alignment test case
Wed Mar 19 14:45:00 GMT 2003
Jeff Sturm writes:
> On Wed, 19 Mar 2003, Andrew Haley wrote:
> > > I can't see any way to fix this without fixing the C++ problem. Well,
> > > we could rewrite gcj to use the new ABI, but that's a long way down
> > > the road.
> > Actually there might be another way: use __attribute__((aligned)) on
> > java.lang.Object. Then a subclass can never be more aligned than its
> > superclass. Wastes memory, though.
> I thought that with hash synchronization, boehm-gc always allocates on an
> 8-byte boundary anyway? Would aligning Object really waste memory?
A little, because every subclass would also be 8-aligned.
But I don't know that would fix the bug anyway. I need to understand
what the C++ compiler is doing.
More information about the Java