Created attachment 11969 [details]
my xorg.conf

Created attachment 11970 [details]
log diff between working 6.6.3 and failling 6.7.194

Btw, I've another very same computer but with a CRT instead of flat screen and there's no sync bug there.

For the record, it happens since at least 6.7.191 but it may have happened with 6.6.19x too)

Can you please attach the full logs?

*** Bug 12771 has been marked as a duplicate of this bug. ***

Created attachment 11972 [details]
full xorg log

I've just committed a fix to ati git which should at least get the connector table right, however it seems DDC tends to not work on XPRESS chips so you probably won't get an edid, as such you'll have to manually specify an monitor section and preferred mode.  Something like this:

Section "Monitor"
    Identifier "monitor1"
    VendorName "Generic"
    ModelName "Flat Panel 1280x1024"
    HorizSync 31.5-90
#    VertRefresh 60
    VertRefresh 50-70

Modeline "1280x1024"   90.75  1280 1328 1360 1440  1024 1027 1034 1054 +hsync -vsync
EndSection

Section "Device"
    Identifier "device1"
    VendorName "ATI Technologies Inc"
    BoardName "ATI Radeon 9250 and earlier"
    Driver "ati"
    BusID "PCI:1:5:0"
    Option "DPMS"
    Option "XaaNoOffscreenPixmaps" "1"
    Option "monitor-DVI-0" "monitor1"
EndSection

Current git fixed the CRT bug (#12771). I'll check for the LCD one

Created attachment 11988 [details]
log with the LCD: it failed miserably (worse than 6.7.193: no screen found)

Please attach full logs rather than diffs.

Also, as I said you will need to edit your config as per comment #8 since DDC is not working on your chip.

which changes nothing:

@@ -510,7 +512,7 @@
 (II) RADEON(0): Port1: DDCType-1, DACType-2, TMDSType-1, ConnectorType-4
 (II) RADEON(0): Output VGA-0 using monitor section monitor1
 (II) RADEON(0): I2C bus "MONID" initialized.
-(II) RADEON(0): Output DVI-0 using monitor section monitor1
+(II) RADEON(0): Output DVI-0 has no monitor section
 (WW) RADEON(0): No External TMDS Table found
 (II) RADEON(0): I2C bus "DVO" initialized.
 (II) RADEON(0): I2C device "DVO:RADEON DVO Controller" registered at address 0x70.

Created attachment 11990 [details]
6.7.195's logs

<sigh> seems the server doesn't parse the monitor section properly (or at all for that matter).  This is a server bug.  I'll see if I can look into it this weekend.

In the meantime, you can add the following to radeon_output.c::radeon_detect() ~line 1649:

if (radeon_output->MonType == MT_DFP)
    return XF86OutputStatusConnected;

that should at least get the driver up.

in RADEONProbeOutputModes

Backtrace:
0: /etc/X11/X(xf86SigHandler+0x6d) [0x4806cd]
1: /lib64/libc.so.6 [0x34a64301c0]
2: /etc/X11/X [0x4b657e]
3: /etc/X11/X(xf86DoEDID_DDC2+0xef) [0x4b678f]
4: /usr/lib64/xorg/modules/drivers//radeon_drv.so(RADEONProbeOutputModes+0x359) [0x2b68d90282a9]
5: /etc/X11/X(xf86ProbeOutputModes+0x252) [0x4ad7d2]
6: /etc/X11/X(xf86InitialConfiguration+0x78) [0x4aeab8]
7: /usr/lib64/xorg/modules/drivers//radeon_drv.so(RADEONPreInit+0xa18) [0x2b68d9011778]
8: /etc/X11/X(InitOutput+0x9aa) [0x4630ea]
9: /etc/X11/X(main+0x295) [0x433db5]
10: /lib64/libc.so.6(__libc_start_main+0xf4) [0x34a641dae4]
11: /etc/X11/X(FontFileCompleteXLFD+0x229) [0x4332d9]

Fatal server error:
Caught signal 11.  Server aborting

(gdb)
#0  DDCRead_DDC2 (scrnIndex=0, pBus=0x0, start=0, len=128) at xf86DDC.c:344
#1  0x00000000004b678f in xf86DoEDID_DDC2 (scrnIndex=0, pBus=0x0)
    at xf86DDC.c:299
#2  0x00002b3bd23f42a9 in RADEONProbeOutputModes (output=0xa093a0)
    at radeon_modes.c:227
#3  0x00000000004ad7d2 in xf86ProbeOutputModes (scrn=0x9f26f0, maxX=2560,
    maxY=1200) at xf86Crtc.c:1307
#4  0x00000000004aeab8 in xf86InitialConfiguration (scrn=0x0, canGrow=0)
    at xf86Crtc.c:1569
#5  0x00002b3bd23dd778 in RADEONPreInit (pScrn=0x9f26f0,
    flags=<value optimized out>) at radeon_driver.c:2782
#6  0x00000000004630ea in InitOutput (pScreenInfo=0x7da120, argc=2,
    argv=0x7fffd9e81038) at xf86Init.c:601
#7  0x0000000000433db5 in main (argc=2, argv=0x7fffd9e81038,
    envp=0x7fffd9e81050) at main.c:370

whoops, sorry, in radeon_modes.c::RADEONProbeOutputModes() you can do one of the following (in addition to the previous change):

1.add:

    return modes;

between lines 226 and 227

or 

2. add:

    RADEONAddScreenModes(output, &modes);
    return modes;

between lines 226 and 227

1. will just return a null set of nodes and the server will add some defaults.  2. will add the modes specified in the screen section of your config.

the result should look like this:

if (radeon_output->type == OUTPUT_DVI || radeon_output->type == OUTPUT_VGA) {
    RADEONAddScreenModes(output, &modes);
    return modes;
}

Can't you just provide a patch please?

(In reply to comment #19)
> Can't you just provide a patch please?
> 

I can provide a patch later today, but I don't have access to the source at the moment.

hmmm...something else just occurred to me, perhaps the XPRESS chips are using the LVDS interface for DVI somehow.  Try adding the following driver options to the device section of your config running an un-patched version of the driver from git:

Option "ConnectorTable" "1,2,2,1,1,1,2,2"
Option "PanelSize" "1280x1024"

I got "no signal"...

Created attachment 11997 [details]
here's the changes in log

Ok.  I'll post some patches for you to try.  Also, in the future, please just attach the full log rather than diffs.

Can you try with all these options as well:

Option "ConnectorTable" "1,2,2,1,1,1,2,2"
Option "PanelSize" "1280x1024"
Option "LVDSBiosNativeMode" "FALSE"

no signal

Created attachment 11999 [details]
with "ConnectorTable" and "PanelSize" options

Created attachment 12000 [details]
log diff after adding "LVDSBiosNativeMode" option

Ok, I've committed everything you'll need.  try latest git with the following options in your config (try them individually):

- Option "ConnectorTable" "1,2,2,1,1,1,2,2"

- Option "ConnectorTable" "1,2,0,4,1,1,2,2"

- Option "ConnectorTable" "1,2,1,4,1,1,2,2"

I got a working server with Option "ConnectorTable" "1,2,2,1,1,1,2,2"
I'll try the others

With Option "ConnectorTable" "1,2,0,4,1,1,2,2", cursor beacmse a big blug square.
With Option "ConnectorTable" "1,2,1,4,1,1,2,2", nothing is displayed, I cannot kill the server, I cannot switch vt and I can only reboot

(In reply to comment #30)
> I got a working server with Option "ConnectorTable" "1,2,2,1,1,1,2,2"
> I'll try the others
> 

Does it work well, or is there flickering or sync problems?

> With Option "ConnectorTable" "1,2,0,4,1,1,2,2", cursor beacmse a big blug
> square.

Aside from the cursor, is the image stable and usable?

Can you attach the full logs for both of these options?

Created attachment 12018 [details] [review]
xorg log with ConnectorTable 1,2,2,1,1,1,2,2

Created attachment 12019 [details] [review]
xorg.conf for the above

(In reply to comment #32)
> Does it work well, or is there flickering or sync problems?

It works well
 
> > With Option "ConnectorTable" "1,2,0,4,1,1,2,2", cursor beacmse a big blug
> > square.
> 
> Aside from the cursor, is the image stable and usable?

I didn't make much test

(In reply to comment #35)
> (In reply to comment #32)
> > Does it work well, or is there flickering or sync problems?
> 
> It works well

Can you attach your full log with this setting?

(In reply to comment #33)
> Created an attachment (id=12018) [details]
> xorg log with ConnectorTable 1,2,2,1,1,1,2,2
> 

Stefan, can you try without the "panelsize" option?  If that doesn't help, can you also try the following connectortable settings?

Option "ConnectorTable" "1,2,0,4,1,1,2,2"

Option "ConnectorTable" "1,2,1,4,1,1,2,2"

Created attachment 12511 [details]
xorg-log-with-6.7.196

new version, new luck. I get a picture on DVI output now, but it is only 1024x768. The Display supports up to 1280x1024

(In reply to comment #38)
> Created an attachment (id=12511) [details]
> xorg-log-with-6.7.196
> 
> new version, new luck. I get a picture on DVI output now, but it is only
> 1024x768. The Display supports up to 1280x1024
> 

It should work if you manually specify the 1280x1024 mode.  the problem is that the driver is not able to get an EDID from your monitor so it doesn't know what modes it supports.  Either specify a modeline and sync ranges for your monitor in your config or add the mode at runtime with xrandr --newmode and xrandr --addmode.

Can you try xf86-video-ati from git master?

closing due to lack of feedback.