Re: Help with understanding strict aliasing rules

[Andrew Haley <aph at redhat dot com>]

Martin Lucina writes:
 > aph@redhat.com said:
 > > Please read the actual standard too.  The rules about pointer type
 > > conversions are at 6.3.2.3.  The appropriate paragraphs are paragraphs
 > > 1 and 7.
 > 
 > So according to paragraph 1, the following is legal (obviously, this is
 > basic C):
 > 
 > void *return_item (int *rc);
 > 
 > ...
 > item_t *item_p;
 > int rc;
 > 
 > item_p = return_item (&rc);
 > 
 > but the case 
 > 
 > int return_item (void **item_p);
 > 
 > ...
 > item_t *item_p;
 > int rc;
 > 
 > rc = return_item ((void **)&item_p);
 > 
 > is not legal since (void **) doesn't fall under the specification of
 > paragraph 1, but will work fine as long as item_p is only ever
 > dereferenced as an item_t?

Well, the phrases "is not legal" and "will work fine" are clearly
contradictory!  It can't be both.

 > In other words, (void *) is convertible to any incomplete or object
 > type, but (void **) is not, despite the fact that *(void **) is a (void *)?

That's right.

 > > (unless you cast it back to its original type first).  This warning
 > > should be interpreted as saying that your interfaces are badly designed,
 > > and the correct way to avoid the warning is to redesign them in a way
 > > where you do not need to cast between conflicting types.  (Even if you
 > 
 > Unfortunately I can't change the interfaces since they're used by
 > legacy code.  I'd just like to get rid of the warnings since if
 > people get used to ignoring the warnings in this code then they are
 > likely to miss other more important problems.

-fno-strict-aliasing will do that, but will disable some
optimizations.  Your call.

Andrew.