Q-FUNK | bryce_: did OLPC's Bernie manage to find you? | 13:36 |
---|---|---|
bryce_ | yes, we chatted a bit | 13:37 |
bryce_ | more about inkscape than -amd though | 13:37 |
Q-FUNK | ah | 13:37 |
bryce_ | I talked with ogra about -amd briefly as well - he said "the current -amd fails to work for 80% of the people, and that we should go to the previous -amd, which worked fine." | 13:38 |
bryce_ | I asked him if he'd talked to you about this, and he said no, so I told him he needed to speak with you about it first. | 13:38 |
Q-FUNK | it's not a question of previous -amd as much as current X | 13:38 |
bryce_ | it was the first I'd heard of an issue | 13:38 |
Q-FUNK | ogra fails to see the big picture: X since 7.2 leaves many drivers that work just fine with 7.1 in the dust, -amd being just one of many. | 13:39 |
Q-FUNK | it gets even worse with 7.3 | 13:40 |
Q-FUNK | modularisation allows upstream X to crank out new releases of the core more often, but this comes at the price of chipset drivers requiring more frequent updates. | 13:41 |
Q-FUNK | for widespread drivers that sometimes include paid coders from the manufacturer, such as ati, mga, intel and nvdia, this works well. | 13:42 |
Q-FUNK | for all other chipsets, their stability falls behind one API/ABI chunk at atime. | 13:42 |
* bryce_ nods | 13:44 | |
bryce_ | so it's a case that even if we went back to the previous -amd, it'd still be broken? | 13:44 |
bryce_ | that is sort of what I suspected | 13:45 |
Q-FUNK | yup | 13:49 |
Gadi | bryce: ping | 13:50 |
Q-FUNK | I tried going back to really old CVS snapshots and I get the same result: the X core simply has changed in too many ways startign with 7.2 | 13:50 |
Q-FUNK | 7.2 was sort-of compatible with 7.1, but 7.3 definitely isn't | 13:51 |
Q-FUNK | bryce_: Gadi here has his ThinCan with him at UDS | 13:51 |
Q-FUNK | bryce_: can you introduce him to Bernie? | 13:52 |
Q-FUNK | bryce_: and sbalneav is the one that merged -amd support for edubuntu. he has a ThinCan too, though not with him at UDS | 13:53 |
* Gadi is sitting along the window with Dave Trask atm | 13:54 | |
Gadi | and I have to leave in a few mins | 13:54 |
Gadi | so.... | 13:54 |
Q-FUNK | oh... | 13:55 |
Gadi | i am happy to hand off the thincan | 13:55 |
Gadi | I am meeting up with the rest of the LTSP gang in Maine tomorrow | 13:55 |
Gadi | and can get it back then | 13:55 |
Q-FUNK | ok | 13:55 |
Q-FUNK | that could work well too | 13:55 |
Q-FUNK | does the LTSP gang ever gather up in EU? | 13:57 |
bryce_ | heya sbalneav | 13:58 |
bryce_ | I've got a session starting now but think I'm free after | 13:58 |
tepsipakki | Q-FUNK: in what way is the -amd driver broken atm? | 14:02 |
tepsipakki | majority of the drivers have not changed in between xserver releases, so I wonder why those still work | 14:03 |
sbalneav | Hey bryce_ | 14:04 |
Q-FUNK | tepsipakki: auto-configuration fails on recent X versions that support it. vice-versa, on recent X, ctrl-alt-Fx fails to change console. | 14:04 |
sbalneav | yeah, changing the console hangs up the whole box. | 14:04 |
Q-FUNK | well, i get the exact same problem on an unrelated driver -siliconmotion | 14:05 |
Q-FUNK | here, it doesn't hang the whole box, but I cannot get an image back after. I can still ssh into the box, thoguh. | 14:06 |
sbalneav | Ah, well, when I say "hang", I haven't tested to see if I can still ssh in. Keyboard becomes unresposive, however. | 14:07 |
Q-FUNK | my problems with -siliconmotion are described in a bug I filed on LP | 14:07 |
Q-FUNK | the problems described by sbalneav with -amd are similar to those I experience with -siliconmotion. those problems did not exist on Feisty, which has an older X release. | 14:07 |
Q-FUNK | otherwise, almost the same source code, in both cases. | 14:08 |
Gadi | ok | 14:09 |
tepsipakki | Q-FUNK: ok | 14:09 |
Gadi | I gave the thincan to Dave Trask | 14:09 |
Q-FUNK | I've had to pin xserver-xorg-core and xserver-xorg-dev to 7.1 in both cases | 14:09 |
Gadi | bryce_: you're welcome to it for the day, but please give it back to him or another LTSP person headed to Maine | 14:09 |
tepsipakki | actually there is a problem with -sis that it hangs on logout | 14:10 |
tepsipakki | using feisty | 14:10 |
Q-FUNK | 7.2 works but you cannot change console and if the X built-in screensaver kicks-in, you can never recover video. | 14:10 |
Q-FUNK | hang on logout is also something i experience with siliconmotion, if I let it upgrade to gutsy | 14:10 |
Q-FUNK | 7.3 is even worse. | 14:11 |
Q-FUNK | something tells me that -amd might need a partial upgrade to work with the new XrandR | 14:11 |
Q-FUNK | it might also need a shuffle of the BIOS env parts, as decribed in the bug report for -amd | 14:12 |
Q-FUNK | I wouldn't be surprised if other drivers need a similar cleanup | 14:13 |
bryce_ | that seems likely | 14:23 |
bryce_ | do you have additional information on specifics of what changes are needed? | 14:23 |
bryce_ | Q-FUNK: btw, I'm going to be in and out today so don't think I should take the thincam (not really sure even how to use it) | 14:27 |
sbalneav | bryce_: if it would help, I'd be happy to lend you my thincan, and ship it to you, when I get back to winnipeg. | 14:32 |
bryce_ | I suspect I need the high level overview first; I've never heard of 'thincan' before | 14:33 |
Q-FUNK | bryce_: a thin client based on the AMD Geode chipset. | 15:13 |
Q-FUNK | works well with LTSP, except for the new bugs introduced by X >=7.2 | 15:15 |
Q-FUNK | bryce_: dtrask has the thincan left by Gadi, if you need it | 15:16 |
Q-FUNK | tepsipakki: alternately, since you live here, it could be very easy for me to loan you one, if you feel like working on it. | 15:16 |
tepsipakki | Q-FUNK: hmm :) | 15:26 |
mvo | hello! what is the current plan for xlib with xcb support? is there any timeline for this yet? we want to switch to compiz 0.7 and it will be a requirement there | 15:27 |
tepsipakki | Q-FUNK: I'm not sure if I'd be able to find the problem though | 15:28 |
tepsipakki | mvo: uploaded! | 15:28 |
mvo | rock! | 15:28 |
* mvo hugs tepsipakki | 15:28 | |
tepsipakki | libx11-xcb* is sitting in new :) | 15:28 |
bryce_ | mvo, as soon as the xserver build is through, you should be good to go | 15:28 |
* tepsipakki hugs mvo back | 15:28 | |
tepsipakki | bryce_: it doesn't really need the server, it's just a lib :) | 15:30 |
Q-FUNK | :) | 15:32 |
bryce_ | ah | 15:33 |
bryce_ | Q-FUNK: do you have a detailed write-up of the work needed for -amd? | 15:44 |
Q-FUNK | the details provided by Anti in the bug are as close to a checklist as I've got | 15:48 |
bryce_ | hmm, I don't remember the bug id offhand | 15:58 |
bryce_ | ok, going out for a bit. bbl | 15:58 |
ubotu | New bug: #159556 in xorg-server (main) "[Hardy] xserver-xorg-core incorrectly conflicting with xserver-xorg-video?" [Undecided,New] https://launchpad.net/bugs/159556 | 16:36 |
tepsipakki | hah | 16:38 |
tepsipakki | that bug ^^ | 16:38 |
tepsipakki | someone actually running hardy | 16:38 |
_bernie | hello | 19:23 |
Q-FUNK | sbalneav: I found Bernie :) | 19:24 |
_bernie | anythink I can do... | 19:24 |
_bernie | are you maintainers of amd_drv in Xorg? | 19:25 |
Q-FUNK | _bernie: Jordan said that you might be able to help us bring -amd in line with X.org 7.3 specs | 19:25 |
_bernie | we currently just rebuild it with no patches | 19:25 |
Q-FUNK | we currently have a situation where it fails autoconfiguration on recent X releases. | 19:26 |
_bernie | but we don't use Xorg as upstream | 19:26 |
Q-FUNK | you have oyur own fork? | 19:26 |
_bernie | http://dev.laptop.org/git?p=xf86-amd-devel;a=summary | 19:27 |
_bernie | unfortunately | 19:27 |
_bernie | jordan says the version on fd.o is not suitable for us | 19:27 |
Q-FUNK | https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-amd/+bug/140051 | 19:27 |
_bernie | probably some DCON changes, I never got around to diff them. | 19:27 |
ubotu | Launchpad bug 140051 in xserver-xorg-video-amd "amd driver fails to autoconfigure" [High,Triaged] | 19:27 |
Q-FUNK | yes, DCON. we don't have those | 19:28 |
_bernie | if you have commit access on the fd.o git repository, it would be nice to pull from us and resync the source | 19:28 |
_bernie | I hate forks | 19:28 |
Q-FUNK | I cannot sync as-is | 19:28 |
Q-FUNK | you guys have too many OLPC-specific changes that would need to be #ifdef'ed | 19:29 |
_bernie | I had that segfault too! | 19:29 |
_bernie | I can't remember how I fixed it though... just a moment | 19:29 |
Q-FUNK | that's why Jordan has been acting as a sort of gate keeper between both trees | 19:29 |
Q-FUNK | if/when you get around haivng proper #ifdef on all OLPC-specific stuff, we 'd love to merge trees and maintian everything at f.d.org | 19:30 |
_bernie | are you sure you're building with 100% 7.3 headers? | 19:30 |
Q-FUNK | yup | 19:30 |
Q-FUNK | but 7.2 already is problematic | 19:30 |
_bernie | I had to delete my /usr/include/X11 and /usr/include/xorg dirs at some point when I was testing stuff locally | 19:30 |
Q-FUNK | building against anything more recent than 7.1 cuases all sorts of issues | 19:30 |
_bernie | I can't test on anything but OLPC. I have no other amd boards... this is why I can't easily help making our patches available for the fd.o version | 19:31 |
_bernie | I build against 7.3 and everything is fine | 19:31 |
Q-FUNK | I think that the most pressing issue is to fix auto-configuration. as far as we can tell, this requires shuffling BIOS environmnet parsing around. | 19:32 |
_bernie | so the fix is somewhere in our tree | 19:32 |
_bernie | how about cherry-picking the easy to merge patches and retry? | 19:32 |
Q-FUNK | _bernie: please look for dtrask at UDS. he has a more generic board made with an LX on it. it's the same that LTSP guys have been playing with. | 19:32 |
_bernie | argh, I'm already back at One Cambridge Center | 19:33 |
Q-FUNK | oh | 19:33 |
_bernie | I'm looking for bits where we explicitly require DCON in our driver | 19:34 |
_bernie | I don't see them... We always check for it before any access as far as I see. | 19:34 |
_bernie | we're nearby a big feature freeze. that would be today. | 19:35 |
_bernie | so maybe next week I could start submitting patches | 19:36 |
Q-FUNK | lovely! | 19:36 |
_bernie | have you being building from a pristine git's master? Or do you have debian or ubuntu specific patches on top of it? | 19:37 |
Q-FUNK | nothing debian/ubuntu specific | 19:38 |
Q-FUNK | we build against a pristine f.d.org tree | 19:38 |
_bernie | I see there are _several_ patches on fd.o | 19:43 |
_bernie | try pulling from our git repo and see what happens | 19:43 |
_bernie | if your bug goes away, then you know where to look | 19:43 |
Q-FUNK | on fd.o ? | 19:43 |
Q-FUNK | the current 2.7.7.3 includes all the latest. | 19:43 |
Q-FUNK | ...except for OLPC additions | 19:44 |
Q-FUNK | git head on fd.o = 2.7.7.3 | 19:44 |
_bernie | I mean, try building the source from git://dev.laptop.org/xf86-amd-devel | 19:46 |
Q-FUNK | I just diffed it | 20:02 |
Q-FUNK | very few differences, oddly enough | 20:02 |
Q-FUNK | mostly DCON stuff | 20:02 |
_bernie | looking | 20:03 |
Q-FUNK | I also added that diff to the bug. | 20:07 |
Q-FUNK | and includes that have been either eliminated or shuffled around | 20:08 |
_bernie | I don't see anything that would be clearly responsible for your particular segfault | 20:09 |
_bernie | well, try building our driver and see what it does | 20:10 |
_bernie | can you do a quick test? | 20:10 |
Q-FUNK | sure | 20:12 |
Q-FUNK | midn you, you guys are using a fixed resolution with a config, aren't you? | 20:13 |
Q-FUNK | I'm also wondering why you removed amd.h from so many files | 20:13 |
ubotu | New bug: #34146 in xserver-xorg-input-hyperpen (universe) "appearance of constant full pressure" [Medium,New] https://launchpad.net/bugs/34146 | 20:32 |
_bernie | http://dev.laptop.org/git?p=xf86-amd-devel;a=commitdiff;h=1d6f6bdcb0aac722e50ac58ff94e4df5c5f0220b | 20:41 |
_bernie | dan williams says it caused a build error | 20:41 |
_bernie | I recall something messing with _X_INLINE | 20:41 |
_bernie | http://dev.laptop.org/git?p=xf86-amd-devel;a=commitdiff;h=63d72f8c1ff1b7e534fc533a6aae8683e83fc0ca | 20:42 |
_bernie | ^^^ this was also a good idea | 20:42 |
_bernie | Q-FUNK: yes, we hardcode the resolution in xorg.conf, since the LCD size is fixed. | 20:43 |
_bernie | but we shouldn't. I'd like to get rid of xorg.conf alotgether | 20:43 |
Q-FUNK | for that, you'd need to manage to fix the above ubuntu bug :) | 20:44 |
ubotu | New bug: #159617 in xserver-xorg-input-hyperpen "tablet does not work" [Undecided,New] https://launchpad.net/bugs/159617 | 20:45 |
_bernie | :-/ | 20:45 |
Q-FUNK | basically, BIOS parsing has to be cleaned up and parts of it shuffled around, for autoconfiguration to work. | 20:50 |
Q-FUNK | our Anti did most of the ground work in his coments to the bug | 20:50 |
Q-FUNK | _bernie: I get a big hard freeze with the olpc tree, as soon as X tries to load | 22:08 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!