[PATCH] malloc: Cleanup libc_realloc

[Wilco Dijkstra]

Hi DJ,

> Why is this unneeded?  Growing an mmap'd chunk into the rest of its page
> seems a useful optimization here.

See mremap_chunk():

  /* No need to remap if the number of pages does not change.  */
  if (total_size == new_size)
    return p;

And if that fails, we do:

      /* Note the extra SIZE_SZ overhead. */
      if (oldsize - SIZE_SZ >= nb)
        return oldmem;   

So it's repeating the same thing 3 times in slightly different ways... We likely
want to merge these 2 checks as well. And perhaps we could use madvise if
mremap fails. And we might want to do this not for a single page in large
allocations but use some slack.

(and yes, I believe the same would be good for small blocks too - obviously
both for growing and shrinking blocks given that software often calls realloc
too frequently for no good reason).

Cheers,
Wilco