bryceh | # Classic DRI and Gallium DRI are mixed up together here | 00:57 |
---|---|---|
bryceh | # Remove the whole tree to avoid false-positives in --list-missing, and | 00:57 |
bryceh | # install the right files manually. | 00:57 |
bryceh | rm -r debian/tmp/dri/usr/lib/dri | 00:57 |
bryceh | dh_install -s --list-missing | 00:57 |
bryceh | cp: cannot stat `debian/tmp/dri/usr/lib/egl/st_GLESv1_CM.so': No such file or directory | 00:57 |
bryceh | dh_install: cp -a debian/tmp/dri/usr/lib/egl/st_GLESv1_CM.so debian/libgles1-mesa/usr/lib/egl/ returned exit code 1 | 00:57 |
bryceh | make: *** [binary-arch] Error 2 | 00:57 |
bryceh | that's after disabling gles overlay and openvg, which krh recommended for mesa to be compatible with wayland | 00:57 |
bryceh | RAOF_, would it be wrong of me to drop GLESv1* from mesa's .install's? | 00:58 |
bryceh | ( http://launchpadlibrarian.net/59576070/buildlog_ubuntu-maverick-i386.mesa_7.10.0%2Bgit20101118.3dcc3153-0ubuntu1~bryce6_FAILEDTOBUILD.txt.gz ) | 00:58 |
bryceh | I'm going to comment out dri/usr/lib/egl/st_GLESv1_CM.so usr/lib/egl in libgles1-mesa.install | 00:59 |
bryceh | RAOF_, btw, I found that even though the wayland server started working after our changes last night, the clients now fail to launch | 01:01 |
bryceh | 01:02 | |
bryceh | errors look like: | 01:02 |
bryceh | Mesa warning: couldn't open libtxc_dxtn.so, software DXTn compression/decompression unavailable | 01:02 |
bryceh | Internal error: Could not resolve keysym SunProps | 01:02 |
bryceh | Internal error: Could not resolve keysym SunFront | 01:02 |
bryceh | Internal error: Could not resolve keysym SunOpen | 01:02 |
bryceh | failed to create cairo drm surface | 01:02 |
bryceh | libEGL debug: Display 0x87f89f8 is destroyed with resources | 01:02 |
bryceh | disconnect from client 0x993db08 | 01:02 |
bryceh | 01:02 | |
bryceh | I'm wondering about the libtxc_dxtn.so bit - that looks odd | 01:02 |
bryceh | however it's the keysym errors that directly precede the termination so I wonder if something's wrong on the input side of things | 01:03 |
bryceh | RAOF_, can you confirm that we definitely do want --enable-glx-tls ? | 01:03 |
bryceh | It looks like we have that on in xorg-server so presumably we want it in mesa too | 01:04 |
RAOF_ | I think we do want --enable-glx-tls, yes. | 01:12 |
Sarvatt | ...this is just going in the PPA right? you're building cairo-drm? yeah we want --enable-glx-tls, no disabling gles and openvg isn't an option in the archive | 01:12 |
RAOF_ | Because ARM at least wants GL|ES & OpenVG, and we probably don't want to drop those libraries anyway. | 01:14 |
bryceh | Sarvatt, yeah just the PPA for now | 01:15 |
bryceh | I'm not building cairo-drm, that's not needed anymore afaik | 01:15 |
bryceh | instead doing cairo-gl | 01:15 |
=== RAOF_ is now known as RAOF | ||
bryceh | hm ok | 01:15 |
bryceh | wonder why krh suggested disabling them? | 01:16 |
RAOF | I don't know. Except that it'll probably take less build time. | 01:17 |
RAOF | I don't see why they would interfere with EGL/GL. | 01:17 |
bryceh | <bryceh> krh, will --enable-openvg cause troubles if it is on? | 01:20 |
bryceh | <krh> bryceh: it wont wroth with --disable-gallium-egl | 01:20 |
bryceh | <krh> s/wroth/work | 01:20 |
bryceh | <bnf> there is --enable-openvg | 01:21 |
bryceh | <Darxus> Hehe. | 01:21 |
bryceh | <krh> yeah, but I'm pretty sure that require gallium egl | 01:21 |
Sarvatt | because he's only made things work on intel with what fedora uses and it gets complicated once you start enabling gallium stuff? :) | 01:21 |
RAOF | Right. But didn't our little expedition into the wonderful world of libudev fix the egl_dri2 problem anyway? | 01:21 |
bryceh | RAOF, yes-ish | 01:22 |
bryceh | it fixed it for the server, which now launches properly | 01:22 |
bryceh | but the clients don't work | 01:22 |
Sarvatt | we have things build-depping on our mesa gles/vg dev packages, I think most of it is arm/linaro specific though outside of mesa-utils | 01:22 |
bryceh | bryce@lynmouth:~/src/Wayland/wayland$ wsmoke | 01:23 |
bryceh | libEGL warning: failed to create DRM screen | 01:23 |
bryceh | Mesa warning: couldn't open libtxc_dxtn.so, software DXTn compression/decompression unavailable | 01:23 |
bryceh | Internal error: Could not resolve keysym SunProps | 01:23 |
bryceh | Internal error: Could not resolve keysym SunFront | 01:23 |
bryceh | Internal error: Could not resolve keysym SunOpen | 01:23 |
bryceh | so I'm poking around for what else needs fixed | 01:23 |
Sarvatt | attach gdb and get a backtrace? | 01:23 |
RAOF | Does it work if you run as root? | 01:23 |
Sarvatt | might just be harmless error spam and its stuck doing something else | 01:24 |
RAOF | It could be a simple failure to authenticate the DRM connection. | 01:24 |
bryceh | RAOF, nope | 01:24 |
RAOF | Sarvatt: I wouldn't think so: “libEGL warning: failed to create DRM screen” looks likely to be fatal :) | 01:24 |
Sarvatt | yeah somehow parsed that out and meant the keysym errors :) | 01:25 |
bryceh | gdb doesn't help as it's not segfaulting | 01:25 |
RAOF | Strace might? | 01:25 |
bryceh | strace: http://pastebin.com/f4qA4nKg | 01:26 |
RAOF | And can EGL be a little more verbose? :) | 01:26 |
bryceh | yes | 01:26 |
Sarvatt | if you attach it should stop the execution until you cont, you can just bt full after attaching? | 01:26 |
bryceh | wait no that's all EGL gives | 01:26 |
Sarvatt | yeah theres tons of egl debug env vars | 01:26 |
bryceh | here's the full run log, showing both server output and client: http://pastebin.ubuntu.com/536116/ | 01:27 |
Sarvatt | EGL_LOG_LEVEL=debug | 01:27 |
bryceh | this is with: | 01:27 |
bryceh | export MESA_DEBUG=1 | 01:27 |
bryceh | export EGL_LOG_LEVEL=debug eglinfo | 01:27 |
Sarvatt | bryceh: do you have X open? | 01:29 |
Sarvatt | did you sudo service gdm stop first I mean? | 01:29 |
RAOF | Hm. Looks like it's correctly opening egl_dri2, which is correctly opening i965_dri | 01:29 |
bryceh | Sarvatt, I have X open | 01:29 |
bryceh | I might have stopped gdm and restarted once since last reboot, I don't remember | 01:30 |
Sarvatt | almost positive you can't have X running when you start it? | 01:30 |
bryceh | untrue; I've had that working fine | 01:30 |
bryceh | in fact I've not yet gotten it running outside X | 01:31 |
bryceh | although I've not yet tried very hard | 01:31 |
* bryceh tries | 01:31 | |
bryceh | whoa that's weird | 01:32 |
bryceh | screen went tie-died and then wiped to black | 01:32 |
RAOF | I suspect the cairo-drm message is the problem | 01:32 |
bryceh | libEGL debug: the best driver is DRI2 (score 100) | 01:33 |
bryceh | Mesa warning: couldn't open libtxc_dxtn.so, software DXTn compression/decompression unavailable | 01:33 |
bryceh | failed to add socket: Address already in use | 01:33 |
bryceh | libEGL debug: Display 0x88999c8 is destroyed with resources | 01:33 |
bryceh | same thing whether running as user or root | 01:33 |
RAOF | Failed to add socket? | 01:33 |
Sarvatt | are you in the video group? | 01:34 |
bryceh | video:x:44:bryce | 01:34 |
bryceh | RAOF, yeah haven't seen that error before | 01:35 |
Sarvatt | you got that error in a paste earlier | 01:35 |
Sarvatt | http://pastebin.ubuntu.com/536116/ | 01:35 |
RAOF | So, I don't particularly think this is a mesa problem; it's loading egl_dri2 properly which is then loading the correct dri driver. | 01:36 |
bryceh | oh wait, socket error just meant I had a stray wayland server process | 01:36 |
RAOF | And it doesn't work, even without a stray wayland? :) | 01:37 |
Sarvatt | tried EGL_PLATFORM=drm EGL_DRIVER=egl_dri2 ./wstart with no X running? | 01:40 |
Sarvatt | well your strace says libEGL warning: failed to create DRM screen is from /usr/lib/egl/pipe_i915.so not existing | 01:41 |
RAOF | Sarvatt: Yeah, but if you read the strace it then goes on to load egl_dri2 and i965_dri. | 01:41 |
bryceh | ok, without a stray wayland it does launch the server, but still no clients | 01:41 |
bryceh | actually fairly cool, I can set a background and get a movable mouse cursor | 01:41 |
bryceh | Sarvatt, yeah those env vars don't seem to make a difference | 01:43 |
bryceh | what generates pipe_i915? | 01:43 |
Sarvatt | are your binaries setuid? | 01:43 |
Sarvatt | i915g | 01:44 |
bryceh | it sounded like from krh that the pipe stuff is used only in certain situations unrelated to this | 01:44 |
bryceh | I haven't fiddled with that at all, should I? | 01:44 |
Sarvatt | yeah just saying you can ignore that warning it looks like | 01:44 |
RAOF | Yeah. The pipe_* drivers are the gallium hardware drivers; they'd be used if egl_gallium was happening. | 01:45 |
bryceh | RAOF, where did you see the cairo-drm bit? | 01:45 |
bryceh | afaik that shouldn't make an appearance anywhere | 01:45 |
RAOF | write(2, "failed to create cairo drm surfa"..., 35failed to create cairo drm surface | 01:45 |
Sarvatt | <bryceh> failed to create cairo drm surface | 01:45 |
bryceh | oh right | 01:46 |
RAOF | You could play with i915g; I understand that it's not entirely broken, like i965g is :) | 01:46 |
bryceh | that's an error message from the client that it gets after trying to create a surface using cairo | 01:46 |
bryceh | hrm | 01:46 |
bryceh | ok if you guys are out of clue on this too, I'll just shelve it until next week | 01:47 |
bryceh | happy to try more idea though if you got 'em | 01:47 |
RAOF | I think the interesting question is why cairo's failing to create a drm surface. | 01:48 |
Sarvatt | yeah looks like the problems in the way cairo is built to me even though cairo-drm isn't enabled | 01:49 |
bryceh | hm ok | 01:53 |
bryceh | alright, lastly, gonna try newer wayland in case recent git changes fixed something | 02:00 |
bryceh | WHOA | 02:00 |
bryceh | that did it | 02:00 |
bryceh | sweetness | 02:00 |
RAOF | Heh | 02:00 |
bryceh | shoulda tried that earlier! | 02:00 |
Sarvatt | http://cgit.freedesktop.org/wayland/commit/?id=32ff69017ab003911b754982772d0644b1cd23d4 | 02:01 |
RAOF | So that sounds like we can use the archive mesa configuration and still have everything work. | 02:01 |
Sarvatt | hah was my next idea | 02:01 |
Sarvatt | (are you *sure* the udev rules are getting used?) | 02:01 |
bryceh | for the record I still see these warnings when firing up the clients: http://pastebin.ubuntu.com/536127/ | 02:02 |
bryceh | Sarvatt, I express certainty on nothing | 02:02 |
bryceh | oh | 02:02 |
bryceh | on that, I had been forcing install of the udev rules in debian/rules anyway | 02:03 |
bryceh | I can probably clean some of that up now | 02:03 |
RAOF | Yeah; the dxtn library warning is totally harmless. I guess unless a wayland client wants to upload DXT-compressed textures :) | 02:04 |
bryceh | would be nice to suppress that | 02:04 |
bryceh | I've a pet peeve about harmless-but-technobabbly warnings | 02:05 |
Sarvatt | you could install it :) | 02:05 |
bryceh | actually not sure that I can | 02:05 |
RAOF | You know what would be a good fix? Show that warning when something tries to fiddle with a DXT texture! | 02:06 |
Sarvatt | yeah you can I just did on a new install a few days ago | 02:06 |
bryceh | bryce@blumonc:~/src/Wayland/wayland/wayland-0.1~git20101124.32ff6901/debian$ apt-cache search dxtn | 02:06 |
bryceh | bryce@blumonc:~/src/Wayland/wayland/wayland-0.1~git20101124.32ff6901/debian$ | 02:06 |
bryceh | is it a natty package? | 02:06 |
Sarvatt | its patent encumbered, have to do it manually | 02:06 |
bryceh | oh, well I don't count that ;-) | 02:06 |
bryceh | well, I'd say either quell the warning or add the installation directions to it | 02:07 |
bryceh | anyway, task for another day | 02:07 |
bryceh | Sarvatt, where do you get the package? | 02:07 |
Sarvatt | do you have LIBGL_DEBUG=verbose exported maybe? | 02:08 |
Sarvatt | digging around for it now, it was on people.freedesktop.org that got nuked awhile back and had to grab it off bugzilla somewhere | 02:10 |
bryceh | no, but I do have MESA_DEBUG=1 exported | 02:10 |
Sarvatt | oh thats why ya got it i bet | 02:10 |
bryceh | export MESA_DEBUG=1 | 02:10 |
bryceh | export EGL_DRIVER=egl_dri2 | 02:10 |
bryceh | export EGL_PLATFORM=drm x11 | 02:10 |
bryceh | export EGL_LOG_LEVEL=debug eglinfo | 02:10 |
Sarvatt | export EGL_LOG_LEVEL=debug eglinfo ? | 02:11 |
bryceh | hmm commenting out both MESA_DEBUG and EGL_LOG_LEVEL but still prints it | 02:12 |
Sarvatt | http://debian-multimedia.org/pool/main/libt/libtxc-dxtn/libtxc-dxtn.php | 02:13 |
Sarvatt | debs! | 02:13 |
bryceh | ok thanks | 02:14 |
* Sarvatt wonders if he can get away uploading that to xorg-edgers :) | 02:14 | |
bryceh | if I get around to it I'll isolate where the warning comes from and add that link, if I don't just quell it outright | 02:14 |
RAOF | Sarvatt: It's free software, right? | 02:14 |
bryceh | is debian-multimedia.org part of Debian? | 02:14 |
Sarvatt | heavily patent encumbered and not suitable for debian :) | 02:15 |
Sarvatt | its debian's medibuntu | 02:15 |
Sarvatt | http://dri.freedesktop.org/wiki/S3TC | 02:16 |
ScottK | Sarvatt: only with lower quality packaging. | 02:18 |
Sarvatt | lower quality? | 02:19 |
ScottK | The debian-multimedia packaging quality is not well thought of in Debian. | 02:19 |
* Sarvatt can't even count how many times he's had broken ffmpeg libs screw up his system from medibuntu | 02:19 | |
bryceh | gotcha | 02:20 |
ScottK | AFAIK, most of the people working on medibuntu are actual Ubuntu devs so I think it's similar to Ubuntu package quality wise | 02:20 |
ScottK | It's been quite some time since I actually paid attention to medibuntu though. | 02:20 |
bryceh | okie dokie guys, get your wayland on: https://launchpad.net/~xorg-edgers/+archive/wayland | 02:22 |
bryceh | IN THEORY, you should be able to just add the ppa, update, install wayland, and then run wayland as described | 02:22 |
bryceh | if not, it's something I should fix | 02:23 |
Sarvatt | ppa-purge no workie in natty, but at least there aren't many packages involved :) | 02:23 |
bryceh | no? why not? | 02:23 |
bryceh | anyway, this is only tested on maverick so far | 02:23 |
Sarvatt | sources are stored compressed now, haven't had a chance to make ppa-purge work with that | 02:24 |
Sarvatt | /var/lib/apt/lists/ppa.launchpad.net_xorg-edgers_wayland_ubuntu_dists_natty_main_binary-i386_Packages.gz | 02:25 |
bryceh | ah | 02:26 |
bjsnider | the medibuntu ffmpeg is almost exactly the same as the ubuntu ffmpeg. hard to believe it could have broken anything | 02:35 |
tjaalton | bryceh: tried the ppa, but looks like wayland needs to depend on libgles2-mesa | 07:52 |
tjaalton | since it can't find libGLESv2.so.2 without it | 07:53 |
RAOF | And it's looking for GLESv2? I thought it was strictly a desktop GL affair. | 07:53 |
vish | RAOF: hi, alacarte upstream is unmaintained.. spoke to Amaranth earlier, he said we can just upload the patch in natty.. re: Bug 395692 | 07:54 |
tjaalton | ah | 07:54 |
ubot4` | Launchpad bug 395692 in alacarte (Ubuntu) (and 2 other projects) "Drag-and-Drop behavior in the menu editor is inconsistent and confusing (affects: 3) (heat: 28)" [Low,Triaged] https://launchpad.net/bugs/395692 | 07:54 |
RAOF | Also, I guess that's a bug in mesa. | 07:54 |
tjaalton | RAOF: yeah I'm trying to run it with the hints from the ppa | 07:54 |
vish | s/natty/ubuntu | 07:54 |
tjaalton | now the compositor runs, but how do I fire up the clients?-) | 07:54 |
RAOF | vish: Yeah, I guessed as much. I was going to go sponsor hunting this evening. | 07:55 |
vish | RAOF: neat! thx :) | 07:55 |
tjaalton | is there a way to launch wayland without X | 08:04 |
tjaalton | ? | 08:04 |
bryceh | tjaalton, there may be a /usr/bin/wstart script | 08:04 |
tjaalton | bryceh: yeah, it just says "no drm device found" | 08:05 |
bryceh | tjaalton, and there are clients named wayland-* you can run if you have the server up | 08:05 |
tjaalton | so maybe something else is wrong | 08:05 |
bryceh | yeah likely | 08:05 |
tjaalton | bryceh: also, it didn't pull libxkbcommon | 08:05 |
bryceh | ok | 08:05 |
* bryceh makes notes | 08:05 | |
tjaalton | the ppa description has old client names, btw (wterminal..) | 08:05 |
bryceh | oh thanks, yeah just redid the naming today | 08:06 |
bryceh | tjaalton, updated | 08:08 |
* bryceh -> bed | 08:08 | |
tjaalton | yeah, night | 08:08 |
=== JanC_ is now known as JanC | ||
=== yofel_ is now known as yofel | ||
=== |Alexia_Death| is now known as Alexia_Death | ||
RAOF | Well, launchpad indicates that mesa doesn't appear to be on fire. That's nice :) | 21:50 |
bryceh | RAOF, should it? | 22:05 |
RAOF | No, of course not :) | 22:06 |
bryceh | RAOF, oh btw I wanted to ask, you mentioned you'd pulled some of the changes for wayland-enabled mesa into the archive... | 22:06 |
bryceh | RAOF, which changes exactly did you pull, and was that into the main archive or the xorg-edgers archive? | 22:06 |
RAOF | I pulled the libudev-dev build-dep into the main archive packages. | 22:07 |
RAOF | So egl_dri2 should actually work. :) | 22:07 |
RAOF | The reasons why Mesa might have caught fire are the switch to r300g by default and libdricore shared linking. | 22:08 |
bryceh | ah gotcha | 22:09 |
bryceh | I've been watching all incoming natty bug reports fairly carefully the last couple weeks | 22:09 |
bryceh | the sense I get is that very few people are running natty at the moment, cause there's hardly any bug reports coming in | 22:09 |
bryceh | http://www2.bryceharrington.org:8080/X/Reports/ubuntu-x-swat/totals-natty-workqueue.svg | 22:10 |
bryceh | there's 1 natty bug against -evdev and one against nvidia-173 | 22:10 |
RAOF | Ah. Mesa hasn't caught fire on people's systems because it's failed to build on the buildds. Thanks, crappy mesa build system! | 22:39 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!