[hal-info patch] Add default video quirks

[Danny Kukawka]

On Dienstag, 24. Juni 2008, Michael Biebl wrote:
> 2008/6/24, David Zeuthen <david at fubar.dk>:
> > On Tue, 2008-06-24 at 10:02 +0200, Danny Kukawka wrote:
> >  > On Montag, 23. Juni 2008, David Zeuthen wrote:
> >  > > On Thu, 2008-05-29 at 12:32 +0100, Matthew Garrett wrote:
> >  > > > The vast majority of machines will not resume without any quirks.
> >  > > > As a result, it makes sense to try a reasonably safe set of
> >  > > > defaults if there are none defined. Machines that explicitly
> >  > > > require that no quirks be run can still define quirks.none.
> >  > >
> >  > > Sounds useful to me; any chance you can add a comment about this in
> >  > > the fdi file including a reference to the thread before committing
> >  > > it?
> >  >
> >  > No, IMO it's not usefull. As already annotated: these quirks are no
> >  > common quirks. If you take a look at the existing quirks this
> >  > combination is never or only in rare cases used. It make absolutely no
> >  > sense to add them to hal-info.
> >
> > Sure, maybe some of the unquirked systems only need a subset of the ones
> >  Matthew selected ... but that's fine as they will resume even if there's
> >  unused quirks. I cannot understand why you are hesitant to change
> >  default behavior for unquirked machines. Let's back up a bit: do you
> >  disagree with the premise that the vast majority of machines will need
> >  quirks to resume?
>
> From my own experience, I know that two many quirks can actually break
> systems. (e.g. I had a few laptops which work with
> vbe_post/vbestate_restore and adding vbemode-restore broke suspend/resume)

I see the same problem. Also with machines wich need e.g. s3_bios/s3_mode and 
all the machines which may need no quirks at all.

> But Matthew claims that there are more systems breaking because of no
> quirks, than systems break because of too many quirks.
> It's hard to argument about that without real data.

IMO, and if you take a look at the existing data these quirks don't work at 
the most machines.

> That's why I suggested to base the default quirk set on the existing
> hal-info database and chose the combination which is used most often.

It make much more sense to base quirks on the typic/common quirks of a vendor, 
but don't add these quirks for all, this produce more trouble than it may 
solve on a small amount of the machines.

Danny