RE: [PATCH] fixes for building kernel 2.5.45 using Intel compiler

[Nakajima, Jun]

> From: Nakajima, Jun [mailto:[e-mail]]
> Sent: Thursday, October 31, 2002 12:17 PM
> 
> This is take 2 of the updated patch against 2.5.45. I'm 
> asking the compiler  team if someone can answer your question:
> 
> > Considering that Intel largely wrote iBCS2, I guess some 
> Intel person can 
> > know what the standard was ;)

This is what some Intel person said:
   The optimization the compiler is working around is perfectly legal.
   It isn't legal C/C++ to "use" a parameter after the function has
returned.
   The optimizer knows this, and that is why the code got removed. The
volatile 
   is necessary in order to make this kind of C code work.

The point is that once sys_iopl(unsigned long unused) returns a value,
unused is 
finished. So nobody can do anything with the data associated with it after
that.

Jun