On Fri, Oct 28, 2005 at 05:00:11PM -0700, Linux Kernel wrote:
> tree 68609a74c3bc43e510f58f9c808a0a74e9d23452
> parent 4153812fc10ea91cb1a7b6ea4f4337dd211c1ef7
> author Grant Coady <[email protected]> Thu, 29 Sep 2005 11:06:40 +1000
> committer Greg Kroah-Hartman <[email protected]> Sat, 29 Oct 2005 05:36:59 -0700
>
> [PATCH] pci_ids: remove non-referenced symbols from pci_ids.h
>
> pci_ids.h cleanup: removed non-referenced symbols, compile tested
> with 'make allmodconfig'
>
> Signed-off-by: Grant Coady <[email protected]>
> Signed-off-by: Greg Kroah-Hartman <[email protected]>
This patch is removing some PCI idents from drivers that are currently
marked BROKEN on some/all architectures. It seems counterproductive
to create even more work to get those drivers fixed.
Especially in the case of for eg, the advansys scsi driver, which
actually works for some people, even though it isn't updated to use
modern scsi layer interfaces.
Dave
Dave Jones wrote:
> On Fri, Oct 28, 2005 at 05:00:11PM -0700, Linux Kernel wrote:
> > tree 68609a74c3bc43e510f58f9c808a0a74e9d23452
> > parent 4153812fc10ea91cb1a7b6ea4f4337dd211c1ef7
> > author Grant Coady <[email protected]> Thu, 29 Sep 2005 11:06:40 +1000
> > committer Greg Kroah-Hartman <[email protected]> Sat, 29 Oct 2005 05:36:59 -0700
> >
> > [PATCH] pci_ids: remove non-referenced symbols from pci_ids.h
> >
> > pci_ids.h cleanup: removed non-referenced symbols, compile tested
> > with 'make allmodconfig'
> >
> > Signed-off-by: Grant Coady <[email protected]>
> > Signed-off-by: Greg Kroah-Hartman <[email protected]>
>
> This patch is removing some PCI idents from drivers that are currently
> marked BROKEN on some/all architectures. It seems counterproductive
> to create even more work to get those drivers fixed.
Nobody cares, the drivers are dying of bit-rot :)
> Especially in the case of for eg, the advansys scsi driver, which
> actually works for some people, even though it isn't updated to use
> modern scsi layer interfaces.
Any positive suggestions? How many years does a driver remain
broken before it gets removed? These drivers don't compile cleanly
thus are not in use, no? Perhaps a set of patches scheduling
removal is in order.
Grant.
On Mon, Oct 31, 2005 at 02:38:13PM +1100, Grant Coady wrote:
> > This patch is removing some PCI idents from drivers that are currently
> > marked BROKEN on some/all architectures. It seems counterproductive
> > to create even more work to get those drivers fixed.
>
> Nobody cares, the drivers are dying of bit-rot :)
Remove the BROKEN, and it builds, and runs just fine on most systems.
(Or it least it did, until this intentional breakage occured).
> > Especially in the case of for eg, the advansys scsi driver, which
> > actually works for some people, even though it isn't updated to use
> > modern scsi layer interfaces.
>
> Any positive suggestions?
Yes. Don't remove symbols that are referenced by code in the rest of the
tree (even if it isn't buildable). It's not as though leaving those
symbols there breaks anything, or even bloats the kernel.
> How many years does a driver remain broken before it gets removed? These
> drivers don't compile cleanly thus are not in use, no? Perhaps a set of
> patches scheduling removal is in order.
At least 2 distros are carrying patches removing the BROKEN attribute
on the advansys Kconfig for some architectures. The users of those kernels
using their advansys controllers without any issue at all.
Even if this were not the case, randomly removing bits of a driver so that
it has no chance of working isn't how we schedule removal.
Dave
Dave Jones wrote:
> At least 2 distros are carrying patches removing the BROKEN attribute
> on the advansys Kconfig for some architectures. The users of those kernels
> using their advansys controllers without any issue at all.
So why are your driver patches not in mainline then?
Grant.
On Mon, Oct 31, 2005 at 04:26:06PM +1100, Grant Coady wrote:
> Dave Jones wrote:
> > At least 2 distros are carrying patches removing the BROKEN attribute
> > on the advansys Kconfig for some architectures. The users of those kernels
> > using their advansys controllers without any issue at all.
>
> So why are your driver patches not in mainline then?
Because it serves as a reminder that it needs fixing.
Dave
On Llu, 2005-10-31 at 14:38 +1100, Grant Coady wrote:
> > This patch is removing some PCI idents from drivers that are currently
> > marked BROKEN on some/all architectures. It seems counterproductive
> > to create even more work to get those drivers fixed.
>
> Nobody cares, the drivers are dying of bit-rot :)
If you look at the -mm tree you will see various drivers are getting
fixed eventually.