Re: PATCH:[darwin] fix load of a misaligned double word

[Andrew Haley <aph at redhat dot com>]

Brad Lucier writes:
 > > On Dec 28, 2003, at 21:26, Brad Lucier wrote:
 > > 
 > > >> Basically it is these two lines which are causing them:
 > > >>
 > > >> *(double*)(((int)(((int*)((___r1)-(1)))+1))+(((((int)(8))<<2))<<(3-2)))=(2*(1+2)<<2);
 > > >>       n = (1<<2) + 17 * sizeof (double) + (5<<2);
 > > >>
 > > >> *(double*)(((int)(((int*)((___r1)-(1)))+1))+(((((int)(9))<<2))<<(3-2)))=(2*n);
 > > 
 > > Also if you did not violate ANSI C aliasing rules this would not really 
 > > be a problem.
 > 
 > I'be been thinking about this.  The memory location is accessed
 > only as a double.  There is no violation of the aliasing rules that
 > I can see.

Regardless of whether this is an aliasing violation, constructing a
misaligned pointer by use of pointer casts and then using that pointer
is not well-defined.  (C99 Section 6.2.2.3, Pointers.)  Even if someone
changes the back end, there's no guarantee that gcc will do anything
useful with this code.

Andrew.