On Wed, Oct 31, 2001 at 11:22:20AM +0100, Roman Zippel wrote:

Hi,

On Tue, 30 Oct 2001, Tom Rini wrote:

quoted

quoted

- pcibios_make_OF_bus_map depends on CONFIG_ALL_PPC

This should just be a warning, since gcc will optmize it out (it'll
always be false on !CONFIG_ALL_PPC.  It'd almost be preferable to do
something like:
#define pcibios_make_OF_bus_map do { } while(0)

Someplace...

That function is in the same file...

Yes, and compiled only on CONFIG_ALL_PPC.  But the if statement will
always be false on !CONFIG_ALL_PPC, so while the compiler will warn
about an implicit declaration, it won't cause a link error.  Eg:

--- 1.34/arch/ppc/kernel/pci.c  Sat Oct  6 11:16:41 2001
+++ edited/pci.c        Wed Oct 31 08:40:26 2001

@@ -743,6 +743,9 @@
                ranges += np;
        }
 }
+#else
+/* Kill a warning */
+#define pcibios_make_OF_bus_map do { } while(0)
 #endif /* CONFIG_ALL_PPC */

 void __init

quoted

quoted

- removed duplicate definition of isa_io_base/isa_mem_base/pci_dram_offset
(note: these variables are protected with CONFIG_PCI in ppc_ksyms.c, but
not in the include files, so there are unconditional now)

This will break 4xx I think...  Can you try doing a
walnut+CONFIG_405_DMA=y compile (in _devel..)

I can try it (when I get home).

Thanks.

quoted

quoted

- removed nested __KERNEL__ ifdefs in io.h, one even outside of the
recursive protection (bad style).

Correct, but it's usually
#ifdef __KERNEL__
#ifndef __FILE_H__
#define __FILE_H__
...
#endif
#endif

Or so.

I would do it the other way around. gcc looks for a ifndef/define/endif
sequence to decide whether it needs to rescan that file next time, but I'm
not sure if gcc finds it in this case.

By and large we do it the way I said anyhow.  Is this in current gcc or
when we get the precompiled headers bits?

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/