Re: the removal of Packard

On Fre, 2003-03-21 at 23:41, Christoph Hellwig wrote:
> On Fri, Mar 21, 2003 at 10:54:57PM +0100, Michel Dänzer wrote:
> > Are you aware that
> > the DRM is trying to share code between different OSs for example,
> > possibly with different standards?
> 
> Yes.  And I see how badly it's done. Take a look at XFS which is a
> project I work on during daytime and see how it's written for IRIX
> APIs and works for Linux nicely without beeing as horribly ugly
> as drm.

Well, do you think your rants are encouraging anyone to do better? I
don't.

> But it's not just the ugly code, drm has lots of other issues like
> silent changes of ioctl ABIs between XFree releases, 

I'm not aware of this happening in a while except for a minor problem in
the 4.2 i810 DRM which could be fixed after the fact. Were there others?
Somehow, I wouldn't be surprised if there were without me knowing, as
others seem to know more than DRI developers...

> and very often there seems to be a lack of understanding the Linux 
> APIs (e.g. many driver oopsed for a while because they hadn't the 
> unlock_page prototype and thus got the calling convetion wrong, it 
> opencoded some kernel functionality I don't remember anymore and thus 
> broke when the implementation was changed, etc..)

Speaking for myself, I always appreciate being educated about such
things, but I doubt ranting on random mailing lists serves this purpose.

> Another very bad thing about the DRI kernel code is the CVS menality.
> If the DRI folks would submit one patch for a change thos things could
> be fixed a lot easier (and earlier!) instead of the big merges happening
> from time to time.

Fair enough, but this also applies in the opposite direction.

> > I find it funny that I rarely if ever see these enhancements on the
> > dri-devel list, where the DRM maintainers are.
> 
> This ios mostly fixes.  Look into -ac or Red Hat's kernel tree to
> find them.

You're saying maintainers are supposed to collect fixes? I had the
impression Linux kernel development worked differently.


-- 
Earthling Michel Dänzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member   /  CS student, Free Software enthusiast