Patch: Standard comformant arithmetic

[Tom Tromey]

>>>>> "Andrew" == Andrew Haley <aph@pasanda.cygnus.co.uk> writes:

>> For systems without memory management support, when -Os is used, it
>> might make the most sense to put the check inside the callee and
>> not the caller.  Of course, this would mean that you'd have to
>> build your entire application the same way in order to get
>> consistent null pointer checks.  Does this sound reasonable?

Andrew> What exactly do you mean by "the callee and not the caller?"  I can
Andrew> think of several different ways that the check could be done.

Instead of checking at each call site whether this==null, we would put
code to do the check just after the prologue of each function.  This
would make a smaller executable at the cost of speed (since presumably
the provably-not-null optimization would win in many cases).

This doesn't suffice for field accesses.  But field accesses from
other objects ought to be rare compared to method calls.

Tom