[PATCH 3/4] xfree86: fbdevhw: remove superfluous pci pointer from probe function

[Vignatti Tiago (Nokia-MS/Helsinki)]

On Thu, Sep 09, 2010 at 10:00:40PM +0200, ext Matt Turner wrote:
> On Thu, Sep 9, 2010 at 3:57 PM, Mark Kettenis <mark.kettenis at xs4all.nl> wrote:
> >> Date: Thu, 9 Sep 2010 11:01:59 +0300
> >> From: "Vignatti Tiago (Nokia-MS/Helsinki)" <tiago.vignatti at nokia.com>
> >>
> >> Mark,
> >>
> >> On Tue, Sep 07, 2010 at 11:25:26PM +0200, ext Mark Kettenis wrote:
> >> > > From: Tiago Vignatti <tiago.vignatti at nokia.com>
> >> > > Date: Tue,  7 Sep 2010 15:18:48 +0300
> >> > >
> >> > > The only drivers using this function are fbdev and glint. Though those driver
> >> > > initialize with NULL argument. Other drivers do some kind of implicit probe
> >> > > using fbdevHWInit instead.
> >> > >
> >> > > API break.
> >> >
> >> > Sorry, but I really fail to see the purpose of this change.
> >>
> >> It's not used by anyone, it would open space for other major
> >> cleanup, then why not just kill? I thought I was clear, sorry.
> >
> > I'm not conviced.  I don't see at all how this helps you to clean
> > things up.  And actually this sort of ABI break is the worst kind
> > imaginable.  Changing the signature of a function means that the
> > dynamic linker will not notice a problem, but when old binaries
> > actually call the function, you'll almost certainly crash.
> 
>  This seems like a vacuously true statement.

and I don't have breath anymore for those :/


> Maybe I'm wrong, but aren't the two drivers that call this function
> going to _have_ to be recompiled against a new xserver anyway?

yes. And it's an API break. 


             Tiago