On Fri, 2006-02-17 at 18:00 -0500, Paul Donohue wrote: > > I grabbed the 12/20/05 version of the driver from cvs (I think this should > correspond to 6.9.0): > 'cvs -d :pserver:[EMAIL PROTECTED]:/cvs/xorg co -D "2005-12-20" > driver/xf86-video-ati' > I installed the autoconf2.13 and libtool packages.
Which version of automake? I suspect some of the problems are due to the autoconfage only having been added relatively recently; it might be better to contain cvs -D to the src directory and use the current code for the rest. > So, I checked in Debian Experimental, but didn't find an xserver-xorg-dev > package. I went to ubuntu and manually downloaded and installed the > xserver-xorg-dev package instead. I later found I needed a bunch of the > x11proto-<something>-dev packages as well; I installed core, fonts, xext, > xf86misc, video, xinerama, render, randr, xf86dri, and gl from > http://packages.ubuntu.com/dapper/x11/ dpkg wasn't very happy about > installing the X7.0 dev package on a system running X6.9, so I just used > 'dpkg -i --force-depends <package>.deb' > I also installed the mesa-common-dev, libdrm-dev, and libdrm2 packages from > Debian. > Then I ran 'make' again > ...and make still wasn't able to find the xorg header files... The main problem here is probably that x11-common 1:1.0.x needs to be installed before any of the modular -dev packages. The experience should have been smoother when using APT instead of dpkg directly though. > (II) Loading /usr/X11R6/lib/modules/drivers/radeon_drv.so > dlopen: /usr/X11R6/lib/modules/drivers/radeon_drv.so: undefined symbol: > gRADEONEntityIndex [...] > (II) Loading /usr/X11R6/lib/modules/drivers/radeon_drv.so > (EE) LoadModule: Module radeon does not have a radeonModuleData data object. If the corresponding ati_drv.so was installed as well, these are most likely results of the previous problems. Or mabe there are lingering incompatibilities that prevent older driver code from working with the current server. :( > If you know of a step-by-step procedure I can follow to just make this > work (so I can just pick a bunch of cvs revisions and try them), that > would be ideal. Otherwise, it might be a long while before I have > another day to spend banging my head on the desk trying to make the > cvs source work. I'm afraid I don't have any specific suggestions to offer; sorry this has been so painful. As a little consolation, let me point out that this kind of thing would usually require at least a bunch of rebuilds of the whole monolithic X tree before the modularization, and that this should be much easier once all the modularized pieces have landed in the archive. -- Earthling Michel Dänzer | Debian (powerpc), X and DRI developer Libre software enthusiast | http://svcs.affero.net/rm.php?r=daenzer
