Sarvatt | sure, pass it along RAOF :) | 01:40 |
---|---|---|
RAOF | Sarvatt: It's in xorg-edgers. Seems to work for me, but my nv40 cards have ceased to be an interesting test :) | 02:03 |
Sarvatt | would be nice to know why we have to remove quiet from the boot line on nouveau :D | 02:07 |
RAOF | Note: that definition of "we" does not include me :) | 02:07 |
RAOF | That particular strangeness got fixed somewhere in the last weeks for me; probably it was mad framebuffer related craziness. | 02:08 |
Sarvatt | oh? must be nice! | 02:08 |
Sarvatt | maybe if i blacklist vga16fb, still havent done that on that machine | 02:08 |
RAOF | I think the -9 kernel fixed stuff, too. Maybe. | 02:09 |
Sarvatt | removed my firmware files, upgrading now | 02:09 |
Sarvatt | got a ton of dmesg spam for 5 seconds about 20 seconds in but seems to be working ok, firmwares there and got loaded | 02:22 |
RAOF | Superlative. | 02:27 |
* Ng sulks. I was copying a bunch of data to a USB disk and it took out X | 12:11 | |
mac_v | Ng: karmic? | 12:36 |
hyperair | took out meaning X lagged like hell or something else? | 12:56 |
hyperair | imo we've got some big issues with IO and interactivity | 12:57 |
* mac_v had somewhat similar X crashes when system was doing heavy read/write operations... but using lucid kernel 32.* solved it .. | 13:11 | |
mac_v | x restarts* | 13:11 |
Ng | mac_v: lucid. looks like the kernel oopsed becuse of something relating to the USB disk and the xlog suggests it then segfaulted, but I could still see it, which was a bit odd. ssh'd in and rebooted | 13:35 |
hyperair | hmm i wouldn't know, i didn't use .31 for long. had a great number of ext4 issues | 13:46 |
Ng | (I couldn't interact with it in any way though, so I can well believe that X had died and the framebuffer just stayed as it was) | 13:50 |
Ng | quite annoying though, I bought the USB disk so I'd be able to take regular rdiff-backup snapshots of my laptop while I'm running lucid on it ;) | 13:51 |
tjaalton | tseliot1: I noticed you updated the halsectomy-spec. there's no working wacom driver atm, so you sed evdev which doesn't work properly with it | 16:01 |
tseliot1 | tjaalton: yes, I also tried to contact pitti and you but I was disconnected from IRC | 16:02 |
tseliot1 | tjaalton: aah, it's evdev then | 16:02 |
tjaalton | there will be a new xf86-input-wacom RSN, and I hope ron finds it good enough to package for experimental | 16:06 |
tseliot1 | tjaalton: well, I think it would be better than using evdev anyway | 16:07 |
tjaalton | I tried to find an intuos4 M/L, but they were out of stock.. | 16:08 |
tjaalton | tseliot1: it sure would | 16:08 |
tseliot1 | tjaalton: (just FYI) this is the only change that I would need to apply to X in order to get the alternatives system to work: http://pastebin.ubuntu.com/345406/ | 16:09 |
tseliot1 | I don't even have to touch mesa | 16:10 |
=== mac_v is now known as _vish | ||
tjaalton | tseliot1: do libdri/libglx need to be moved to another path for it to work? | 16:22 |
tseliot1 | tjaalton: they live in /usr/lib/xorg/modules/extensions/standard/ and, when you use open drivers, you will have links to them in /usr/lib/xorg/modules/extensions/ | 16:23 |
tseliot1 | in the case of nvidia you use libdri.so from /usr/lib/xorg/modules/extensions/standard/ and libglx.so from /usr/lib/nvidia-current/xorg/libglx.so | 16:24 |
tseliot1 | then a simple update-alternatives --set gl_conf $master link and you can switch between them | 16:25 |
tseliot1 | where $master_link is /etc/standard-x11/standard.conf in the case of X11 and open drivers | 16:26 |
tseliot1 | therefore, in order to switch to open drivers (the default): | 16:26 |
tseliot1 | update-alternatives --set gl_conf /etc/standard-x11/standard.conf | 16:26 |
tseliot1 | while, for example, for nvidia-current: | 16:26 |
tseliot1 | update-alternatives --set gl_conf /etc/nvidia-current/ld.so.conf | 16:27 |
tseliot1 | of course jockey will deal with this | 16:27 |
tseliot1 | oh and of course "ldconfig" is required after these commands | 16:28 |
tjaalton | ok | 16:29 |
tjaalton | why not just installlibdri/glx directly to the correct path instead of moving them around? | 16:30 |
tjaalton | and if the confs aren't meant to be admin-configurable, shouldn't them be under /usr somewhere? | 16:31 |
tjaalton | *they | 16:31 |
tseliot1 | tjaalton: update-alternatives doesn't create links if the file already exists | 16:33 |
tseliot1 | tjaalton: what do you mean by admin-configurable? You can do it manually if you want | 16:34 |
tseliot1 | i.e. install the packages, update alternatives, set the driver in xorg.conf and reboot | 16:34 |
tjaalton | I mean the file content probably will never change | 16:35 |
tseliot1 | tjaalton: what file? | 16:35 |
tjaalton | and cleaning conffiles from /etc is more troublesome | 16:35 |
jcristau | /etc/nvidia-current/ld.so.conf < that file | 16:36 |
tseliot1 | aah | 16:36 |
tseliot1 | those files are not a problem unless something links to them | 16:36 |
tseliot1 | and when you uninstall the packages the alternative is removed and so are the links | 16:37 |
jcristau | but the file isn't | 16:37 |
tseliot1 | why not? | 16:38 |
jcristau | because conffiles aren't removed on package removal | 16:38 |
jcristau | they're removed on purge, except when they become obsolete and you forget about them | 16:38 |
tjaalton | exactly :) | 16:38 |
tseliot1 | aah, now I see your point | 16:39 |
jcristau | so for something that's not supposed to be configurable, installing it in /etc leads to pain down the road | 16:39 |
jcristau | or cruft, or whatever | 16:39 |
tseliot1 | I can either make the postrm deal with that or simply install them in /usr/... | 16:40 |
tseliot1 | I remove the alternative in the postrm, therefore I think it's better if I remove the file there | 16:41 |
tjaalton | better in /usr, otherwise it would also "pollute" the namespace in /etc | 16:42 |
jcristau | you should remove the alternative in prerm | 16:42 |
tseliot1 | that would be a solution too | 16:43 |
tseliot1 | good idea | 16:43 |
tseliot1 | shall I use /usr/share/nvidia-current etc, or /usr/lib/nvidia-current (that I already use). The former makes more sense to me | 16:46 |
jcristau | lib imo | 16:46 |
tseliot1 | right, after all is an ld.so.conf file, which is relevant to that directory | 16:47 |
tseliot1 | also, is there a point in adding a prerm.in instead of a prerm in X? | 16:52 |
tseliot1 | it's not a template | 16:52 |
tseliot1 | oh, there's the #INCLUDE_SHELL_LIB# thing | 16:55 |
jcristau | prerm gets removed by xsfclean | 16:55 |
tseliot1 | ok, I'll use a .in file | 16:55 |
tseliot1 | that's why I asked | 16:55 |
tseliot1 | jcristau: are templates (i.e. .in files) automatically detected? | 17:01 |
tseliot1 | or shall I add the prerm.in file somewhere? | 17:03 |
jcristau | see the genscripts rule in xsfbs.mk | 17:07 |
=== _vish is now known as \vish | ||
tseliot1 | ah, nice, thanks | 17:13 |
tseliot1 | jcristau, tjaalton: better? http://pastebin.ubuntu.com/345439/ | 17:27 |
jcristau | not sure what the point of /usr/lib/standard-x11/standard.conf | 17:28 |
jcristau | is | 17:28 |
jcristau | oh. default alternative. /dev/null would work as well :) | 17:29 |
tseliot1 | it's an empty file since we're going to use the libraries that come with X/mesa | 17:30 |
tseliot1 | e.g. libGL.*.so, etc. | 17:30 |
tseliot1 | while fglrx and nvidia use their own libraries | 17:31 |
tseliot1 | and yes, that's for alternatives, as you said | 17:31 |
superm1 | tseliot1, rather than relying on jockey to update the alternatives, why not do it right in the postinst? | 18:02 |
superm1 | the less custom work that jockey has to do, the better (imo) | 18:02 |
jcristau | i guess because alternatives have static priorities, which can't take into account the actual hardware on the machine, so can't decide between nvidia-current or nvidia-last-year or fglrx-current or fglrx-legacy, or or.. | 18:04 |
superm1 | well if you run update alternatives and they do have static priorities, then that will be a no-op | 18:10 |
superm1 | eg current should always be newer than last year which should always be newer than last-last-year | 18:10 |
superm1 | but if you run jockey it can override such priorities? | 18:10 |
jcristau | what will be a no-op? if you have multiple nvidia driver packages for different generations of chips which one gets higher priority? | 18:12 |
tjaalton | or fglrx over nvidia? | 18:12 |
tjaalton | that's why I didn't think alternatives would work, but making the user choose via jockey is the only way | 18:14 |
jcristau | tseliot1: maintainer scripts have update-alternatives in $PATH no need to use the full path | 18:18 |
jcristau | also no need to use $(CURDIR) when you don't need an absolute path | 18:22 |
jcristau | and echo \n is not portable :) | 18:22 |
superm1 | if they all need to have the same priority set, then maybe a debconf to select among the priorities | 18:26 |
superm1 | *debconf question | 18:26 |
jcristau | tjaalton: not sure if you saw, the testsuite failure on sparc/ppc should be fixed in master now | 18:33 |
tjaalton | jcristau: ah, good. didn't see that | 18:41 |
tjaalton | superm1: that just moves the question to a different layer | 18:42 |
superm1 | tjaalton, but then you can interactively ask | 18:42 |
superm1 | if there is only one option available, then you wouldnt have to, but if there are a few, then you can present them to the user | 18:43 |
tjaalton | superm1: if you only have one blob installed manually, it'll have a higher priority than the default | 18:43 |
tjaalton | so no need to ask anything there | 18:44 |
superm1 | right, and then if you have two blobs, then you ask the question | 18:44 |
tjaalton | well there should be a way to not ask it if the choice is already made by jockey | 18:45 |
tjaalton | or rather, by using jockey | 18:46 |
superm1 | jockey uses debconf noninteractively currently | 18:46 |
superm1 | so it would have to preset that value | 18:46 |
superm1 | this can just be for a second iteration of changes though, because it will require work on the jockey end that may be non-trivial | 18:47 |
tjaalton | sure | 18:52 |
* bryce_ waves | 19:14 | |
tseliot1 | superm1: priorities work when in automatic mode | 19:36 |
tseliot1 | in jockey I would switch to manual mode by using update-alternatives --set | 19:36 |
tseliot1 | do we really want to use debconf for this? | 19:37 |
tseliot1 | hey bryce_ | 19:37 |
tseliot1 | jcristau: ah, thanks, I'll just use update-alternatives without the full path then | 19:42 |
bryce_ | heya tseliot1 | 19:45 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!