[Bug inline-asm/10153] [3.3/3.4 regression] selection of %dil or %sil on ia32 by valid C source

[rth@gcc.gnu.org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=10153


rth@gcc.gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID


------- Additional Comments From rth@gcc.gnu.org  2003-06-12 18:18 -------
I'm of the opinion that the source is invalid.  Using a "=q" constraint
instead of an "=r" constraint properly shows that the compiler must choose
one of the four "low byte" registers.

You might object that the compiler should "know" that it cannot place a
character object into %esi/%edi/%ebp, but that's not true at all -- we 
cannot store a character to memory from those registers, but we can hold
a character value in those registers.