keithp at keithp.com
Sat Mar 17 12:23:01 PDT 2007
On Sat, 2007-03-17 at 13:18 +0100, Pavel Troller wrote:
> I was playing with those, tried various combinations of timeout, but no success.
> Then, I studied the code a bit more and found a RETRIES macro defined as 4. At
> first, I've changed it to 400 :-). The server started about 10 seconds then but
> EDID was always successful. Then I decreased it to 40. Now it starts faster and
> from about 20 attempts it failed once to read EDID.
40 tries is still fairly bogus; it'd be nice to know what's failing. If
you are up for it, there's a pile of I2C tracing code available in both
the driver (i830_i2c.c) and the core I2C code hw/xfree86/i2c/xf86i2c.c;
changing the I2C_TRACE and I2C_TIMEOUT macros from /*(x)*/ to (x) will
dump piles of debug output to the log; perhaps we can sift out why your
monitor is failing from that.
> Then I've found that my PreferredMode option contained a typo (1599 instead
> of 1600, yes, all the right-hand fingers were off by one :-) ). After that, it
> starts in 1600x1200 with EDID (it still uses 1152xsomething without it).
That's good to know. As reported in my previous message, the 1152 mode
is being selected as it is closer to 96 dpi than 1600x1200.
> The DPI and screen size reported are correct now. I think that the problems
> which I observed formerly were caused by the fact that the server reported
> different things depending on EDID being ok or not. This could confuse the KDE
> info center, fontconfig xft and other packages.
Ok, excellent. We'll treat everything except the I2C as solved, with
major progress (thanks for finding the timeout typo, btw) in the I2C
area as well.
Thanks very much for your patience in this process; I'm glad you've got
the 'odd' monitor and are willing to help make it work.
keith.packard at intel.com
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 189 bytes
Desc: This is a digitally signed message part
More information about the xorg