Implement support for non-coalescing maps on MinGW.

- Do not reallocate huge objects in place if the number of backing
  chunks would change.
- Do not cache multi-chunk mappings.

This resolves #213.

src/chunk.c

@@ -337,6 +337,7 @@ chunk_record(arena_t *arena, extent_tree_t *chunks_szad,
 	extent_node_t *node, *prev;
 	extent_node_t key;
 
+	assert(maps_coalesce || size == chunksize);
 	assert(!cache || !zeroed);
 	unzeroed = cache || !zeroed;
 	JEMALLOC_VALGRIND_MAKE_MEM_NOACCESS(chunk, size);
@@ -421,6 +422,11 @@ chunk_dalloc_cache(arena_t *arena, void *chunk, size_t size)
 	assert(size != 0);
 	assert((size & chunksize_mask) == 0);
 
+	if (!maps_coalesce && size != chunksize) {
+		chunk_dalloc_arena(arena, chunk, size, false);
+		return;
+	}
+
 	chunk_record(arena, &arena->chunks_szad_cache, &arena->chunks_ad_cache,
 	    true, chunk, size, false);
 	arena_maybe_purge(arena);

src/huge.c

@@ -304,6 +304,9 @@ huge_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra,
 		return (false);
 	}
 
+	if (!maps_coalesce)
+		return (true);
+
 	/* Shrink the allocation in-place. */
 	if (CHUNK_CEILING(oldsize) >= CHUNK_CEILING(usize)) {
 		huge_ralloc_no_move_shrink(ptr, oldsize, usize);