hi michael,

> Hello,
> 
> 
>>i finally ran the X server provided by the URL above and the kernel
>>patches you sent me. thanks a lot!
>>
>>now, exiting X, my monitor does not complain on the sync frequency out
>>of range (it says 268kHz ;), using your patches/X server it says that it
>>does not get *vertical* sync -- horizontal sync is okay. furthermore,
>>the signal seems not out of range (AFAICS).
> 
> Hmm, so the mode restoration still goes wrong - too bad I don't have a radeon or a G4 Mac to see what exactly happens. 
> Is the result in any way different from the old Xserver? 

yes, as i said in the lines above, there is a difference in the signals 
the gfx card pumps to the monitor after leaving X...

> 
> 
>>i'll provide my X logfile, maybe you are able to localize the problem...
> 
> 
>>(--) Assigning device section with no busID to primary device
>>(--) Chipset ATI Radeon 9000/PRO If (AGP/PCI) found
> 
> 
>>(II) RADEON(0): MMIO registers at 0x90000000
>>(WW) xf86EnableIO 6
> 
> Ok, so it was able to map what it thinks is the PCI IO space. 
> 
> 
>>(II) RADEON(0): vgaHWGetIOBase: hwp->IOBase is 0x03b0, hwp->PIOOffset is
>>0x0000
> 
> That's the VGA register stuff... I wonder why it's 0x3b0 - it should be 0x3d0. 0x3b0 is used by original VGAs in monochrome mode, 0x3d0 is for colour ( don't ask me why... ) but maybe the Radeon always uses 0x3b0 or both or whatever.
> 
> So, what could have happened is this:
> - the kernel patch didn't really work for you and didn't map the right address range.
> - the radeon driver doesn't use the IO instruction stubs provided by XFree and so it still can't access the VGA registers ( should be easy to verify, depends on the code )
> - the VGA register base is wrong ( not too likely since it was able to change resolution, but maybe it just stores what it finds in the VGA registers and writes it back on exit hoping to restore the video mode - this would fail if the register base is indeed wrong )
> 
> I'm afraid I can't do a lot without a radeon.

where are you located? maybe we can provide you a gfx card (will do 
*any* type of radeon?)...

> So I'll write a small program that just dumps the address space it gets when asking ofb for the IO range - this should tell us if the problem is in the kernel or in the Xserver.
> 

please notify me when this lil proggie exists -- i am very willing to 
test and get this stuff running ;)

> thanks tor trying
> Michael

michael, thank you *very much* for your help -- your work is very much 
appreciated! thanks again!

regards,

timo