modular -> monolithic

[Dave Airlie]

On Jan 22, 2008 1:29 PM, KAMALNEET SINGH <kamalneet.s at samsung.com> wrote:
>  David Miller wrote:
> > From: Eric Anholt <eric at anholt.net>
> > Date: Mon, 21 Jan 2008 17:14:27 -0800
> >
> >> Some portions of the tree, sure, using the rough criteria of hardware
> >> that we can't find any instances of and therefore can't test.
> >
> > You don't need access to the hardware to build the driver source.
> >
> > Issues like testability and what drivers to mark as "deprecated"
> > or whatever are secondary to making sure the build works.
>
> In my opinion, if you cannot be confident that the driver actually runs
> properly after the change, it is of little use to make it build. Making
> it build may make it easier for someone else to fix it later. But for
> others, it is much more frustrating to find that something doesn't run
> when you expect it to. It is less frustrating to find that the driver
> doesn't build, it is clear that it is not tested with the current X
> code.
>

But you lose out on potential testers if they can't even get the thing
to build, a building driver will allow people to report regression, a
driver that doesn't build is just a useless piece of code, considering
mostly the only person who can quickly fix them is the person who
broke them in the first place..

testers aren't developers and nor should we expect them to be.. I've
done sis and cirrus pciaccess ports with no hardware but one willing
tester..

Dave.