Bug 10121

Created attachment 8898 [details]
my xorg.conf

Created attachment 8899 [details]
Xorg.0.log with enabled DDC

Created attachment 8900 [details]
Xorg.0.log with disabled DDC

i had the exact same problem with my 855gm, but http://people.freedesktop.org/~anholt/i8xx-clocks.diff fixed it for me; give that a try.

Thanks for your advice but I had no luck. I'll attach the new xorg log and a diff between the different logs. (one with the clockfix, one without)

Created attachment 8906 [details]
Xorg.0.log with enabled DDC and applied clockfix patch

Created attachment 8907 [details]
diff between the xorg logfiles (with and without clockfix patch)

Is that from a fresh boot?

okay, thanks for trying!

(did you do this after a fresh boot? if not, could you please try that?)

I tried it again with a fresh boot and experienced no difference. The logfiles are identical. (I can also show up on IRC if needed)

Yeah, if you could hop on #xorg-devel of freenode.net, that would be great.

Created attachment 8911 [details]
xf86-video-intel-VBIOS - as requested on irc

Created attachment 8912 [details]
log with i8xx-clocks.diff

this one works for me with i8xx-clocks.diff: x40, 1024x768, 855gm.

OK, I finished that program I'd talked about writing today.  So, if you can install libpciaccess, then update and rebuild the modesetting branch, you should get a program in xf86-video-intel/src/reg_dumper/intel_reg_dumper.  Run that as root from the console and when the working Xorg is running, and attach those logs, and hopefully that'll help me figure out what's going on here.

Created attachment 8924 [details]
output from intel_reg_dumper

After adding #include <stdint.h> into reg_dumper.h compiling worked fine. (Daniel Stone gave me the hint.)

I also commented out the following lines in main.c:
//    if (dev->vendor_id != 0x8086)
//      errx(1, "Graphics card is non-intel");

Regards,
Tobias

Created attachment 8925 [details]
output from intel_reg_dumper while on vt

Proposed patch for i855 LVDS issues at http://people.freedesktop.org/~anholt/i8xx-clocks-magic.diff -- please test with the latest driver from master with that diff applied.

(still works for me: hasn't broken the committed driver.)

Created attachment 9043 [details]
Xorg.0.log with enabled DDC and applied clockfix-magic patch

Still the same symptons. (except that I can switch back to VT after a fresh boot)

Created attachment 9044 [details]
diff between xorg logfiles (with i8xx-clocks patch and i8xx-clocks-magic patch)

Created attachment 9051 [details]
Xorg.0.log with enabled DDC and applied clockfix-magic patch (xserver 1.2.99.901 (1.3.0 RC 1))

I installed the latest experimental xserver packages available from my distro. The screen is now completly black and switching to vt does not work. The log is from a fresh boot.

Created attachment 9123 [details]
log from i855GM, i855-clocks-magic.diff applied to 14ee9...

this one still works for me, but here's the log.

OK, master is updated with the clocks-magic patch plus some additional changes that could help with mysterious issues.

Ok, I'll give it a try within the next days.

Compiling the driver by myself resulted in a crash of the server. So I tried the latest experimental builds from my distro containing xserver 1.2.99.902 and xf86-video-intel 12.92.92 (which seems to be 1.9.92) and everything except the DPI settig works like a charm. Really great!

Currently the DPI settings seems to be wrong. The Xorg.0.log says "(==) intel(0): DPI set to (75, 75)", but I believe it is not. My fonts are much larger. If I execute "xrandr --verbose --dpi 75/LVDS" I get "screen 0: 1024x768 346x260 mm  75.00dpi", after that new opened windows have the correct font sizes (the same as I had with the old driver). One more thing: If 346x260 is the screen size in millimeters this is wrong. It is ~305x229.

Again thanks so far.

The wrong font size might be concerned to https://bugs.freedesktop.org/show_bug.cgi?id=10304 - so I believe my original report seems to be fixed.

Original issue is fixed, so closing.  Follow up in a separate bug for display size issues.

I want to give RC3 a try. If then the issue is still there I'll file a new bug.