2.6.15-rc2 pci_ids.h cleanup is a pain

2.6.15-rc2 pci_ids.h cleanup is a pain

[David Fox]

I'm sure I'm not the only person that applies patches to the kernel that 
use some of the 500 plus PCI IDS eliminated from pci_ids.h by rc2.  I 
would like to see the PCI ids that were removed simply because the don't 
occur in the main kernel source restored.  Is there a rationale for 
removing them that I'm not aware of?

Re: 2.6.15-rc2 pci_ids.h cleanup is a pain

[Phil Oester]

On Mon, Nov 21, 2005 at 02:27:23PM -0800, David Fox wrote:
> I'm sure I'm not the only person that applies patches to the kernel that 
> use some of the 500 plus PCI IDS eliminated from pci_ids.h by rc2.  I 
> would like to see the PCI ids that were removed simply because the don't 
> occur in the main kernel source restored.  Is there a rationale for 
> removing them that I'm not aware of?

As long as you are patching the kernel, is there some reason you can't
add back the PCI IDs you require?

Phil

Re: 2.6.15-rc2 pci_ids.h cleanup is a pain

[Greg KH]

On Mon, Nov 21, 2005 at 02:27:23PM -0800, David Fox wrote:
> I'm sure I'm not the only person that applies patches to the kernel that 
> use some of the 500 plus PCI IDS eliminated from pci_ids.h by rc2.  I 
> would like to see the PCI ids that were removed simply because the don't 
> occur in the main kernel source restored.  Is there a rationale for 
> removing them that I'm not aware of?

They were not being used.  Why would you want them in there?

And, what pending patches do you have that patched this file?  Is there
a reason you have not submitted it for inclusion?

thanks,

greg k-h

Re: 2.6.15-rc2 pci_ids.h cleanup is a pain

[Andrew Morton]

Greg KH <greg@kroah.com> wrote:
>
> On Mon, Nov 21, 2005 at 02:27:23PM -0800, David Fox wrote:
> > I'm sure I'm not the only person that applies patches to the kernel that 
> > use some of the 500 plus PCI IDS eliminated from pci_ids.h by rc2.  I 
> > would like to see the PCI ids that were removed simply because the don't 
> > occur in the main kernel source restored.  Is there a rationale for 
> > removing them that I'm not aware of?
> 
> They were not being used.  Why would you want them in there?

Because they contained useful information which had been accumulated by
many people over a long period of time.

Throwing that information away seemed rather pointless, especially as the
cost of retaining it was so low.

Re: 2.6.15-rc2 pci_ids.h cleanup is a pain

[Grant Coady]

Hi Andrew, Greg,

On Tue, 22 Nov 2005 16:25:58 -0800, Andrew Morton <akpm@osdl.org> wrote:

>Greg KH <greg@kroah.com> wrote:
>>
>> On Mon, Nov 21, 2005 at 02:27:23PM -0800, David Fox wrote:
>> > I'm sure I'm not the only person that applies patches to the kernel that 
>> > use some of the 500 plus PCI IDS eliminated from pci_ids.h by rc2.  I 
>> > would like to see the PCI ids that were removed simply because the don't 
>> > occur in the main kernel source restored.  Is there a rationale for 
>> > removing them that I'm not aware of?
>> 
>> They were not being used.  Why would you want them in there?
>
>Because they contained useful information which had been accumulated by
>many people over a long period of time.
>
>Throwing that information away seemed rather pointless, especially as the
>cost of retaining it was so low.

There's an out-of-tree reference, the pci.ids website, that carries 
this information, do we need the reference info in the kernel as well?  

So far two people raised an objection, the other wants to maintain 
an out-of-tree driver, D. Fox didn't say why he needs the symbols.

There's some other cleanups to be done yet, as >100 files define 
own PCI_* symbols instead of referencing the header...


Would you prefer to revert the patches or rather me put the removed 
symbols back again via a new patch series?

Cheers,
Grant.

Re: 2.6.15-rc2 pci_ids.h cleanup is a pain

[Andrew Morton]

Grant Coady <grant_lkml@dodo.com.au> wrote:
>
> Would you prefer to revert the patches or rather me put the removed 
>  symbols back again via a new patch series?

It's probably too late now - I'm just having a little grump.

Re: 2.6.15-rc2 pci_ids.h cleanup is a pain

[Dave Jones]

On Wed, Nov 23, 2005 at 12:51:45PM +1100, Grant Coady wrote:
 > Hi Andrew, Greg,
 > 
 > On Tue, 22 Nov 2005 16:25:58 -0800, Andrew Morton <akpm@osdl.org> wrote:
 > 
 > >Greg KH <greg@kroah.com> wrote:
 > >>
 > >> On Mon, Nov 21, 2005 at 02:27:23PM -0800, David Fox wrote:
 > >> > I'm sure I'm not the only person that applies patches to the kernel that 
 > >> > use some of the 500 plus PCI IDS eliminated from pci_ids.h by rc2.  I 
 > >> > would like to see the PCI ids that were removed simply because the don't 
 > >> > occur in the main kernel source restored.  Is there a rationale for 
 > >> > removing them that I'm not aware of?
 > >> 
 > >> They were not being used.  Why would you want them in there?
 > >
 > >Because they contained useful information which had been accumulated by
 > >many people over a long period of time.
 > >
 > >Throwing that information away seemed rather pointless, especially as the
 > >cost of retaining it was so low.
 > 
 > There's an out-of-tree reference, the pci.ids website, that carries 
 > this information, do we need the reference info in the kernel as well?  
 > 
 > So far two people raised an objection, the other wants to maintain 
 > an out-of-tree driver, D. Fox didn't say why he needs the symbols.

Three. I already mentioned we broke the compilation of the
advansys driver because of this.

		Dave

Re: 2.6.15-rc2 pci_ids.h cleanup is a pain

[Grant Coady]

On Tue, 22 Nov 2005 23:19:17 -0500, Dave Jones <davej@redhat.com> wrote:

>Three. I already mentioned we broke the compilation of the
>advansys driver because of this.

Nah, two.  Dave J. is the other, broken BROKEN driver, out of tree 
for working driver, patches not in mainline?  

I may have misunderstood...  You want advansys back?  Perhaps my 
boo-boo for not checking with allyesconfig + BROKEN re: unused 
symbols?

Grant.

Re: 2.6.15-rc2 pci_ids.h cleanup is a pain

[Grant Coady]

On Tue, 22 Nov 2005 23:19:17 -0500, Dave Jones <davej@redhat.com> wrote:

>On Wed, Nov 23, 2005 at 12:51:45PM +1100, Grant Coady wrote:
> > Hi Andrew, Greg,
> > 
> > On Tue, 22 Nov 2005 16:25:58 -0800, Andrew Morton <akpm@osdl.org> wrote:
> > 
> > >Greg KH <greg@kroah.com> wrote:
> > >>
> > >> On Mon, Nov 21, 2005 at 02:27:23PM -0800, David Fox wrote:
> > >> > I'm sure I'm not the only person that applies patches to the kernel that 
> > >> > use some of the 500 plus PCI IDS eliminated from pci_ids.h by rc2.  I 
> > >> > would like to see the PCI ids that were removed simply because the don't 
> > >> > occur in the main kernel source restored.  Is there a rationale for 
> > >> > removing them that I'm not aware of?
> > >> 
> > >> They were not being used.  Why would you want them in there?
> > >
> > >Because they contained useful information which had been accumulated by
> > >many people over a long period of time.
> > >
> > >Throwing that information away seemed rather pointless, especially as the
> > >cost of retaining it was so low.
> > 
> > There's an out-of-tree reference, the pci.ids website, that carries 
> > this information, do we need the reference info in the kernel as well?  
> > 
> > So far two people raised an objection, the other wants to maintain 
> > an out-of-tree driver, D. Fox didn't say why he needs the symbols.
>
>Three. I already mentioned we broke the compilation of the
>advansys driver because of this.

Nope, advansys.* don't appear to use PCI_*  Defines its own ASC_PCI* 
instead?

Checked:  linux-2.6.13.4, linux-2.6.15-rc2

Grant.

Re: 2.6.15-rc2 pci_ids.h cleanup is a pain

[Dave Jones]

On Thu, Nov 24, 2005 at 08:41:12AM +1100, Grant Coady wrote:

 > >Three. I already mentioned we broke the compilation of the
 > >advansys driver because of this.
 > 
 > Nope, advansys.* don't appear to use PCI_*  Defines its own ASC_PCI* 
 > instead?

I stand corrected.

		Dave