when CONFIG_PCI is not defined

[Jeff Garzik]

Martin,

Please do not copy private mail from me to a public list.  Very bad
netiquette.


Martin Mares wrote:
> Jeff Garzik wrote:
> > Would you mind reviewing pci.h to make sure that all necessary functions
> > are exported when CONFIG_PCI is not defined?

> > I think that some of the recent PCI additions were not available for the
> > !CONFIG_PCI case.  This is wrong IMHO because pci.h is specifically
> > designed to provide support to drivers even when CONFIG_PCI is
> > undefined.  It leads to much cleaner driver code.

>    I'm not sure at all that this is the right way to go.
[...]
> in which case nothing gets optimized away when compiling without CONFIG_PCI
> and you end up with the driver full of dead code. Therefore to be on the
> clean side of the glass, you need #ifdef CONFIG_PCI anyway, so there isn't
> much sense in providing pci_register_driver() stub for PCI-less cases.

Well, if that is your and Linus' feeling, then you should rip out all
the inlined compatibility code.  As it stands now, there is a random
collection of functions which work correctly for the !CONFIG_PCI case,
and a random collection of functions which break if CONFIG_PCI is not
defined.  Bug city.

I fully realize that inlined functions with the new API would create a
SMALL bit of dead code.  If you look at driver usage of the old APIs,
this condition has existed for a long time.  Dead code with !CONFIG_PCI
was always a possibility.

Linus seems to favor _removing_ #ifdef trees from the main source code. 
Your proposal adds #ifdefs to code, making driver code less clean (if
0.00001% more efficient for the !CONFIG_PCI case).

Regards,

	Jeff



+
-- 
Jeff Garzik         | Andre the Giant has a posse.
Building 1024       |
MandrakeSoft, Inc.  |

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo en vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/