hi Sarvatt [17:12] how goes psb by the way? an old friend was asking me about it and i'm not sure how far you guys got it [17:13] might try to convince him to swap his psb netbook for my aspire one that I've used as my main machine for 1.5 years now, dont think i'm right in the head :) [17:13] I got stuck trying to "free" the psb driver from the moblin or meego repositories [17:13] I've swapped my A1 for my 1101HA... :) [17:14] yeah its gonna need a bunch of work, i'd just wait for the meego release with mrst support [17:14] hopefully its not too far off, they didnt release it yet.. its the handset variant [17:14] Yes, they dropped psb from the main release [17:14] the interesting part is that they are merging psb with pvr [17:15] i think it was only in ivi before [17:15] wasn't in the main netbook one, pretty sure of that [17:15] it was in the test builds in the main core [17:15] which aa1 did you have? [17:15] the first one, 8GB slow SSD [17:16] ah yea thats what i have except the hdd model, i love this thing to death [17:16] which I later replaced with a 32GB CF [17:16] got 2 10 hour batteries for it and it fits in a coat pocket :) [17:16] It is very nice, but the 7 hour battery is huge [17:16] I now get 6h +- on the 1101 [17:17] and a lovely 1366x768 screen [17:18] As for meego/psb, my main problem was that the driver is a large kernel patch, and splitting it to build as dkm module is taking a lot of work. I hadn't done any programming in 10+ years, and it shows [17:20] Sarvatt: did you try building the xorg driver? [17:20] yeah i haven't been able to justify screwing with it without an actual machine to try it on :) [17:20] Bernardo: if you do use the meego psb x driver be sure to configure it with --disable-dri2 [17:20] (dri2 needs gallium) [17:20] Yes, i tried passing it to the build. But I get a strange error [17:21] why dont ya try booting that moblin image with psb on it from january and extract all the blobs it needs from there? i couldn't find it in any source packages but its probably named funky [17:21] what error? [17:21] shouldn't be hard to fix, look at other drivers in git during the transition to xorg 7.5 [17:22] should just need the xext check and wrapping some headers behind xserer version check ifdefs, can see what to do from the psb patches you have packaged up now [17:23] no, that I did [17:23] wait [17:23] oh good clutter checking for GLX_BLIT_COMPLETE_INTEL is wrapped in an #if 0 so that doesnt matter [17:24] I'll turn on the netbook to see if I can reproduce the error [17:25] oh you got an error after it built? thought ya meant compiling it [17:28] no, building it [17:28] in a header [17:28] I'll just try to reproduce and pastebin it [17:31] oh yea need to build against crap you dont want to install on a non psb system, gotcha :) [17:31] :) [17:31] also this machine is x86-64 [17:31] so I'd need a chroot [17:38] ok, I'm building now. Hopefully it won't take much to get to the build error [17:42] http://pastebin.com/1DeAXKx2 is the error, and http://pastebin.com/0iFw3AER is the header it complains of [17:42] and my limited c knowledge doesn't let my find any error in that header [17:43] oh [17:43] you dont use libdrm-poulsbo anymore with those packages [17:43] I added line 582 jut to try to find it [17:43] no, but moblin provides its own psb headers [17:43] you need the psb-kernel-source srpm headers [17:43] yeah, you're using those? [17:43] yes, I used those [17:44] unless things got corrupted, I'll try reinstalling the deb I did for them [17:44] psb-headers: /usr/include/psb_drm.h [17:44] should be right, but... [17:44] are the headers from the same release you grabbed the ddx from? [17:44] the ones out of the srpm i have handy are #define PSB_PACKAGE_VERSION "5.1.0.32L.0120.1" [17:45] yours is #define PSB_PACKAGE_VERSION "5.1.0.32L.0124" [17:45] yes, they are all from moblin, built from the moblin repository [17:46] they have at least 10 different repos though [17:46] its so darn convoluted [17:46] :( [17:46] I know [17:46] the headers were from psb-headers-1.0.13-1.1.moblin2.src.rpm [17:46] whats the xf86-video-psb version? [17:47] xf86-video-psb-5.1.0.0112-1.3.moblin2.src.rpm [17:47] argh my headers were 1.0.12 and from the latest meego, wtf [17:47] forgot i had moblin stuff in another directory [17:48] I looked at meego, but the 0.9 builds had only the kernel and the headers, not the xorg driver [17:49] did you build libwsbm too? [17:49] yes [17:49] to prefix /usr? [17:49] how about if you use --enable-dri2? [17:50] "/usr/lib/libwsbm.so.1.1.0" [17:50] "/usr/lib/libwsbm.so.1" [17:50] did my deb from the libwsbm-1.1.0-3.11.src.rpm [17:51] ok, time to try that [17:51] have a /usr/lib/pkg-config/libwsbm.pc? [17:51] err pkgconfig [17:51] i'd just remove that and install it locally [17:53] nah it aint libwsbm anyway [17:53] yes, I also created a dev package with that, the header files and /usr/lib/libwsbm.so (from the rpm spec) [17:53] enable-dri2 doesn't change, it fails the same way [17:56] doh [17:56] #define OV_REGRWBITS_OVADD (1 << 0) [17:56] then later uint32_t OVADD; [17:57] I wonder if the moblin guys ever tried to compile this driver or checked these headers [17:57] do you have any psb junk in /usr/include/drm or /usr/include/X11/dri? [17:58] (i'd uninstall libdrm-poulsbo too) [17:58] no, nothing there [17:58] let me try removing libdrm-poulsbo then [17:59] hmm your psb-drm.h is different than mine from the same source package? [18:00] struct { [18:00] uint32_t OVADD; [18:00] uint32_t OGAMC0; [18:00] uint32_t OGAMC1; [18:00] bah nevermind i'm looking at the meego ones still [18:00] no, i'm not actually [18:00] I still don't know why gcc doesn't like that uint32_t OVADD; [18:01] I added a line before that, to see if it was some non-printing char between { and uint32_t OVADD; [18:01] you pastebinned this [18:01] struct { [18:01] uint32_t maria_amelia; [18:01] uint32_t OVADD; [18:01] uint32_t OGAMC0; [18:01] different from mine [18:01] the unedited file is http://pastebin.com/gLcKnJV8 [18:02] oh ok [18:02] Still fails the same way, without libdrm-poulsbo [18:03] I need to add a svn or bzr branch with what I've done so far, so that someone competent can look at it [18:03] think i'll set up a chroot to test, mind is going numb from working on xtrace all morning [18:03] brb [18:03] :) [18:03] Bernardo: can you send me all your source? [18:03] wife's calling [18:03] Sarvatt: yes, pm me and I'll tar it and send it to you in 30m [18:03] okie, need to set up a new chroot anyway === ripps_ is now known as ripps [21:14] ok, something went dramatically wrong with 10.04 nvidia the other day [21:15] as in glx is busted and causes gdm / X11 to segfault wrong [21:15] i will say, its nice that i network manager works on the command line [21:16] i'm going to blame 195.36.24-0ubuntu1~10.04 0 500 http://archive.ubuntu.com/ubuntu/ lucid-proposed/restricted Packages [21:19] what does glxinfo give you? [21:20] right now? [21:20] mesa because i installed all that crap [21:20] uninstalled [21:20] i mean, i hope it's mesa [21:20] heh [21:20] glxinfo is not installed =P [21:21] server glx vendor string: SGI [21:21] server glx version string: 1.2 [21:21] does glxinfo work without X11 running? [21:23] server glx vendor string: SGI [21:23] server glx version string: 1.2 [21:23] doh [21:23] http://pastebin.com/zXM8DZSv [21:26] maybe i'll give nouveau another spin [21:31] what card is this? [21:33] 6600gt [21:33] did i get hit by another round of obseletion? [21:35] it's on the backside of the supported cards list [21:35] i heard once that the 6600 series had some issues but that was a long time ago [21:36] you'd get opengl 2.1 probably instead of 3.3 or 4 if you have a newer card [21:38] the thing to do is to install the blob and troubleshoot any issue that arises, not to put nouveau in there because you don't feel like having to do so [21:52] pwnguin: if glx was busted the alternatives just werent set up right most likely, might want to just try reenabling it in jockey and it'll probably go away :D [21:52] hah Bernardo, moblin psb doesnt even compile on moblin, same error [21:54] it makes it one more file before failing with the same error [21:54] /bin/sh ../libtool --tag=CC --mode=compile gcc -DHAVE_CONFIG_H -I. -I.. -I/usr/include/xorg -I/usr/include/pixman-1 -I/usr/include/drm -I/usr/include/drm -I/usr/include/X11/dri -I/usr/include/wsbm -Wall -I.. -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m32 -march=i386 -mtune=generic -fasynchronous-unwind-tables -MT psb_ws_driver.lo -MD -MP -MF .deps/psb_ws_driver.Tpo -c -o psb_ws [21:54] _driver.lo psb_ws_driver.c [21:54] In file included from psb_dri.h:37, [21:54] from psb_driver.h:70, [21:54] from psb_dga.c:12: [21:54] /usr/include/psb_drm.h:582: error: expected identifier or '(' before numeric constant [21:54] /usr/include/psb_drm.h:583: error: expected ';' before 'uint32_t' [21:54] ok thought that was less lines, sorry :) [21:58] Sarvatt: interesting [22:03] is that just sloppy code? [22:03] they left out a semicolon? [22:09] Sarvatt, http://www.nvnews.net/vbulletin/showthread.php?t=151479 [22:09] that's a spicy meat-a-ball [22:09] rawr! [22:10] complete file layout overhaul again right? :) [22:10] yeah, i'm sure [22:10] that would be funny [22:10] just run into nvidia's headquarters and yell "is there a psychiatrist in the house?" [22:11] so whats fixed? [22:11] that powermizer issue [22:11] * Sarvatt is too busy packaging to look [22:11] ah [22:12] that's all they put in the changelog, but they leave things out of it these days, so there could be more. [22:15] oh i didnt even notice that was all the put about it, the one line blended in with the rest of the crap [22:17] yeah but they have gotten into the habit of leaving things out [22:18] hmm they removed the dithering controls from nvidia-settings too [22:18] remember when they used to put 'improved compatibility with recent kernels' in there? that helped us figure out when we could drop the kernel patches [22:19] lol [22:19] i think they stopped because they actually keep on top of it now :) [22:20] nope [22:21] well not as bad as it used to be at any rate, nothing like fglrx :) [22:24] it was never that bad [22:24] they are always compatible with kernels in use in the distros, yes. but they are not always compatible with newly released kernels [22:27] "update-alternatives: warning: skip creation of /usr/lib32/vdpau/libvdpau_nvidia.so.1 because associated file /usr/lib32/nvidia-current/vdpau/libvdpau_nvidia.so.1 (of link group gl_conf) doesn't exist." [22:29] yeah thats just a harmless message [22:29] well you mentioned alternatives so i just wanted to make sure [22:29] bjsnider: for 2.6.29 it took them well into 2.6.30 to catch up if i remember right [22:29] err 2.6.28 i mean [22:30] well, i'll reload and see if anything magically is fixed [22:30] pwnguin: sudo update-alternatives --config gl_conf [22:30] is nvidia-current picked? [22:31] i have two [22:31] one is manual ode and one is auto mode [22:31] there's another problem with the nvidia packaging that i've seen, all the headers are in /usr/include/nvidia-whatever/ and no alternatives back to /usr/include/ [22:32] is one of them picked? [22:32] has a* [22:32] yep [22:32] yea just dont change it then, you're good to go after a reboot [22:32] i hope [22:32] lets find out [22:32] brb [22:32] unless you did something else like install fglrx... [22:35] If intel_iommu=off prevents the crash, then this is a bug which should be fixed in the next release. [22:35] oh boy :) [22:36] new releases out the wazoo [22:39] im not crazy [22:40] i know its hard to tell right now, but im not generally clueless [22:45] its not working? [22:46] not yet but i'll get it there. lemme try a few things first [22:46] or was that in response to the fglrx comment.. I'm slow :) [22:46] fglrx [22:47] this box is going to be a bit quirky [22:47] its been dist-upgraded since warty [22:47] wow those new nvidia blobs built fast in x-updates [22:48] alright, i gave jockey a chance to configure things as it sees fit, we'll see if its smarter than me [22:48] brb again [22:48] did you make any changes in /etc.X11/xorg.conf? [22:48] before when the crashes started [22:51] well here's the deal [22:51] the desktop hardlocked [22:51] so i rebooted [22:51] came back and decided it might be video playback and checked for updates to banshee and nvidia [22:52] low and behold, nvidia had updates [22:52] updated from proposed and when i came back it tanked hard. [22:52] hmm mesa update in lucid-updates? [22:52] it was all kinds of messed up [22:53] dpkg's available packages file was nicely corrupted or something [22:53] so apt-get wasn't working on install [22:53] then the filesystem crashed and went readonly [22:54] so im not really ready to blame anything at the moment other than me or my old computer [22:55] anyways, i moved my xorg.conf out of the way and removed all the nvidia components, and then gdm stopped crashing [22:56] if i can reproduce the problem then it's just the worlds least fortunate series of events [23:04] ah, sweet relief [23:04] server glx vendor string: NVIDIA Corporation [23:04] server glx version string: 1.4 [23:04] fonts look odd though [23:06] maybe it's just an illusion from dealing with crappy resolution [23:06] Sarvatt: thanks for the tips and attention [23:06] worst dist-upgrade evar [23:07] and thats including the time someone uploaded a busted nvidia that spwaned the invention of -proposed [23:30] looks like i find some other kind of bug with busted nvidia, just spent the past hour fighting that machine while the wife waited to use it.. lol [23:31] activate nvidia-173, reboot, purge nvidia-173 and upgrade kernel [23:31] wouldn't let me install nvidia-current again, i was getting those funky nvidia-173 already removed messages in the jockey log some guy had in here the other day [23:33] like it updated the wrong initramfs and left 173 stuff in the old one [23:39] does your wife prefer linux or some other operating system? [23:41] shes using one of my laptops now so she doesnt get a choice :D only thing she misses is photoshop though [23:44] photoplop