Nightwulf|work: hi all
obv1us: hi there
obv1us: aha.. there are people here.. anyone knowledgeable ?
Nightwulf|work: depends on what you're calling "knowledgeable" and what kind of questions you have :)
edman007: obv1us, do you like shiny things too?
obv1us: ok :) I've been trying to compile xf86-video-radeonhd and got some errors. I don't think I've compiled much in linux before...
obv1us: edman, like shiny new gfx cards ?
edman007: obv1us, pastebin the error(s)
obv1us: Let me first give you the context.. I probably have a bit of a strange setup.
obv1us: What's happened is my old 32bit motherboard became unreliable, so I replaced it.. and that meant new cpu, ram ... and gfx card.
obv1us: I'm on Kubuntu 7.04 feisty.
obv1us: so I replaced everything with the new hardware, and booted up. No X.
obv1us: (so I'm in the console now.)
obv1us: I'm adverse to upgrading my system at the moment, as it's too much software change, so I want to try to get the video card working on Feisty.
obv1us: Anyway, so the OS is 32bit running on 64bit hardware.
edman007: ...and the errors?
obv1us: The compile errors I get involve 'uint8_t' and 'uint32_t'
Nightwulf|work: if it's PC based, there is nothing like pure 64bit hardware btw
obv1us: (ok, that makes me feel a bit better.)
edman007: obv1us, pastebin everything...
obv1us: ok... pastebin pastebin... um.. checking help
edman007: obv1us, everything you see (a few screen fulls or so) should get pasted onto pastebin.ca, then give us the link
obv1us: how can I do that? I'm stuck in the console..
edman007: obv1us, you don't have another box with a gui that you can use to ssh and copy/paste the error?
obv1us: I have a laptop sitting here looking at pastebin.ca. ssh I can try...
obv1us: just getting the make output into a file... including errors...
obv1us: ok, got stderr output into the file, now to get it to the laptop...
obv1us: while I'm installing openssh-server, here's apparently the first error:
obv1us: In file included from rhd_driver.c:123:
obv1us: radeon_accel.h:36: error expected specifier-qualifier-list radeon_accel.h:36: error: expected specifier-qualifier-list
obv1us: before 'uint32_t'
edman007: are there any lines above that saying that it could not find some file?
obv1us: (got the file across using scp)
airlied: do you have /usr/include/stdint.h
airlied: sounds like missing -dev packages
obv1us: yes I do
obv1us: probably I have not installed all dev environment properly
obv1us: it makes me think that stdint.h isn't in the path or something.
obv1us: what should I check?
obv1us: The way I set up the dev environment:
obv1us: sudo apt-get install git-core
obv1us: git-clone git://anongit.freedesktop.org/git/xorg/driver/xf86-video-radeonhd
Nightwulf|work: one question concerning the git version....all the changes happening last days in radeon_xaa.*, radeon_textured_video.* etc....are they still all r5xx only or should they work on 6xx too?
libvde: hrm... why were those uint* things necessary in the first place
libvde: X has used CARD* for 20 years or so
obv1us: sudo apt-get install autoconf xorg-dev automake xorg-build-macros libtool
obv1us: then I did ./autogen.sh which completed successfully, then make... which led to the errors above
obv1us: Have I installed the wrong packages perhaps, or need to install any more ?
airlied: libvde: CARD32 is broken for printf without warnings..
airlied: libvde: it is type unsigned long on 32-bit and unsigned int on 64-bit or some such stupidity.
airlied: nobody wanted to change it in case some corner case broke.
obv1us: Nightwulf, the reason I went for the git version was because I saw in the changelogs that RV6xx support was being added, which is what I wanted it for. Not sure about those particular files you mention.
ssam: I have a 3650. i have just tried the radeonhd driver on the latest ubuntu intrepid live cd. it is version 1.2.1, but X locks at a black screen. is it even worth bothering with this old version?
obv1us: I don't think so.
obv1us: I think the 3650 uses RV630 chipset.
ssam: if i were to install a distro on a spare partition to test radeonhd, what would ubuntu or debian be a good choice?
ssam: RV635 i think
ssam: wikipedia says it is a die shrink of the RV630
Nightwulf|work: uses a 3650 too with 1.2.1 and doesn't have any locks
obv1us: ok, check /var/log/Xorg.0.log .. way down the driver should list the supported cards/chipsets
ssam: if i start X with radeonhd or radeon, then the machine locks up. i am unable to get to a VT
obv1us: Anything interesting in the Xorg.0.log ?
libvde: airlied: but now we have broken builds everywhere...
ssam: http://launchpadlibrarian.net/16518878/Xorg.0.log when running with VESA
libvde: ssam: a log of vesa tells us nothing :)
ssam: if i install it, will it store anything in the log before it locks up? i guess i'll have to try
libvde: 635 is quite different from the 630
libvde: maybe not on the 3d front, but a lot of stuff got changed on the display/modesetting engine
libvde: another person was talking about 620/635 locking up on him
libvde: ssam: is this a 32 or 64bit machine?
libvde: ssam: you're using the latest git?
ssam: no, 1.2.1 from a live cd
libvde: hrm... even with that? that's bad
ssam: but i am happy to do some testing on a spare partition
libvde: i asked the user from yesterday to test 1.2.1, so here we have our result i guess
obv1us: i don't mind broken builds - as long as it's the "right thing" to do.
obv1us: we can work through the problems..
libvde: ssam: well, i'll get out the cards as soon as i get to the office
libvde: obv1us: could you verify that, with our latest driver, this is also true?
ssam: is there a recommended distro for testing. something that plays well with installing hot new version od X and whatever
libvde: the information on how to get and install this, is provided in the wiki link in /topic
libvde: you don't need a hot new version of X
libvde: our driver should build and work against many X versions, apart from some typing issues recently
obv1us: libvde: I did the git clone on the weekend.
obv1us: I'll try another git pull
obv1us: "Already up-to-date."
libvde: obv1us: yeah, no commits during this weekend :)
obv1us: Xorg -version --> 7.2.0
obv1us: :) yeah I didn't think so.
ssam: ok thanks
obv1us: libvde, what exactly did you want me to verify ? that the errors I reported above are still there ? I'll try again now.
libvde: obv1us: i got you and ssam mixed up, sorry, i still need to get my heavy dose of caffeine :)
libvde: ssam: could you go and try our git version?
ssam: yes i'll give it a go. thanks
obv1us: ah ok :) it's morning for you I suppose..
obv1us: Are there any developers around ? Anything come to light on the uint*_t errors ?
ndim: obv1us: SHould be fixes since several days.
ndim: obv1us: http://gitweb.freedesktop.org/?p=xorg/driver/xf86-video-radeonhd;a=commit;h=6b04da151fb96c6307864d040b805567e1235c0e
egbert: ndim: but you reported problems with this, too, didn't you?
ndim: egbert: I could confirm the original problems, and I could confirm that these problems went away with that commit.
ndim: So as far as I can tell, the issue is fixed.
Nightwulf|work: ndim: the work just happening in radeon_textured_video.* and radeon_xaa.* is limited to r5xx and doesn't do anything on 6xx, right?
ndim: Nightwulf|work: Yes (with 85% confidence)
Nightwulf|work: ndim: thx...i just wondered why my xorg-log told me it "forced DRI" on my r680 but glxinfo tells me it hasn't got DRI (which would be what I expect atm)
Daun: hello, I'm running a asus m2a-vm hdmi board with builtin x1200 (rs690 I believe) and just built xorg and the driver from git. I just wanted to mention that its working pretty nicely for an incomplete driver!
Daun: Looking forward to when opengl is fully implemented with this code ^_^
ndim: Daun: Thanks for the report. That is nice to here.
gentoofu: x1200 + hdmi = x1250
libv: ndim: i will be sticking your git_version.sh into -unichrome soon :)
Daun: ndim: no problem, I'll be sure to mention any problems I find down the road... gentoofu: I was just going off of what lspci told me, nice to know more about my chipset though
libv: gentoofu: heh, that's rather stupid marketing :)
libv: gentoofu: it's just a riser card for hdmi iirc
libv: suddenly the silicon on the northbridge gains 50?
Daun: libv: not sure if ati's naming scheme was ever all that logical anyway... good cards surrounded by so much confusion :/
ndim: libv: You're welcome to, even if I begin to doubt the concept and parts of the implementation. :-)
libv: why are you doubting it?
libv: it's rather nice imho
libv: we get very detailed info in the log about what the user is using
ndim: I'm seeing curious phenomena which look like errors in program logic, and some projects (like autoconf) do something very interesting using "git describe".
ndim: libv: "git clone git://fedorapeople.org/~ndim/nbb.git" for an example improved somewhat over autoconf
ndim: I guess something unifying both methods would be possible and give optimum information... but I don't have the patience to do that right now :)
ssam: hi i was having trouble earlier with a 3650. i now have the latest GIT of radeonhd install, and it still locks my machine (sometimes reboots it). no log gets saved to disk
ssam: is there anything i can do to help debug
peterbrestak: who had radeonHD for download?
ssam: i have tried sshing in when the computer locks up, but it does not respond
pb1302: where do i get RadeonHD to install my 2600PRO
ssam: or from your distro
pb1302: ok... well went to the site and have NO idea what to click on to get the needed files.
yangman: pb1302: that's because there is no such a link. I recommend you check to see if your distro already has it packaged
pb1302: Kalyway 10.5.1... not so much
Obscene_CNN: pb1302, look at section 6 on http://www.x.org/wiki/radeonhd
yangman: er... hackingtosh?
pb1302: so i just get that GIT thing installed and then use GIT to retrieve it?
pb1302: hackintosh yes....
yangman: you're on the wrong IRC server
yangman: you're looking for #radeonhd on, IIRC, irc.osx86.hu
Obscene_CNN: you can use git or download a tar ball
pb1302: cool thanks... much appreciated for the help
Obscene_CNN: git would be more current
yangman: np. we've had a lot of people like you come through, although not so much recently
pb1302: off i go...
airlied: Daun: that card supports GL fine now.
airlied: Daun: at least as much as any r300 engine with no vertex shader hw.
Daun: airlied: oh, nifty... I haven't had a chance to really try it out yet but that's nice to know
libv: CS branch just got exa composite/render accel :)
ndim: CS stands for command submission?
rehabdoll: any news on r600 docs?
Obscene_CNN: hmmm... 6 new git commits
libv: Obscene_CNN: :)
ndim: Obscene_CNN: On the CS branch, just like libv reported 17 minutes ago
Honk: what'3s the CS branch anyway? ;P
Obscene_CNN: he didn't say there were 6 of them though ;-)
libv: Honk: clean infrastructure to submit register writes and command packets into the hardware :)
Honk: ah yes
Obscene_CNN: are there docs available yet for RS780 acceleration?
yangman: getting endless "R300CheckComposite: Component alpha not supported with source alpha and source value blending." in Xorg.log with the new CS HEAD
yangman: hm... maybe not endless, but enough to have the logfile expand in size an worrying amount
ndim: libv: You might want to read yangman in the backlog when you are back.
libv: let me check that
libv: hrm... i should silence the RADEON_FALLBACK thing, apparently
libv: weird thing is, some of those RADEON_FALLBACK things seem to point to a broken exa implementation or broken setup, others are meant to happen, apparently
libv: if only this was used in a more discriminating manner
libv: yangman: testbuilding a "silent" version now
marcheu: libv: well that's just a fallback, so it's not broken
marcheu: libv: it just means you should implement it if you want full accel
marcheu: and yeah, disabling fallback messages is indeed a good idea, only devs want to know what fallbacks
libv: but most of the things hidden behind the fallback messages here are more errors than anything else
marcheu: hmm no, if you decide to fallback it means you didn't implement something, ti's not an error
marcheu: sometimes the hw can't do it
libv: sure, but not all the things hidden behind RADEON_FALLBACK are actual fallbacks :)
libv: also, my upload function at one point borked, as it got told to upload a 0 size pixmap :)
marcheu: I guess it's better than no code at all. IME the fallbacks are very convenient in debugging the misc functions
libv: i'm sure it was because it was an older server, but it was kind of surprising
marcheu: don't complain, some old EXA used to send you pixmaps with negative coods :)
marcheu: (to copy)
marcheu: but now EXA is really solid & fast
marcheu: it's becoming kinda boring actually :)
libv: yangman: pushed btw
yangman: libv: yeah, I saw the email. thanks
libv: marcheu: :)
yangman: I'm procrastinating studying for my final exam. I've really got no other reason to try the branches ;)
libv: yangman: hehe :) good thing you did though, as more people will run into this apparently
libv: anyway, off home now, almost 01:00 here
DJ-N: how can i checkout CS branch with git? i tried $ git checkout CS but nothing happens
cire: has ATI/AMD passed the necessary documents for the R6xx 3D support to the developers yet?
airlied: cire: some of us have pre-release docs under NDA.
cire: ahh, okay. I often read about the improvements at phoronix, but mostly it is talked about R5xx support, that's why I am asking.
libvde: cire: CS was major infrastructural work, which will pay off for r6xx and up as well
airlied: libvde: how much faster/cpu usage lower is MMIO vs CP in VRAM?
libvde: airlied: so far i spent most of my time running into the wall that is rs690
cire: great to hear. In the end I don't really need 3D support, but was just wondering. At the moment I really love the driver's functionality - especially xrandr support and those small autoconfig xorg.conf files :)
cire: Dual Head was never easier.
libvde: airlied: CSQ vs RBBM interaction is broken, and i have only a few alleys left to try
airlied: libvde: ouch.. is it some coherency problem with RBs in VRAM?
airlied: libvde: or just generally busted..
libvde: airlied: but... it is more about being able to do render and textured video without drm (older distributions) than anything else
libvde: airlied: i tried main, indirect _and_ feeding the CSQ directly, all the same thing
airlied: for old RH we just backport the drm as well.
airlied: its not like kernel team isn't used to backporting drivers.
airlied: granted for unsupported users it might be useful indeed.
Daun: I can hardly believe it but I managed to get these drivers to run world of warcraft
Daun: ... kind of
Daun: I didn't think it would work personally without the shaders and such
Daun: definitely not playable yet though, but that's alright
edman007: is it a known problem that radeonhd does not initialize my vid card properly (hd 3870), i have to attempt fglrx (which crashes) and then try radeonhd, i'm on 64-bit and using xorg 1.4.2 along with the radeonhd that ships with it
Diablo-D3: did you file a boog?
edman007: figured it might be a bit old too, its not the latest and greatest
obv1us: who -c #radeonhd
edman007: is not here