> Linus' vanilla 2.3.30 -- that vanilla one doesn't compile
> either..)
because Linus didn't applyed my patches and he merged a patch (not from
me) around 2.3.30pre6 that forgot to remove some of the `kill_arch:'
fields. It's basically only a trivially fixable compilation problem.
So just go with Linus's tree plus my alpha-1.gz and alpha-1-to-2.gz.
Everything after `2' is very unstable. The `4' revision happens to be rock
solid in UP on tsunami though, so it looks at least a bit promising ;).
If you all agree I'll send to Linus an alpha-1.gz plus alpha-1-to-2.gz
merged in one whole patch (so there won't be any pci common code change).
The only thing that Linus can be interested about is the new
____cacheline_aligned macro. I think it's less misleading to call it
__gcc_cacheline_aligned so there will be less confusion between the elf
section trick and the compiler way. I am also not 100% sure if
__gcc_cacheline_aligned works fine with gcc-2.7.2.3 or not. Or maybe we
could implement the __cacheline_aligned in the compiler way as Linus's
tree is doing... (potentially wasting some memory).
Comments?
Andrea