[MLton] gc question

[Matthew Fluet]

> Is this code (in foreachObjptrInObject) correct?
> 
>   else if (WEAK_TAG == tag) {
>     p += bytesNonObjptrs;
>     if (1 == numObjptrs) {
>       if (not skipWeaks)
>         callIfIsObjptr (s, f, (objptr*)p);
>       p += OBJPTR_SIZE;
>     }
>   }

Yes.

> We only advance p past the weak pointer if it is valid, but this does
> not seem to square with the description of foreachObjptrInObject:
> 
>   Returns pointer to the end of object, i.e. just past object.
> 
> or with the way it is called in foreachObjptrInRange. Am I missing
> something?

See runtime/gc/weak.h.

The object type indexed by the header of the weak pointer object 
determines whether the weak is valid or not.  When a weak goes dead, we 
update its header.

If the type has numObjptrs == 1, then the weak pointer is valid and 
bytesNonObjptrs == 4 (or 8 for -align 8).  Hence, in this case the code 
above advances by 4 and then again by 4.

If the type has numObjptrs == 0, then the weak pointer is invalid and 
bytesNonObjptrs == 8 (or 12 for -align 8).  Hence, in this case, the 
code above advances by 8.

We do things this way so that in many situations, we can treat a weak 
object just like a normal object, using the bytesNonObjptrs and numObjptrs.