On Fri, 2005-01-14 at 11:23 +0000, Ciaran McCreesh wrote: > On Fri, 14 Jan 2005 12:10:31 +0100 Paul de Vrieze <[EMAIL PROTECTED]> > wrote: > | These ebuilds are broken. Instead of "use sse" they should use "use > | x86 && use sse". All workarounds are broken by proxy. > > Nonsense. The 'sse' USE flag is used to indicate the availability of > optional x86 sse assembly code. If an arch is incapable of running x86 > sse assembly code then it should use.mask the sse flag.
I don't see anything about the flag being x86-only in use.desc, at all. Especially since the instruction set is available on a non-x86 CPU (kinda) > What next? (use x86 || use x86-obsd || use x86-fbsd || use x86-od) && > use x86? You're not thinking of the big picture here. A single purpose Actually, that would be much more correct, in this instance. Especially considering that amd64 *is* a supported platform for the SSE/SSE2 instruction sets. When we have code (non-sse) that does something stupid and arch-specific, we either fix the code to be arch-neutral, or only allow it on the architecture where it works. How is this any different? Because it is assembly? Because it is x86? > USE flag which is use.mask'ed on non-supporting profiles is far cleaner. Yes, as it should be on sparc, alpha, mips, etc. but *not* on amd64 because it *is* supported. > Now, you could argue that x86-sse would be a better name for the flag, > but since we can't really do USE renames it's kinda moot. If that were the intention of the flag, then yes, that would have been a better name. -- Chris Gianelloni Release Engineering - Operational/QA Manager Games - Developer Gentoo Linux
signature.asc
Description: This is a digitally signed message part
