[MLton] osx compiles again in x86_64 branch

[Wesley W. Terpstra]

On May 24, 2006, at 12:44 AM, Wesley W. Terpstra wrote:
> Things start going wrong after an attempt to remapHeap fails  
> (because on osx HAS_REMAP is FALSE).

void *GC_mremap (void *start, size_t oldLength, size_t newLength) {
         void *out;
         size_t smallest;

         if (oldLength < newLength)
                 smallest = oldLength;
         else    smallest = newLength;

         out = mmap (0, newLength, PROT_READ | PROT_WRITE,  
MAP_PRIVATE | MAP_ANON, -1, 0);
         memcpy (out, start, smallest);
         munmap (start, oldLength);

         return out;
}

ottawa:~/src/mlton/mlton.x86_64/mllex terpstra$ ./mllex
no files
Usage: mllex file.lex ...
ottawa:~/src/mlton/mlton.x86_64/mllex terpstra$

Tada!

Anyways, it's definitely the code-path which works around a missing  
mremap, as I hope I've just demonstrated. So set HAS_REMAP FALSE on  
linux, and I bet you'll get exactly the same problem. For whatever  
reason, the stack is pointing into the region which gets freed by the  
first use in growHeap. I imagine that either you're freeing the wrong  
heap, or you didn't update the stack pointer to the new heap after  
the move.

(It later crashes when trying to access the address which got freed)