Further discussion of default video quirks

[Matthew Garrett]

On Sat, Nov 29, 2008 at 11:25:08PM +0100, Danny Kukawka wrote:

Independently, though:

> As already said in the thread (IIRC): 
> - As you already pointed out: This can break machines which may work without 
> any quirks. And there are such machines, we have seen such cases in the past, 
> where quirks prevented a machine from suspend/resume which worked without any 
> quirk.

The only cases anyone's brought up were using the Intel driver. Running 
the quirks on Intel is the wrong thing to do anyway, so pm-utils strips 
them out. As a result, setting them by default can't break a machine.

> - This combination isn't really common (I couldn't find any machine with this 
> combination in a first looks at the quirk list), it would make more sense (if 
> any) to find vendor specific default quirks, since the chance is much better 
> to get a working machine.

The perfect is the enemy of the good enough. It's possible that 
optimising the list per-vendor would result in a larger set of working 
machines, but covering all vendors would be a substantial amount of work 
for undetermined benefit. Again, the idea is to improve the current 
situation rather than anything else. If users need different values, 
they'll still need to submit their own data.

> - It prevents may finding bugs/problems with suspend. Since it may make 
> machines working which currently don't work because of e.g. a kernel bug.

The quirks aren't run on any drivers that have suspend/resume methods 
that are expected to reprogram the video, so there's no risk of them 
hiding kernel bugs.

I think it would be a mistake for any distribution to fail to ship this 
file, but that's obviously a question for the individual distributions.
-- 
Matthew Garrett | mjg59 at srcf.ucam.org