nha: who is webmaster of radeonhd.org, and do you think it would be possible to sort the IRC archive listings (http://www.radeonhd.org/?page=archive_listings&m=7&y=2008) by date?
glisse: nha: it's Michael of phoronix i think
nha: michaellarabel: ping?
nha: ah yes, it says powered by phoronix.com, so it seems likely
ndim: nha: Hehe. Those look to be sorted by alphabetical MD5 hash or something :)
michaellarabel: nha: Yes it can be improved, right now it's not sorted at all
cjb: Hi! The tinderbox is seeing odd build failures on radeonhd.
cjb: On one machine, they went away after restarting from scratch. On another, they haven't. One of the suspicious messages is:
cjb: fatal: ref HEAD is not a symbolic ref
cjb: git_version.sh: Output is new file, creating git_version.h
cjb: And then the errors seen are:
cjb: radeon_video.c: In function 'RADEONCopyMungedData':
cjb: radeon_video.c:427: error: 'struct RHDRec' has no member named 'surface_cntl'
cjb: radeon_video.c:465: error: 'struct RHDRec' has no member named 'ModeReg'
cjb: Any ideas?
yangman: cjb: have you tried rerunning autogen.sh?
cjb: yangman: I've tried fully deleting the build environment.
libvde: maybe, just maybe, quick and dirty 2d was not ready for master
agd5f: I never tested the build on big endian
agd5f: fix is coming
cjb: agd5f: oh, cool, okay. yes, that machine is ppc.
cjb: I didn't realize endian problems can cause struct members to go away.
agd5f: cjb: ifdef big endian stuff
agd5f: cjb: should be fixed
cjb: agd5f: Great. Actually, there's a tinderbox run that should get to radeonhd in about five minutes; will let you know.
cjb: agd5f: Seems to have failed again in the same way.
cjb: radeon_accelfuncs.c: In function 'RADEONSetupForScanlineCPUToScreenColorExpandFillCP':
cjb: radeon_accelfuncs.c:780: error: 'struct RHDRec' has no member named 'ChipFamily'
cjb: radeon_accelfuncs.c:780: error: 'CHIP_FAMILY_R300' undeclared (first use in this function)
cjb: radeon_accelfuncs.c: In function 'RADEONSubsequentScanlineCP':
cjb: radeon_accelfuncs.c:911: error: 'struct RHDRec' has no member named 'ChipFamily'
cjb: radeon_accelfuncs.c:911: error: 'CHIP_FAMILY_R300' undeclared (first use in this function)
cjb: radeon_accelfuncs.c:917: error: 'struct RHDRec' has no member named 'scanline_bpp'
cjb: radeon_accelfuncs.c: In function 'RADEONSetupForScanlineImageWriteCP':
cjb: radeon_accelfuncs.c:981: error: 'struct RHDRec' has no member named 'ChipFamily'
ndim: cjb: How does "the tinderbox" (whatever that may be) get the sources? I'm interested in the "fatal: ref HEAD is not a symbolic ref" thing, even though it should not hurt the build in any way but print a not quite correct source code version.
cjb: ndim: I think everything it's running is shown in the build log: http://tinderbox.freedesktop.org/builds/2008-07-27-0020/logs/xf86-video-radeonhd/#build
ndim: agd5f, libvde: Except for the first ifdef issue, quick and dirty 2d works awesomely for me :)
cjb: so, git checkout origin/master, git pull origin master. and the first time, it does a straight git clone.
cjb: Maybe the git pull line is responsible for HEAD being odd?
ndim: cjb: Yupp, the working tree is not a branch.
cjb: interesting. that code comes from jhbuild.
ndim: "git checkout origin/master" checks out the remote master, and "git pull origin master" does not check that.
ndim: "git checkout origin/master" checks out the remote master, and "git pull origin master" does not change that.
cjb: Right. It should do git checkout -b master or something?
ndim: So "git branch" tells us "(no branch)".
ndim: cjb: I'm not sure what the jhbuild is attempting to achieve, but I would do a a clone initially (which gives you a checked out local master branch tracking origin/master).
ndim: To update, I'd run "get pull", still in the master branch, which will then fetch the new commits from origin/master and fast-forward the local master to origin/master.
cjb: yup. We do the clone, so sounds like it's just switching to git pull. I should give that a try.
ndim: Might depend on git >= 1.5, though.
ndim: Anyway... if you don't have a branch checked out, it's not my fault when none is detected.
agd5f: cjb: ok fixed
cjb: thanks :)
cjb: agd5f: same error on the next run, I'm afraid
brot: hi everyone. i have a radeon hd3850. How well is that card supported for some basic desktop usage (kde4)
kurtr: I'm hoping to find help here with this Xpress 200 on a Dell Vostro 1000 laptop. Am I in the right place?
ndim: kurtr: #radeon is the right place.
roberth: havent amd delivered doc for 2d accel, textured video for r600 yet?
airlied: roberth: there is no 2D or video docs.
airlied: roberth: its all 3D
roberth: airlied: okay, and amd hasnt delivered that yet for r600?
airlied: the 3D docs are quite large, so they haven't been released yet.
roberth: okay, when was the last documentation drop from amd?
airlied: roberth: they released some of the r600 ISA docs not long ago
airlied: in June..
roberth: one more question, is this driver ment to support r700 in the same way as it supports r500 and r600?
airlied: roberth: yes r700 is similiar to r600
roberth: will r700 require its own 3d accel documentation?
airlied: not likely.
bridgman: we're actually using 700 to help figure out how to get 600 running ;)
matteo: have you r700 docs?
matteo: my 4870 will arrive tomorrow
matteo: i'd like to have it working just after installing it
matteo: even non accelerated
bridgman: non-accel should work, and if it doesn't it won't be hard to make it work
bridgman: differences between 6xx and 7xx are pretty small from programming POV
roberth: cant wait untill radeonhd supports 2d and xv with r600, tired of the xv tearing with the propietary one
airlied: roberth: we have teraing with the open one as well, its a pain..
roberth: airlied: ...what is causing it?
airlied: roberth: we don't have any way to sync to vblank inside the X server.
airlied: or at least no nice way.
roberth: airlied: is that the same issue for the propietary one?
airlied: roberth: pretty much..
airlied: roberth: the future plan is just to use a compositing manager that syncs to vblank.
netman87: is it possible to run mobility radeon hd 2300 dri/3d
bridgman: I think so; 2300 is like an rv515 for most programming IIRC - I think the IDs were added recently
netman87: then i need help to set this shit up :)
bridgman: ok, where are you now - driver installed & doesn't work ?
netman87: xorg.0.log: http://pastebin.com/m5c279a3e
netman87: drm,mesa,radeonhd all compiled from git
netman87: 2.6.26 krnl,debian unstable
netman87: what information needed?
bridgman: are there any symptoms which are not obvious from the log ?
netman87: sorry, cant understand..
bridgman: what is it doing, ie what is the problem ?
netman87: everythink looks like OK, but direct rendering just not..
netman87: any files where i can find some information about that?
netman87: or post to pastebin
bridgman: this is one of the many areas where I don't know enough -- looks like the driver is coming up fine but there's probably an "always indirect" flag set somewhere and I don't know where it hides
netman87: jeah, i hope there not gonna be loong waiting
netman87: i just cant install fglrx (Xorg crash if using opengl or compile errors)
glisse: netman87: do LIBGL_DEBUG=verbose glxinfo
airlied: sounds like missing pciids probably
glisse: as adviced in the output
glisse: but might be that
airlied: where is the Xorg log
bridgman: his first pastebin was xorg log, second pastebin was (theoretically) verbose glxinfo
airlied: bridgman: nice naming on the hd 2300 :)
bridgman: yes, I'm sure someone is very proud of that
netman87: 5sec im gonna restart X then i post xorg.conf,xorg.log to pastebin
bridgman: at least internally it was the rv550
ndim: Hmm. Sometimes b0f470aa produces intermittent image corruption in 2D stuff, and flickering noise lines in 3D.
netman87: what a fck
netman87: there is some shit from fglrx :O
netman87: ... from is maybe wrong word...
bridgman: ahh, that would explain it
netman87: how i can remove it?
netman87: some fglrx-powersave / fglrx-powermode
bridgman: did it show up in the logs ?
bridgman: let me step back -- did you have fglrx installed earlier, and did you uninstall or just install a new driver over top ?
netman87: i have fglrx earlier
netman87: installed with amd intaller and from apt-get
netman87: then i remove it with apt-get
netman87: looks like there is somethink and somewhere in my system from fglrx package
bridgman: where did you see fglrx appear ? I didn't see it in the xorg log
netman87: bridgman: i boot just... then i trying to shutdown X and check xorg.conf then trying to start again but there was one already
netman87: so i type top
netman87: Xorg taking totally one core and fglrx-power* named procs was there
bridgman: ahh -- can you do top >
bridgman: or won't that work...
bridgman: have you rebooted since you removed fglrx ?
netman87: few times
netman87: i think i cant
bridgman: ok, if there's nobody on here right now who remembers the tricks for removing fglrx you might want to post on Phoronix, there are some real fglrx exorcism experts there ;)
netman87: ah, its connected acpi
netman87: bridgman: open url again and look fglrx named file
netman87: i can just delete those files?
bridgman: I would try renaming /etc/acpi/fglrx-powermode.sh
bridgman: I tend to regret deleting lots of files
bridgman: netman87, can you do the same for usr/local/lib/libdrm* ?
netman87: damn, my brother wares on run
netman87: sec vnc 2 close torrent
bridgman: also do you have a usr/lib/fglrx ?
netman87: need to sleep
bridgman: ok, will just leave you with a thought... guessing that usr/lib/libdrm.so.2.3.1 is the problem but only a guess
netman87: i can try somethink
netman87: 10mins and i put my laptop to sleep
bridgman: I don't really know what all the versions of libdrm mean but the 2.3.1 version seems to be smaller than the versions you presumably built in usr/local/lib
bridgman: what I don't know is whether something is hardcoded to pick up 2.3.1 or whether it just picks up the highest version in usr/lib or if this is totally the wrong direction
bridgman: but it seems to me that renaming the 2.3.1 version to something totally different would be interesting
bridgman: then again I'm only guessing that what you built in usr/local/lib is correct...
bridgman: note to self... need to play with DRI some more
bridgman: ok, gotta run for a while, good luck
cire: hi guys. Yesterday I wondered about my transformed cursor with radeonhd. Today I was using radeon (not radeonhd), and I have the exact same problem. (messed up cursor). Seems to be an X issue. Just for your information. (ndim and Fuddl)
ndim: cire: Ah, thanks.
cire: no problem ;)
ndim: cire: However, it could also be code shared between radeon and radeonhd. :)
cire: yeah - in this case two teams would be tracking this bug, which is good for me ;)
cjb: agd5f: radeon_exa.c: In function 'RADEONPrepareAccess':
cjb: radeon_exa.c:237: error: 'drmRadeonSurfaceAlloc' undeclared (first use in this function)
cjb: agd5f: radeon_exa.c:240: error: 'drmsurfalloc' undeclared (first use in this function)
cjb: radeon_exa.c: In function 'RADEONFinishAccess':
cjb: radeon_exa.c:279: error: 'drmRadeonSurfaceFree' undeclared (first use in this function)
dmb: fail :P
dmb: cjb, did you install libdrm?
dmb: oh hmm
agd5f: cjb: none of the surface stuff is ported over yet to radeonhd
agd5f: I'll see what I can do