Re: [PATCH][2.6.6-rc3] gcc-3.4.0 fixes

["H. Peter Anvin" <hpa@zytor.com>]

Linus Torvalds wrote:
> 
> I don't see any point where we cast any function pointers to anything 
> else.
> 
> We cast data pointers all over the place, but that is actually guaranteed
> to work in C for some "large enough" integer type, and "unsigned long" is
> pretty much it. 
> 

It would be nice at some point to switch that to [u]intptr_t, before someone 
comes up with 128-bit machines (in other words, no rush whatsoever, but after 
seeing Sony build processors with 128-bit integer registers I'm willing to 
believe it's just a matter of time...)  The other thing about it is that it's 
nice to be explicit about the "pointer-sized integerness" of it all.

> And even function pointers should be safeish. The fact that some broken
> architecture (can you say "ia64"?) has totally idiotic calling conventions
> and requires the caller to load the GP value is _their_ problem. The
> architecture will either die or hide the fact that it's being silly. For
> now it's hiding it.
> 
> Repeat after me: practice is more important than theory. A _lot_ more 
> important.

Indeed.

	-hpa