sigp239: How do I install the radeon driver in ubuntu 9.10? I have an ATI Technologies Inc R481 [Radeon X850XT-PE]. Thanks.
MostAwesomeDude: sigp239: It's already installed.
sigp239: MostAwesomeDude, Then why does my X log show me using the VESA driver?
Hiryu: sigp239: you might take a look at xorg edgers
Hiryu: youch!
Hiryu: nm
Hiryu: apt-get install xserver-xorg-video-radeon
sigp239: xserver-xorg-video-radeon is already the newest version.
MostAwesomeDude: sigp239: Pastebin your Xorg log.
Hiryu: does your /etc/X11/xorg.conf define any video card drivers to be used?
Hiryu: sigp239: also do what MostAwesomeDude said
sigp239: Hiryu: I don't have an /etc/X11/xorg.conf
Hiryu: hmm
Hiryu: so xorg is auto deciding you have vesa
sigp239: MostAwesomeDude, Here it is: http://pastebin.com/05bkunL1
sigp239: (WW) Warning, couldn't open module ati
MostAwesomeDude: sigp239: $ sudo apt-get install xserver-xorg-video-ati
Hiryu: xserver-xorg-video-ati
Hiryu: MostAwesomeDude: heh
Hiryu: seems like xserver-xorg-video-ati should be a dependency for xserver-xorg-video-radeon
sigp239: What is the diff between xserver-xorg-video-ati and xserver-xorg-video-radeon ?
Hiryu: I ran into this on monday... I should've known better
MostAwesomeDude: The former is a wrapper.
sigp239: Okay I did that now what?
sigp239: MostAwesomeDude, ^
MostAwesomeDude: Restart X.
sigp239: Thanks
Hiryu: when you log out of gdm, does it start your latest X server? or does it use the one that was installed when gdm 1st started? kdm uses the one when you 1st login so you have to hit alt-e to reload it
Hiryu: he's taking a while, probably rebooted
mcgregor: since yesterday when I update ddx I am getting those error randomly but quite often from kernel: [drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation !
suokko: agd5f: Funny problem. If I set from bios that onboard r600 card is prefered kernel module initializes both cards but ddx is only trying to initialize discreed one
suokko: But at least I have KMS console to work with :)
airlied: suokko: DDX will only ever use one card by default
airlied: you need xorg.conf you do anything more
airlied: suokko: did you post the dmesg from offboard preferred? in theory it should disable or put bios in vram on the igp
suokko: airlied: nope. I can post it now
suokko: I still have bios prefering onboard card. But another interesting part is that early kernel messages aren't rendered to IGP either
suokko: airlied: Both logs are now at http://people.freedesktop.org/~suokko/ (dmesg_*_preferred.log)
airlied: suokko: they both look fine
airlied: but I'd expect console to end up on the onboard okay
suokko: yes. But problem is that early boot and X is selecting the 740 as primary card and I have monitor connected to onboard
suokko: hmm. In need to reboot a few times -> moving irssi to another machine
airlied: suokko: the bios picks the discrete card for messages?
airlied: even when you prefer it
airlied: prefer onboard?
airlied: suokko: I expec you are missing a config option got your .config?
suokko: airlied: It is bios bug.
suokko: I toggled "surround view" optino from enabled to auto in bios and now onboard is primary
suokko: But discrete card is enabled too when radeon module is loaded
suokko: In any case it would be improvement if X would try to autoconfigure all cards
[Enrico]: mhm i just realized that i can't switch to the tty anymore, the screen is black but the system is not freezed. if i try to switch back to X it takes long but at some point it will switch. i'm using .33 vanilla kernel and mesa, libdrm and video-ati are from git master of 6 days ago. is this a known bug already solved? if not is there something can i do?
MrCooper: [Enrico]: can you put up the X log file for us to look at?
[Enrico]: MrCooper: of course. and of course i'm using UMS, KMS works fine :)
[Enrico]: MrCooper: do you want the xorg log after a switch am i right ?
MrCooper: yeah, but VT switching is just generally much more fragile with UMS
[Enrico]: yeah i know
[Enrico]: MrCooper: but i can't use kms since the PM patch is...... well not 100% ready (laptop here). give me a minute ti finish a system update then i will switch to a vt and i will paste the log
[Enrico]: ok it is time to try the switch
[Enrico]: :/ it worked now........ so i guess it fails only from kdm..... let's try
[Enrico]: MrCooper: brb
[Enrico]: MrCooper: ok here is the log http://pastebin.ca/1851377 there are 2 switchs, the first from kde worked, but the second one from kdm returned to me a black screen (then i switched again to kdm and logged)
[Enrico]: it is intresting that until now, only switchs from kdm gave me a black screen
[Enrico]: no wait i just retried from kde -> black screen
[Enrico]: MrCooper: oh when i use radeon in UMS i use vesa fb for the tty
Pallokala: hi, http://pastebin.com/P2g4GZNP
Pallokala: is it normal that my "Default" and "Performance" states have the same engine/memory-clock?
Pallokala: card being rv630 agp
Pallokala: passively cooled
rah: waits for "tv standard" property to be fixed :/
twnqx: waits for r635 to do 2D without locking up immediately
soreau: waits for the static to get fixed on right half of screen when 3D happens
adamk: waits for KMS to arrive on FreeBSD :-)
glisse: twnqx: you are unlucky all rv635 we have access too works fine
twnqx: :(
twnqx: waits for the fedora live cd
twnqx: maybe it's really something else, like kernelconfig or something in userspace
[Enrico]: waits for a usable PM algorithm and finally he will switch to kms for everyday use
kdekorte: twnqx, what issue are you having with an rv635?
dileX: radeontool 1.6.1 released
dileX: http://airlied.livejournal.com/72345.html
suokko: kdekorte: I think I can't reproduce eog problems. Only slow stuff is if window under eog is slow to repaint
twnqx: kdekorte: gpu locks.
twnqx: that do not even trigger glisse's lock detection, but are soft (e.g. reboot from an ACPI hotkey works)
twnqx: kdekorte: and 100% guaranteed when i try to start the like of filezilla or thunderbird, not xchat or firefox, though
twnqx: and only with KMS
suokko: twnqx: hmm. btw, Are you sure it is GPU hang and not X lockup?
twnqx: well
twnqx: fade to white is a pretty strong indicator
suokko: Because I got X hang when overflowing the command buffer in ddx
suokko: Then kernel jsut rejected the commands for me
suokko: But for me restarting X from ssh fixed the problem
twnqx: shouldn't it log something in dmesg then?
suokko: yes. It should be in dmesg
twnqx: nothing there
suokko: But just in case it is X hanging effect could be similar to GPU hang
twnqx: gonna try today evening, when i have a second comp to ssh from...
twnqx: but should that even trigger fade to white?
suokko: fade to white sounds that something is messing with framebuffer so probably not
twnqx: no, not that fade... the one that makes you fear for your hardware.
MrCooper: blooming?
kdekorte: twnqx, I run thunderbord all the time under KMS
twnqx: yeah, i feel SO alone with my problem :(
twnqx: hence i'm gonna try something more mainstream
Jonimus: I saw my first tearing since starting to use this driver today. :)
Jonimus: 5 internets for whoever guesses what caused it.
twnqx: claims pebkac
Jonimus: twnqx: nope Flash
twnqx: heh
Jonimus: it wasn't even full screen
suokko: tough that worst problems with flash were fixed
twnqx: hm
suokko: exluding that flash is not supporting linux well ;)
twnqx: fedora 13 alpha cd iso has 840MB for x86_64
twnqx: that doesn't quite fit on a CD :X
Jonimus: twnqx: yeah thats total fail
suokko: twnqx: 1G memory stick is standard install media ;)
twnqx: i don't want to install a junk distro
twnqx: i just want to boot it to test KMS
suokko: ok. 1G memory stick is standard live cd media
twnqx: and i'm too clueless to make an ISO boot from usb memory
mjg59: twnqx: dd
twnqx: hm
mjg59: The Fedora live images all use isohybrid, so you can just dd them directly onto a USB stick
twnqx: looks what on this 4GB stick
suokko: twnqx: https://fedorahosted.org/liveusb-creator/ easy tool
twnqx: and 13alpha is supporting KMS on rv635, right?
twnqx: i wonder why the KDE version is smaller than the "normal" version...
twnqx: yey, this usb stick only has redhat ES 4 update 3 to install from
twnqx: suokko: is there a non-qt version of that tool?
suokko: twnqx: There is gtk tool too but I don't remember where to find it.
twnqx: kk
Obscene_CNN: ponders the idea of using several usb sticks to form a raid
Jonimus: lol Obscene_CNN USB would be the bottleneck not the actual flash chip
Obscene_CNN: writes take a while on the flash sticks though
agd5f: kdekorte: does the patch on http://bugs.freedesktop.org/show_bug.cgi?id=27284 fix the slideshow probably you are seeing?
kdekorte: agd5f, I'll try it... should I apply to git master HEAD?
agd5f: yes
kdekorte: agd5f, nope it does not fix it
agd5f: ok
suokko: btw. When I see slow performance in window move pixman trapezoid rendering is the slowest part but with only 15% share of cpu
suokko: That is when I move window over firefox
kdekorte: agd5f, I'm tracking that bug now...
glisse: agd5f: you tested new i2c code with r100/r200 ?
agd5f: glisse: yes. works on my cards
agd5f: glisse: but doesn't work on some others
glisse: guess i need to look why it fails here
agd5f: there's a bug about it
agd5f: the prescale need to be tweaked
agd5f: glisse: http://bugs.freedesktop.org/show_bug.cgi?id=26430
evil_core: agd5f: you know that pm2 is broken at T60p with r500?
agd5f: evil_core: what's broken?
evil_core: sets core to low speed, and all is fuzzy
evil_core: unbtil you run glxgears or something
agd5f: evil_core: fuzzy? is the screen is out of focus or flickering or something else?
evil_core: like Zajec firsts patch, but your sometimes when browsing www not only makes fuzzy, but I got some white frames and flassesh
evil_core: out of focus a bit, and fuzzy
evil_core: its not stable(at constant core speed)
evil_core: even on console
agd5f: evil_core: try commenting out the memory reclock
agd5f: in r100.c
evil_core: dont know if you released fixed version or something
evil_core: I got it about week ago
evil_core: and applied all ptches from pm2 dir
evil_core: somebody told me your code is using values from powerplay table, is it true?
agd5f: evil_core: yes
agd5f: evil_core: new versions posted on monday
agd5f: in my pm2 directory
evil_core: not sure, but remember that I were able to go only 500 or 15000MHz down without fuzzy screen, but it were hardcoded all
evil_core: is your version flicker-free?
evil_core: agd5f: Are there any other patches that would be usable for r500 against d-r-t from your site?
agd5f: evil_core: my pm2 patches are against drt
agd5f: although some of the earlier ones are already applied
evil_core: names changed, but hopefully theres bash history :D
evil_core: patch=10; for i in pm2/*; do echo -e "Patch$patch:\t$i"; ((patch++)); done
evil_core: (I am adding them to SPEC file)
thansen: should displayport work out of the box with 2.6.32 (on ubuntu 10.04)
agd5f: thansen: yes
thansen: :(
agd5f: thansen: what card?
thansen: agd5f: firepro 7740 mobile
DanaG: arghjkl, I keep getting this:
DanaG: [ 798.790220] [drm:radeon_fence_wait] *ERROR* fence(ffff8800b58fae80:0x0001B6D2) 510ms timeout going to reset GPU
DanaG:
DanaG: ANd kernel is dead... heartbeat LED is not "beating".
agd5f: thansen: using kms?
thansen: agd5f: yes
thansen: gnome-display-properties is detecting the screen with the correct res etc
thansen: I just don't get a picture and monitor says it isn't receiving a signal
agd5f: thansen: can you pastebin your xorg log and dmesg?
DanaG: firepro 7740... what actual gpu is that?
DanaG: er, M7740
agd5f: rv740
thansen: DanaG: don't remember, I pulled it up on wikipedia once..just got the laptop today
agd5f: thansen: what res monitor?
thansen: 2560x1440
thansen: new dell 27"
DanaG: lspci
DanaG: ah.
agd5f: thansen: native dp or dp to dvi adapter?
thansen: native dp
thansen: I don't have adapter to test with either :(
thansen: http://www.pastebin.ca/1851676
thansen: xorg.log
thansen: http://www.pastebin.ca/1851675
thansen: lspci
thansen: http://www.pastebin.ca/1851673
thansen: dmesg
DanaG: hmm, what would a FirePro M7820 be?
DanaG: http://hpfansite.com/hp-elitebook/ati-firepro-m5800-appears-elitebook-8540w-dreamcolor
thansen: DanaG: new thing based on evergreen in hp mobile workstation
DanaG: yeah, but which actual GPU?
DanaG: 5400? 5600? 5800? =þ
DanaG: hmm, did 2.6.33 have dynpm in it yet?
DanaG: er, gotta' go.
edwin: Obscene_CNN: good news for you. Your latest patch to xf86-video-ati+libdrm works! (I am in KDE4 right now with that patch applied)
edwin: Obscene_CNN: going to retest your MEsa patch now, lets see if it still hangs the GPU
edwin: Obscene_CNN: there is one caveat with your patch: if I don't upgrade libdrm first, then X will fail to load the radeon driver due to a missing symbol. Shouldn't that be caught at build time? (that radeon driver has a missing symbol not satisfied by any of the libraries it links to)
svenstaro: edwin: it probably can't be caught at build time if you don't build statically
Dice-Man: hi
Dice-Man: themgdn: hey man
Dice-Man: who is the channel's owner ?
edwin: Obscene_CNN: mesa patch looking good too!
shadowmaster: Dice-Man: /msg chanserv info #radeon
mcencora: agd5f: could you push the patch from #27043?
zhasha: airlied: okay so it's not only S-Video that's completely fooked, but VGA outputs garbage... It's getting really weird, and the output is VERY inconsistent
agd5f: mcencora: sure
agd5f: mcencora: pushed
airlied: suokko: X can't autoconfig cards, it has no idea what to do
suokko: So there is default config with one card+one screen
airlied: suokko: the main thing is xinerama vs multi-desktop stuff, you really don't want X to decide you wanted xinerama
suokko: But maybe autodetect could use first device that finds monitor attached
airlied: suokko: it should use the same device as the BIOs uses
airlied: and does for me
suokko: airlied: It does but now use case was if I have no monitor in primary device (for me some BIOS option was toggle to wrong) but in decundary. X doesn't know to configure the secundary device automaticaly
airlied: suokko: it assumes you can set things up to see the bios messages
airlied: and copies thta
airlied: since the console will also be on that display
airlied: though you need to make sure you have the CONFIG_FRAMEBUFFER_CONSOLE_DETECT_PRIMARY
suokko: I jsut want to be lazy and not to greate conf ;)
suokko: When I want to switch to secundary without entering bios
suokko: /greate/create/
suokko: But now I have minimal xorg.conf that works
mjg59: agd5f: Is there any way to request a vline interrupt without sending something to the CP?
agd5f: mjg59: just set the vline reg and enable vline interrupts
mjg59: agd5f: And it just takes a line number? That's shockingly easy.
agd5f: mjg59: it takes a range
agd5f: and you get an interrupt inside or outside the range
mjg59: RADEON_CRTC_GUI_TRIG_VLINE ?
mjg59: Or AVIVO_D1MODE_VLINE_START_END ?
mjg59: Wait, that's a command, not a reg
agd5f: mjg59: yup
agd5f: avivo variant is for r5xx+
agd5f: radeon variant is for r1xx-r4xx
mjg59: agd5f: Heh. Not in the docs I have.
mjg59: Let me look more...
agd5f: mjg59: see the vline stuff in the ddx
agd5f: RADEONWaitForVLine
mjg59: Got it
twnqx: so yeah
twnqx: fedora 13alpha kernel doesn't really work either, tells me something about recursive locks and blurb
airlied: twnqx: does it lockup the same?
twnqx: no
twnqx: but i couldn't test, couldn't log in
twnqx: X didn't even fully crash, i could use the reboot...
twnqx: gonna put the lock debugging in my kernel here
kdekorte: twnqx, have you checked to see if there are an BIOS updates for your motherboard?
twnqx: back when it was IBM you'd get those... with lenovo... well
twnqx: i'll check
twnqx: well. filling a cd-r with 21MB :P
cemc: compiled for 1.7.5, module version = 6.12.191 - on ubuntu lucid. can I somehow change the powersave settings, like with CCC PowerPlay ?
Dice-Man: http://www.youtube.com/watch?v=jlzI3_QADWc
Dice-Man: cool song
mjg59: cemc: Not currently
cemc: mjg59: any ideas when?
mjg59: 2.6.35, with luck
cemc: mjg59: that's why this message appears? (WW) RADEON(0): Option "ForceLowPowerMode" is not used
Obscene_CNN: sync
Obscene_CNN: oops, wropng window again
twnqx: so the bios upgrade doesn't help
twnqx: BUT
twnqx: the X process crashes.
twnqx: and that the screen fades to white.
twnqx: so I can just restart X
twnqx: *** BUG ***
twnqx: pixman_region_union: Malformed region new_reg
twnqx: Set a breakpoint on 'log_region_error' to debug
twnqx: hmmmm
twnqx: wonder if that's from enlightenment
suokko: twnqx: Have you run memchecker?
twnqx: come on, i run gentoo
twnqx: with broken mem i'd have more errors than this single, 100% reproducable one :P
suokko: memtest that is :)
twnqx: i know what you mean
twnqx: guess i'll create a debug build of the X server
suokko: ok so it is some pixman bug
twnqx: and run it in gdb
twnqx: ... what is pixman?
twnqx: :P
suokko: software rasterizer
twnqx: mh
twnqx: upgrade to 0.17.4? or even git?
twnqx: but hey, the KMS driver can't recover from sudden client loss?
suokko: I don't remember if I have crashed X ... But I haven't noticed that KMS wouldn't recover unless there is kernel module bug
twnqx: tests pixman git
twnqx: right... installed...
twnqx: YEY
twnqx: i can start filezilla!
twnqx: \o/
twnqx: so the question would be, why does pixman crash with KMS but not with UMS
twnqx: but it doesn't really matter since git doesn't
twnqx: so now i can go for powersaving, yes?
suokko: twnqx: If it works for you
twnqx: dunno yet :P
agd5f: cemc: tha option is for UMS only
twnqx: turns on dynpm and dynclks
twnqx: brb
cemc: agd5f: UMS ?
agd5f: cemc: usermode modesetting
agd5f: ie, not kms (kernel modesetting)
edwin: yeah I found that confusing too
edwin: before KMS noone talked about UMS
agd5f: cemc: it's also only in git master and the 6.13 rc releases. 6.12.x doesn't support that option
edwin: agd5f: what will be done about the REPEAT_NONE FIXME for 6.13? fix firefox or implement the fixme?
cemc: agd5f: how will this option be different from powerplay ?
cemc: confused ;)
agd5f: edwin: ideally fix firefox. I'm not sure there's a way to use a border color alpha if the surface doesn't have an alpha channel
agd5f: edwin: the spec doesn't match what hw can do
agd5f: cemc: that option is just a stop gap while we develop better power management in kms
cemc: agd5f: will it reduce temps? :)
agd5f: cemc: yes
twnqx: agd5f: does the KMS driver notice power state changes?
cemc: good enough for me :) is it difficult to get it from git and try it ?
agd5f: twnqx: the kms driver implements power state changes
twnqx: no i mean
twnqx: if the tables change
agd5f: twnqx: the tables are in the vbios of your card
agd5f: so they vary from card to card
twnqx: i'm SURE i saw more than the one i see now in earlier KMS boots
suokko: agd5f: What all pixmaps have hiden alpha for acceleration?
agd5f: twnqx: try my new patches. we skip some modes in the current code
agd5f: suokko: ?
twnqx: well, it's the same kernel
twnqx: rebooting :P
suokko: agd5f: just to get alpha for border operations
agd5f: suokko: we force the alpha channel to 1 for XRGB, etc. surfaces
agd5f: so the border color is forced to 1 too
agd5f: but the spec requires it be 0 IIRC
agd5f: suokko: should 0 for areas outside the src that is
suokko: agd5f: But then spec requires alpha 0 for solid color inside, right?
twnqx: agd5f: http://nopaste.info/d3052509a8.html
agd5f: twnqx: yup. 6 power states. looks normal
twnqx: look at the last
twnqx: end*
twnqx: last boot.
edwin: agd5f: does firefox actually use the alpha for that repeat_none though? it doesn't seem to make a difference visually
twnqx: agd5f: -dirty is just the removal of the lvds quirks
twnqx: that's the only difference...
twnqx: and now it only finds one power state...
twnqx: ok, and the inclusion of the irq firmware into the kernel.
agd5f: edwin: it depends on the transform
agd5f: twnqx: ? the pastebin shows 6 power states
twnqx: the pastebin shows to boots
twnqx: one from march 16th, one from martch 24th
twnqx: two*
twnqx: and march* - the one from today only finds one power state
agd5f: twnqx: bisect? and see what changed it?
twnqx: nah, look at the kernel versions...
twnqx: the diff is, as i said, only the lvds quirks removal
twnqx: i guess the new bios suokko made me flash... removed the power states
agd5f: twnqx: does booting the old kernel bring them back?
twnqx: dunno, that kernel killed the LVDS :P
twnqx: but gonna try
twnqx: brb
agd5f: if that patch is the only difference, I'd say the bios changed something
twnqx: nope, they are gone
agd5f: twnqx: new bios may have changed the version of the power table
agd5f: twnqx: can you send me a copy of you vbios?
twnqx: the new one?
twnqx: sure
agd5f: yup
Dice-Man: short
twnqx: agd5f: how'd you like it?
twnqx: attached to a bug?
twnqx: dcc, email?
agd5f: email
agd5f: alexdeucher AT gmail
twnqx: "you have mail"
cemc: agd5f: I'm trying to get 6.13 from git, but I don't see how... I've got something, I've compiled it and it was 6.12.192 and still no ForceLowPowerMode
agd5f: cemc: 6.12.192 is 6.13rc release
agd5f: cemc: pastebin your xorg log
agd5f: twnqx: yup the table version increased
twnqx: yey.
agd5f: looks the same though content wise
cemc: agd5f: http://pastebin.ubuntu.com/400817/
twnqx: are your patches in drm-radeon-next?
agd5f: twnqx: to fix it, change this line in radeon_atombios.c::radeon_atombios_get_power_modes(): } else if (frev == 4) {
agd5f: change it to } else if (frev >= 4) {
cemc: agd5f: I'm not saying I did everything right, but I tried to follow that howto on the x.org/wiki
agd5f: cemc: you are using kms
agd5f: that option is for kms only
agd5f: ums only
agd5f: cemc: you need to load the radeon drm module with modeset=0 to use ums
twnqx: :\
twnqx: 2.6.33-50669-g65965f4-dirty
twnqx: not dirty enough
agd5f: twnqx: my patches are here http://people.freedesktop.org/~agd5f/pm2/
agd5f: the first 4 are already in d-r-t
twnqx: and now
twnqx: your server just went down >_>
twnqx: i could get the directory index, seconds later the httpd was down :P
twnqx: i don't see a patch to enable newer versions of the power tables, though
twnqx: or is it 11?
cemc: agd5f: got it working now, and the option seems to work too, got the temp down from 50 to 45. still a bit high but it's a start ;)
Dice-Man: sometimes radeon chipsets bugs a lot
soreau: I have installed arch and was having the constant problem of, sometimes when I boot and X starts, compiz is extremely slow with all kinds of out of sync issues like it's running everything with swrast. This was happening with mesa 7.7 but now the same is happening with 7.9 latest. Sometimes if I reboot it will be fine but usually not
Dice-Man: yesterday i was teaching pupils how to use openoffice and the retroprojector fail
soreau: but all logs look good including x log, dmesg and glxinfo
soreau: nothing that I can see indicating a fail
soreau: on rv350, x86, Arch, X 7.6, EXA, libdrm, mesa, xf86-video-ati - git
amarsh04: will try drawing an ellipse in GNU paint again, if I drop out, it's because it killed the x server again
twnqx: maybe he should update his pixman, too
twnqx: :]
suokko: Dice-Man: What happened?
Dice-Man: a big fatal error during the lesson
Dice-Man: and all of my pupils was laughing
suokko: Dice-Man: technicaly speaking :)
Dice-Man: i dunno
Dice-Man: i change the graph card
Dice-Man: i take an intel one
suokko: Dice-Man: Did image just disapear or rendered somehow incorrectly?
Dice-Man: disapear
Dice-Man: i'm run under mdv 2010
Dice-Man: boa
Dice-Man: don't worry about that
Dice-Man: i juste make a bad experience with radeon chipset
suokko: Dice-Man: mdv is mandriva?
Dice-Man: yep
suokko: Dice-Man: But we want to fix the problems
Dice-Man: humm well
Dice-Man: i will check out all the references tomorrow
Dice-Man: and check out the logs
amarsh04: is back from drawing an ellipse in GNU Paint, which killed the x server (radeon 9200se)
agd5f: twnqx: you still need to change the line I posted here
twnqx: i see
twnqx: must have missed that at start
suokko: amarsh04: Do you have backtrace in xorg.log
suokko: ?
soreau: hmm.. seems compiz starting too soon was the culprit
amarsh04: suokko, will check my xorg.log
twnqx: ok, the tables are back :P
soreau: Can anyone say why gl looks and feels like LIBGL_ALWAYS_SOFTWARE is set most of the times when I start X but other times it works fine?
soreau: the intermittent nature of this problem and the fact that no logs say anything out of the ordinary afaict is baffling
soreau: on rv350, x86, Arch, X 1.7.6, EXA, libdrm, mesa, xf86-video-ati - git, kernel 2.6.33-ARCH
chithead: if it occurs only with certain apps, try disabling low impact fallback in driconf
Obscene_CNN: well I have glxgears spending 2/3s of its time in the kernel now :)
amarsh04: wonders how best to report the "xserver crash when drawing ellipses in GNU paint with radeon driver"
airlied: amarks_: get a newer X server
airlied: oops
airlied: amarsh04: get a newer X server ;-
amarsh04: I can recheck what debian has in experimental
jcristau: no xserver there.
_ds_: Trying out gallium on an X300… seems that some people have shaders working on lesser hardware, and here I am with glxinfo reporting OpenGL 1.5.
_ds_: Still, everything else seems to be working, if a little slowly.
_ds_: (Setup: X from Debian testing, Radeon driver from experimental, libdrm 2.4.19, mesa git from a few hours ago)
BioTube: _ds_: I think r300c gained support for shaders
_ds_: Hmm.
_ds_: Curiously, Oolite is allowing me to alter its shaders option.
_ds_: Nice. "Assertion `0' failed."
MostAwesomeDude: _ds_: File and line number?
_ds_: r300_fs.c:209
MostAwesomeDude: Hm. Alright, I'll look at it later.
MostAwesomeDude: (You can open a bug, if you like.)
_ds_: sees component "Drivers/Gallium/i915g" but no "r300g"…
_ds_: http://tartarus.org/ds/shady-thargoid.png – not quite how it's supposed to look…
_ds_: (oolite, "freaky thargoids" OXP, shader setting "simple")
BioTube: _ds_: there's a reason r300g isn't built by default
_ds_: Yes, I'd gathered that :-)