Accepted xresprobe 0.4.24ubuntu1 (source)

[Matt Zimmerman]

On Thu, Mar 15, 2007 at 04:54:59PM +0200, Timo Aaltonen wrote:
> On Sat, 10 Mar 2007, Matt Zimmerman wrote:
> 
> > On Thu, Mar 08, 2007 at 12:14:05PM +0200, Timo Aaltonen wrote:
> >> On Wed, 7 Mar 2007, Matt Zimmerman wrote:
> >>> Does this provide us with any useful information?  Does the VESA BIOS
> >>> accurately report the size of an attached LCD?
> >>
> >> No it doesn't, but the server autoprobes the best values it can use.
> >
> > How does it probe this when it's using the VESA driver, if not by way of the
> > BIOS?
> 
> Those are two different things :) The panel can be a gigantic widescreen 
> which the vesa-driver can't drive with it's native resolution. Then the 
> driver would use the best mode the BIOS has. In the case of TP Z61p which 
> has a 1920x1200 capable panel it would use 1600x1200.. stretched but 
> better than 1024x768.

Do you mean to say that the modes reported by the BIOS are not accurate, but
they are useful?  e.g., if the panel is 1920x1200, it will provide larger
modes than if the panel is 1024x768?

> > We should only be changing this part of the infrastructure at this point in
> > the release cycle in very well-defined and predictable ways, and hardware
> > detection is especially tricky.  The change should definitely be limited
> > to cases known to be handled incorrectly if that is possible, and they
> > should also be scrutinized to ensure that they're improving the situation.
> 
> Right, so limiting the autoprobing to vesa is ok?

That seems reasonable.  Note that this now requires a beta freeze exception
to change.  Your original changes have now been in the archive for over a
week; have you received any feedback, especially from laptops which were
being detected correctly in the past?

> > Again, can you describe the problem this change is meant to solve?  Are
> > there bug reports which are relevant to this problem?  How did you decide to
> > make this change?
> 
> I've been trying to find the bug but can't... At least all radeon r5xx 
> chip owners are affected, but also others where the panel allows a better 
> resolution than 1024x768 and the driver used is vesa (for one reason or 
> another).
> 
> The problem in a nutshell: previous xresprobe probes some settings running 
> X with a barebones xorg.conf (with 1024x768 resolution) and then greps 
> some values from the resulting log. Since the vesa-driver cannot identify 
> the panel and it's dimensions, the resulting xorg.conf always uses 
> fallback resolution which is 1024x768. And that is what the change in 
> xresprobe is trying to fix.

Right.  The problem is only with the vesa driver, but the config change
affects *all* drivers, even those which have been working correctly.  It's
those cases we want to avoid changing when we are approaching the release.

-- 
 - mdz