diff --git a/include/mimalloc.h b/include/mimalloc.h
index 38524b8d..0b613c74 100644
--- a/include/mimalloc.h
+++ b/include/mimalloc.h
@@ -401,6 +401,9 @@ mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_new_n(size_t count, s
 mi_decl_nodiscard mi_decl_export void* mi_new_realloc(void* p, size_t newsize)                mi_attr_alloc_size(2);
 mi_decl_nodiscard mi_decl_export void* mi_new_reallocn(void* p, size_t newcount, size_t size) mi_attr_alloc_size2(2, 3);
 
+mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_alloc_new(mi_heap_t* heap, size_t size)                mi_attr_malloc mi_attr_alloc_size(2);
+mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_alloc_new_n(mi_heap_t* heap, size_t count, size_t size) mi_attr_malloc mi_attr_alloc_size2(2, 3);
+
 #ifdef __cplusplus
 }
 #endif
@@ -460,6 +463,124 @@ template<class T> struct mi_stl_allocator {
 
 template<class T1,class T2> bool operator==(const mi_stl_allocator<T1>& , const mi_stl_allocator<T2>& ) mi_attr_noexcept { return true; }
 template<class T1,class T2> bool operator!=(const mi_stl_allocator<T1>& , const mi_stl_allocator<T2>& ) mi_attr_noexcept { return false; }
+
+#if (__cplusplus >= 201103L) || (_MSC_VER > 1900)  // C++11
+#include <memory>      // std::shared_ptr
+
+// STL allocator allocation in a specific heap
+template<class T> struct mi_heap_stl_allocator {
+  typedef T                 value_type;
+  typedef std::size_t       size_type;
+  typedef std::ptrdiff_t    difference_type;
+  typedef value_type& reference;
+  typedef value_type const& const_reference;
+  typedef value_type* pointer;
+  typedef value_type const* const_pointer;
+  template <class U> struct rebind { typedef mi_heap_stl_allocator<U> other; };
+
+  mi_heap_stl_allocator() { 
+    mi_heap_t* hp = mi_heap_new();
+    this->heap.reset(hp, heap_delete);
+  }
+  mi_heap_stl_allocator(mi_heap_t* hp) : heap(hp) { }    /* will not delete or destroy the passed in heap */
+  mi_heap_stl_allocator(const mi_heap_stl_allocator& x) mi_attr_noexcept : heap(x.heap) { }
+  template<class U> mi_heap_stl_allocator(const mi_heap_stl_allocator<U>& x) mi_attr_noexcept : heap(x.heap) { }  
+  mi_heap_stl_allocator  select_on_container_copy_construction() const { return *this; }
+  void deallocate(T* p, size_type) { mi_free(p); }
+
+  #if (__cplusplus >= 201703L)  // C++17
+  mi_decl_nodiscard T* allocate(size_type count) { return static_cast<T*>(mi_heap_alloc_new_n(this->heap.get(), count, sizeof(T))); }
+  mi_decl_nodiscard T* allocate(size_type count, const void*) { return allocate(count); }
+  #else
+  mi_decl_nodiscard pointer allocate(size_type count, const void* = 0) { return static_cast<pointer>(mi_heap_alloc_new_n(this->heap.get(), count, sizeof(value_type))); }
+  #endif
+
+  #if ((__cplusplus >= 201103L) || (_MSC_VER > 1900))  // C++11
+  using propagate_on_container_copy_assignment = std::true_type;
+  using propagate_on_container_move_assignment = std::true_type;
+  using propagate_on_container_swap = std::true_type;
+  using is_always_equal = std::false_type;
+  template <class U, class ...Args> void construct(U* p, Args&& ...args) { ::new(p) U(std::forward<Args>(args)...); }
+  template <class U> void destroy(U* p) mi_attr_noexcept { p->~U(); }
+  #else
+  void construct(pointer p, value_type const& val) { ::new(p) value_type(val); }
+  void destroy(pointer p) { p->~value_type(); }
+  #endif
+
+  size_type     max_size() const mi_attr_noexcept { return (PTRDIFF_MAX / sizeof(value_type)); }
+  pointer       address(reference x) const { return &x; }
+  const_pointer address(const_reference x) const { return &x; }
+
+  void collect(bool force) { mi_heap_collect(this->heap.get(), force); }
+
+// protected:
+  std::shared_ptr<mi_heap_t> heap;
+
+private:
+  static void heap_delete(mi_heap_t* hp)  { if (hp != NULL) { mi_heap_delete(hp); } }
+};
+
+template<class T1, class T2> bool operator==(const mi_heap_stl_allocator<T1>& x, const mi_heap_stl_allocator<T2>& y) mi_attr_noexcept { return (x.heap == y.heap); }
+template<class T1, class T2> bool operator!=(const mi_heap_stl_allocator<T1>& x, const mi_heap_stl_allocator<T2>& y) mi_attr_noexcept { return (x.heap != y.heap); }
+
+
+// STL allocator allocation in a specific heap, where `free` does nothing and 
+// the heap is destroyed in one go on destruction -- use with care!
+template<class T> struct mi_heap_destroy_stl_allocator {
+  typedef T                 value_type;
+  typedef std::size_t       size_type;
+  typedef std::ptrdiff_t    difference_type;
+  typedef value_type& reference;
+  typedef value_type const& const_reference;
+  typedef value_type* pointer;
+  typedef value_type const* const_pointer;
+  template <class U> struct rebind { typedef mi_heap_destroy_stl_allocator<U> other; };
+
+  mi_heap_destroy_stl_allocator() {
+    mi_heap_t* hp = mi_heap_new();
+    this->heap.reset(hp, heap_destroy);
+  }
+  mi_heap_destroy_stl_allocator(mi_heap_t* hp) : heap(hp) { }    /* will not delete or destroy the passed-in heap; nor free any allocated objects it allocates in the heap! */
+  mi_heap_destroy_stl_allocator(const mi_heap_destroy_stl_allocator& x) mi_attr_noexcept : heap(x.heap) { }
+  template<class U> mi_heap_destroy_stl_allocator(const mi_heap_destroy_stl_allocator<U>& x) mi_attr_noexcept : heap(x.heap) { }
+  mi_heap_destroy_stl_allocator  select_on_container_copy_construction() const { return *this; }
+  void deallocate(T* p, size_type) { /* do nothing as we destroy the heap on destruct. */ }
+
+  #if (__cplusplus >= 201703L)  // C++17
+  mi_decl_nodiscard T* allocate(size_type count) { return static_cast<T*>(mi_heap_alloc_new_n(this->heap.get(), count, sizeof(T))); }
+  mi_decl_nodiscard T* allocate(size_type count, const void*) { return allocate(count); }
+  #else
+  mi_decl_nodiscard pointer allocate(size_type count, const void* = 0) { return static_cast<pointer>(mi_heap_alloc_new_n(this->heap.get(), count, sizeof(value_type))); }
+  #endif
+
+  #if ((__cplusplus >= 201103L) || (_MSC_VER > 1900))  // C++11
+  using propagate_on_container_copy_assignment = std::true_type;
+  using propagate_on_container_move_assignment = std::true_type;
+  using propagate_on_container_swap = std::true_type;
+  using is_always_equal = std::false_type;
+  template <class U, class ...Args> void construct(U* p, Args&& ...args) { ::new(p) U(std::forward<Args>(args)...); }
+  template <class U> void destroy(U* p) mi_attr_noexcept { p->~U(); }
+  #else
+  void construct(pointer p, value_type const& val) { ::new(p) value_type(val); }
+  void destroy(pointer p) { p->~value_type(); }
+  #endif
+
+  size_type     max_size() const mi_attr_noexcept { return (PTRDIFF_MAX / sizeof(value_type)); }
+  pointer       address(reference x) const { return &x; }
+  const_pointer address(const_reference x) const { return &x; }
+
+// protected:
+  std::shared_ptr<mi_heap_t> heap;
+
+private:
+  static void heap_destroy(mi_heap_t* hp) { if (hp != NULL) { mi_heap_destroy(hp); } }
+};
+
+template<class T1, class T2> bool operator==(const mi_heap_destroy_stl_allocator<T1>& x, const mi_heap_destroy_stl_allocator<T2>& y) mi_attr_noexcept { return (x.heap == y.heap); }
+template<class T1, class T2> bool operator!=(const mi_heap_destroy_stl_allocator<T1>& x, const mi_heap_destroy_stl_allocator<T2>& y) mi_attr_noexcept { return (x.heap != y.heap); }
+
+#endif // C++11
+
 #endif // __cplusplus
 
 #endif
diff --git a/readme.md b/readme.md
index fe2ead69..64cd0c4b 100644
--- a/readme.md
+++ b/readme.md
@@ -27,6 +27,8 @@ It also has an easy way to override the default allocator in [Windows](#override
   to integrate and adapt in other projects. For runtime systems it
   provides hooks for a monotonic _heartbeat_ and deferred freeing (for
   bounded worst-case times with reference counting).
+  Partly due to its simplicity, mimalloc has been ported to many systems (Windows, macOS, 
+  Linux, WASM, various BSD's, Haiku, MUSL, etc) and has excellent support for dynamic overriding.
 - __free list sharding__: instead of one big free list (per size class) we have
   many smaller lists per "mimalloc page" which reduces fragmentation and
   increases locality --
@@ -42,7 +44,7 @@ It also has an easy way to override the default allocator in [Windows](#override
   similar to randomized algorithms like skip lists where adding
   a random oracle removes the need for a more complex algorithm.
 - __eager page reset__: when a "page" becomes empty (with increased chance
-  due to free list sharding) the memory is marked to the OS as unused ("reset" or "purged")
+  due to free list sharding) the memory is marked to the OS as unused (reset or decommitted)
   reducing (real) memory pressure and fragmentation, especially in long running
   programs.
 - __secure__: _mimalloc_ can be built in secure mode, adding guard pages,
@@ -52,13 +54,12 @@ It also has an easy way to override the default allocator in [Windows](#override
 - __first-class heaps__: efficiently create and use multiple heaps to allocate across different regions.
   A heap can be destroyed at once instead of deallocating each object separately.  
 - __bounded__: it does not suffer from _blowup_ \[1\], has bounded worst-case allocation
-  times (_wcat_), bounded space overhead (~0.2% meta-data, with low internal fragmentation),
-  and has no internal points of contention using only atomic operations.
+  times (_wcat_) (upto OS primitives), bounded space overhead (~0.2% meta-data, with low 
+  internal fragmentation), and has no internal points of contention using only atomic operations.
 - __fast__: In our benchmarks (see [below](#performance)),
   _mimalloc_ outperforms other leading allocators (_jemalloc_, _tcmalloc_, _Hoard_, etc),
-  and often uses less memory. A nice property
-  is that it does consistently well over a wide range of benchmarks. There is also good huge OS page
-  support for larger server programs.
+  and often uses less memory. A nice property is that it does consistently well over a wide range 
+  of benchmarks. There is also good huge OS page support for larger server programs.
 
 The [documentation](https://microsoft.github.io/mimalloc) gives a full overview of the API.
 You can read more on the design of _mimalloc_ in the [technical report](https://www.microsoft.com/en-us/research/publication/mimalloc-free-list-sharding-in-action) which also has detailed benchmark results.   
diff --git a/src/alloc.c b/src/alloc.c
index 6af43b50..d42d4c1a 100644
--- a/src/alloc.c
+++ b/src/alloc.c
@@ -91,7 +91,10 @@ extern inline void* _mi_page_malloc(mi_heap_t* heap, mi_page_t* page, size_t siz
 
 static inline mi_decl_restrict void* mi_heap_malloc_small_zero(mi_heap_t* heap, size_t size, bool zero) mi_attr_noexcept {
   mi_assert(heap != NULL);
-  mi_assert(heap->thread_id == 0 || heap->thread_id == _mi_thread_id()); // heaps are thread local
+  #if MI_DEBUG 
+  const uintptr_t tid = _mi_thread_id();
+  mi_assert(heap->thread_id == 0 || heap->thread_id == tid); // heaps are thread local
+  #endif
   mi_assert(size <= MI_SMALL_SIZE_MAX);
 #if (MI_PADDING)
   if (size == 0) {
@@ -916,20 +919,46 @@ static bool mi_try_new_handler(bool nothrow) {
 }
 #endif
 
-static mi_decl_noinline void* mi_try_new(size_t size, bool nothrow ) {
+static mi_decl_noinline void* mi_heap_try_new(mi_heap_t* heap, size_t size, bool nothrow ) {
   void* p = NULL;
   while(p == NULL && mi_try_new_handler(nothrow)) {
-    p = mi_malloc(size);
+    p = mi_heap_malloc(heap,size);
   }
   return p;
 }
 
-mi_decl_nodiscard mi_decl_restrict void* mi_new(size_t size) {
-  void* p = mi_malloc(size);
-  if mi_unlikely(p == NULL) return mi_try_new(size,false);
+static mi_decl_noinline void* mi_try_new(size_t size, bool nothrow) {
+  return mi_heap_try_new(mi_get_default_heap(), size, nothrow);
+}
+
+
+mi_decl_nodiscard mi_decl_restrict inline void* mi_heap_alloc_new(mi_heap_t* heap, size_t size) {
+  void* p = mi_heap_malloc(heap,size);
+  if mi_unlikely(p == NULL) return mi_heap_try_new(heap, size, false);
   return p;
 }
 
+mi_decl_nodiscard mi_decl_restrict void* mi_new(size_t size) {
+  return mi_heap_alloc_new(mi_get_default_heap(), size);
+}
+
+
+mi_decl_nodiscard mi_decl_restrict inline void* mi_heap_alloc_new_n(mi_heap_t* heap, size_t count, size_t size) {
+  size_t total;
+  if mi_unlikely(mi_count_size_overflow(count, size, &total)) {
+    mi_try_new_handler(false);  // on overflow we invoke the try_new_handler once to potentially throw std::bad_alloc
+    return NULL;
+  }
+  else {
+    return mi_heap_alloc_new(heap,total);
+  }
+}
+
+mi_decl_nodiscard mi_decl_restrict void* mi_new_n(size_t count, size_t size) {
+  return mi_heap_alloc_new_n(mi_get_default_heap(), size, count);
+}
+
+
 mi_decl_nodiscard mi_decl_restrict void* mi_new_nothrow(size_t size) mi_attr_noexcept {
   void* p = mi_malloc(size);
   if mi_unlikely(p == NULL) return mi_try_new(size, true);
@@ -954,17 +983,6 @@ mi_decl_nodiscard mi_decl_restrict void* mi_new_aligned_nothrow(size_t size, siz
   return p;
 }
 
-mi_decl_nodiscard mi_decl_restrict void* mi_new_n(size_t count, size_t size) {
-  size_t total;
-  if mi_unlikely(mi_count_size_overflow(count, size, &total)) {
-    mi_try_new_handler(false);  // on overflow we invoke the try_new_handler once to potentially throw std::bad_alloc
-    return NULL;
-  }
-  else {
-    return mi_new(total);
-  }
-}
-
 mi_decl_nodiscard void* mi_new_realloc(void* p, size_t newsize) {
   void* q;
   do {
diff --git a/src/segment.c b/src/segment.c
index 25e63904..1654d2fd 100644
--- a/src/segment.c
+++ b/src/segment.c
@@ -771,11 +771,81 @@ static mi_page_t* mi_segments_page_find_and_allocate(size_t slice_count, mi_aren
    Segment allocation
 ----------------------------------------------------------- */
 
+static mi_segment_t* mi_segment_os_alloc( size_t required, size_t page_alignment, bool eager_delay, mi_arena_id_t req_arena_id,
+                                          size_t* psegment_slices, size_t* ppre_size, size_t* pinfo_slices, 
+                                          mi_commit_mask_t* pcommit_mask, mi_commit_mask_t* pdecommit_mask,
+                                          bool* is_zero, bool* pcommit, mi_segments_tld_t* tld, mi_os_tld_t* os_tld)
+
+{
+  // Allocate the segment from the OS
+  bool mem_large = (!eager_delay && (MI_SECURE==0)); // only allow large OS pages once we are no longer lazy    
+  bool is_pinned = false;
+  size_t memid = 0;
+  size_t align_offset = 0;
+  size_t alignment = MI_SEGMENT_ALIGN;
+  
+  if (page_alignment > 0) {
+    // mi_assert_internal(huge_page != NULL);
+    mi_assert_internal(page_alignment >= MI_SEGMENT_ALIGN);
+    alignment = page_alignment;
+    const size_t info_size = (*pinfo_slices) * MI_SEGMENT_SLICE_SIZE;
+    align_offset = _mi_align_up( info_size, MI_SEGMENT_ALIGN );
+    const size_t extra = align_offset - info_size;
+    // recalculate due to potential guard pages
+    *psegment_slices = mi_segment_calculate_slices(required + extra, ppre_size, pinfo_slices);
+    //segment_size += _mi_align_up(align_offset - info_size, MI_SEGMENT_SLICE_SIZE);
+    //segment_slices = segment_size / MI_SEGMENT_SLICE_SIZE;
+  }
+  const size_t segment_size = (*psegment_slices) * MI_SEGMENT_SLICE_SIZE;
+  mi_segment_t* segment = NULL;
+
+  // get from cache?
+  if (page_alignment == 0) {
+    segment = (mi_segment_t*)_mi_segment_cache_pop(segment_size, pcommit_mask, pdecommit_mask, &mem_large, &is_pinned, is_zero, req_arena_id, &memid, os_tld);
+  }
+  
+  // get from OS
+  if (segment==NULL) {
+    segment = (mi_segment_t*)_mi_arena_alloc_aligned(segment_size, alignment, align_offset, pcommit, &mem_large, &is_pinned, is_zero, req_arena_id, &memid, os_tld);
+    if (segment == NULL) return NULL;  // failed to allocate
+    if (*pcommit) {
+      mi_commit_mask_create_full(pcommit_mask);
+    }
+    else {
+      mi_commit_mask_create_empty(pcommit_mask);
+    }
+  }    
+  mi_assert_internal(segment != NULL && (uintptr_t)segment % MI_SEGMENT_SIZE == 0);
+
+  const size_t commit_needed = _mi_divide_up((*pinfo_slices)*MI_SEGMENT_SLICE_SIZE, MI_COMMIT_SIZE);
+  mi_assert_internal(commit_needed>0);
+  mi_commit_mask_t commit_needed_mask;
+  mi_commit_mask_create(0, commit_needed, &commit_needed_mask);
+  if (!mi_commit_mask_all_set(pcommit_mask, &commit_needed_mask)) {
+    // at least commit the info slices
+    mi_assert_internal(commit_needed*MI_COMMIT_SIZE >= (*pinfo_slices)*MI_SEGMENT_SLICE_SIZE);
+    bool ok = _mi_os_commit(segment, commit_needed*MI_COMMIT_SIZE, is_zero, tld->stats);
+    if (!ok) return NULL; // failed to commit 
+    mi_commit_mask_set(pcommit_mask, &commit_needed_mask); 
+  }
+  mi_track_mem_undefined(segment,commit_needed);
+  segment->memid = memid;
+  segment->mem_is_pinned = is_pinned;
+  segment->mem_is_large = mem_large;
+  segment->mem_is_committed = mi_commit_mask_is_full(pcommit_mask);
+  segment->mem_alignment = alignment;
+  segment->mem_align_offset = align_offset;
+  mi_segments_track_size((long)(segment_size), tld);
+  _mi_segment_map_allocated_at(segment);
+  return segment;
+}
+
+
 // Allocate a segment from the OS aligned to `MI_SEGMENT_SIZE` .
-static mi_segment_t* mi_segment_init(mi_segment_t* segment, size_t required, size_t page_alignment, mi_arena_id_t req_arena_id, mi_segments_tld_t* tld, mi_os_tld_t* os_tld, mi_page_t** huge_page)
+static mi_segment_t* mi_segment_alloc(size_t required, size_t page_alignment, mi_arena_id_t req_arena_id, mi_segments_tld_t* tld, mi_os_tld_t* os_tld, mi_page_t** huge_page)
 {
   mi_assert_internal((required==0 && huge_page==NULL) || (required>0 && huge_page != NULL));
-  mi_assert_internal((segment==NULL) || (segment!=NULL && required==0));
+  
   // calculate needed sizes first
   size_t info_slices;
   size_t pre_size;
@@ -786,114 +856,42 @@ static mi_segment_t* mi_segment_init(mi_segment_t* segment, size_t required, siz
                             _mi_current_thread_count() > 1 &&       // do not delay for the first N threads
                             tld->count < (size_t)mi_option_get(mi_option_eager_commit_delay));
   const bool eager = !eager_delay && mi_option_is_enabled(mi_option_eager_commit);
-  bool commit = eager || (required > 0); 
-  
-  // Try to get from our cache first
-  bool is_zero = false;
-  const bool commit_info_still_good = (segment != NULL);
+  bool commit = eager || (required > 0);   
+  bool is_zero = false;  
+
   mi_commit_mask_t commit_mask;
   mi_commit_mask_t decommit_mask;
-  if (segment != NULL) {
-    commit_mask = segment->commit_mask;
-    decommit_mask = segment->decommit_mask;
-  }
-  else {
-    mi_commit_mask_create_empty(&commit_mask);
-    mi_commit_mask_create_empty(&decommit_mask);
-  }
-  if (segment==NULL) {
-    // Allocate the segment from the OS
-    bool mem_large = (!eager_delay && (MI_SECURE==0)); // only allow large OS pages once we are no longer lazy    
-    bool is_pinned = false;
-    size_t memid = 0;
-    size_t align_offset = 0;
-    size_t alignment = MI_SEGMENT_ALIGN;
-    
-    if (page_alignment > 0) {
-      mi_assert_internal(huge_page != NULL);
-      mi_assert_internal(page_alignment >= MI_SEGMENT_ALIGN);
-      alignment = page_alignment;
-      const size_t info_size = info_slices * MI_SEGMENT_SLICE_SIZE;
-      align_offset = _mi_align_up( info_size, MI_SEGMENT_ALIGN );
-      const size_t extra = align_offset - info_size;
-      // recalculate due to potential guard pages
-      segment_slices = mi_segment_calculate_slices(required + extra, &pre_size, &info_slices);
-      //segment_size += _mi_align_up(align_offset - info_size, MI_SEGMENT_SLICE_SIZE);
-      //segment_slices = segment_size / MI_SEGMENT_SLICE_SIZE;
-    }
-    const size_t segment_size = segment_slices * MI_SEGMENT_SLICE_SIZE;
+  mi_commit_mask_create_empty(&commit_mask);
+  mi_commit_mask_create_empty(&decommit_mask);
 
-    // get from cache
-    if (page_alignment == 0) {
-      segment = (mi_segment_t*)_mi_segment_cache_pop(segment_size, &commit_mask, &decommit_mask, &mem_large, &is_pinned, &is_zero, req_arena_id, &memid, os_tld);
-    }
-    
-    // get from OS
-    if (segment==NULL) {
-      segment = (mi_segment_t*)_mi_arena_alloc_aligned(segment_size, alignment, align_offset, &commit, &mem_large, &is_pinned, &is_zero, req_arena_id, &memid, os_tld);
-      if (segment == NULL) return NULL;  // failed to allocate
-      if (commit) {
-        mi_commit_mask_create_full(&commit_mask);
-      }
-      else {
-        mi_commit_mask_create_empty(&commit_mask);
-      }
-    }    
-    mi_assert_internal(segment != NULL && (uintptr_t)segment % MI_SEGMENT_SIZE == 0);
-
-    const size_t commit_needed = _mi_divide_up(info_slices*MI_SEGMENT_SLICE_SIZE, MI_COMMIT_SIZE);
-    mi_assert_internal(commit_needed>0);
-    mi_commit_mask_t commit_needed_mask;
-    mi_commit_mask_create(0, commit_needed, &commit_needed_mask);
-    if (!mi_commit_mask_all_set(&commit_mask, &commit_needed_mask)) {
-      // at least commit the info slices
-      mi_assert_internal(commit_needed*MI_COMMIT_SIZE >= info_slices*MI_SEGMENT_SLICE_SIZE);
-      bool ok = _mi_os_commit(segment, commit_needed*MI_COMMIT_SIZE, &is_zero, tld->stats);
-      if (!ok) return NULL; // failed to commit 
-      mi_commit_mask_set(&commit_mask, &commit_needed_mask); 
-    }
-    mi_track_mem_undefined(segment,commit_needed);
-    segment->memid = memid;
-    segment->mem_is_pinned = is_pinned;
-    segment->mem_is_large = mem_large;
-    segment->mem_is_committed = mi_commit_mask_is_full(&commit_mask);
-    segment->mem_alignment = alignment;
-    segment->mem_align_offset = align_offset;
-    mi_segments_track_size((long)(segment_size), tld);
-    _mi_segment_map_allocated_at(segment);
-  }
-
-  // zero the segment info? -- not always needed as it is zero initialized from the OS 
+  // Allocate the segment from the OS  
+  mi_segment_t* segment = mi_segment_os_alloc(required, page_alignment, eager_delay, req_arena_id, 
+                                              &segment_slices, &pre_size, &info_slices, &commit_mask, &decommit_mask, 
+                                              &is_zero, &commit, tld, os_tld);
+  if (segment == NULL) return NULL;
+  
+  // zero the segment info? -- not always needed as it may be zero initialized from the OS 
   mi_atomic_store_ptr_release(mi_segment_t, &segment->abandoned_next, NULL);  // tsan
   if (!is_zero) {
     ptrdiff_t ofs = offsetof(mi_segment_t, next);
     size_t    prefix = offsetof(mi_segment_t, slices) - ofs;
     memset((uint8_t*)segment+ofs, 0, prefix + sizeof(mi_slice_t)*(segment_slices+1));  // one more
   }
-
-  if (!commit_info_still_good) {
-    segment->commit_mask = commit_mask; // on lazy commit, the initial part is always committed
-    segment->allow_decommit = (mi_option_is_enabled(mi_option_allow_decommit) && !segment->mem_is_pinned && !segment->mem_is_large);    
-    if (segment->allow_decommit) {
-      segment->decommit_expire = _mi_clock_now() + mi_option_get(mi_option_decommit_delay);
-      segment->decommit_mask = decommit_mask;
-      mi_assert_internal(mi_commit_mask_all_set(&segment->commit_mask, &segment->decommit_mask));
-      #if MI_DEBUG>2
-      const size_t commit_needed = _mi_divide_up(info_slices*MI_SEGMENT_SLICE_SIZE, MI_COMMIT_SIZE);
-      mi_commit_mask_t commit_needed_mask;
-      mi_commit_mask_create(0, commit_needed, &commit_needed_mask);
-      mi_assert_internal(!mi_commit_mask_any_set(&segment->decommit_mask, &commit_needed_mask));
-      #endif
-    }    
-    else {
-      mi_assert_internal(mi_commit_mask_is_empty(&decommit_mask));
-      segment->decommit_expire = 0;
-      mi_commit_mask_create_empty( &segment->decommit_mask );
-      mi_assert_internal(mi_commit_mask_is_empty(&segment->decommit_mask));
-    }
-  }
   
-
+  segment->commit_mask = commit_mask; // on lazy commit, the initial part is always committed
+  segment->allow_decommit = (mi_option_is_enabled(mi_option_allow_decommit) && !segment->mem_is_pinned && !segment->mem_is_large);    
+  if (segment->allow_decommit) {
+    segment->decommit_expire = _mi_clock_now() + mi_option_get(mi_option_decommit_delay);
+    segment->decommit_mask = decommit_mask;
+    mi_assert_internal(mi_commit_mask_all_set(&segment->commit_mask, &segment->decommit_mask));
+    #if MI_DEBUG>2
+    const size_t commit_needed = _mi_divide_up(info_slices*MI_SEGMENT_SLICE_SIZE, MI_COMMIT_SIZE);
+    mi_commit_mask_t commit_needed_mask;
+    mi_commit_mask_create(0, commit_needed, &commit_needed_mask);
+    mi_assert_internal(!mi_commit_mask_any_set(&segment->decommit_mask, &commit_needed_mask));
+    #endif
+  }    
+  
   // initialize segment info
   const size_t slice_entries = (segment_slices > MI_SLICES_PER_SEGMENT ? MI_SLICES_PER_SEGMENT : segment_slices);
   segment->segment_slices = segment_slices;
@@ -945,12 +943,6 @@ static mi_segment_t* mi_segment_init(mi_segment_t* segment, size_t required, siz
 }
 
 
-// Allocate a segment from the OS aligned to `MI_SEGMENT_SIZE` .
-static mi_segment_t* mi_segment_alloc(size_t required, size_t page_alignment, mi_arena_id_t req_arena_id, mi_segments_tld_t* tld, mi_os_tld_t* os_tld, mi_page_t** huge_page) {
-  return mi_segment_init(NULL, required, page_alignment, req_arena_id, tld, os_tld, huge_page);
-}
-
-
 static void mi_segment_free(mi_segment_t* segment, bool force, mi_segments_tld_t* tld) {
   MI_UNUSED(force);
   mi_assert_internal(segment != NULL);
@@ -1532,7 +1524,6 @@ static mi_page_t* mi_segment_huge_page_alloc(size_t size, size_t page_alignment,
   mi_assert_internal(segment->used==1);
   mi_assert_internal(mi_page_block_size(page) >= size);  
   segment->thread_id = 0; // huge segments are immediately abandoned
-  #if MI_DEBUG > 3
   if (page_alignment > 0) {
     size_t psize;
     uint8_t* p = _mi_segment_page_start(segment, page, &psize);
@@ -1546,7 +1537,6 @@ static mi_page_t* mi_segment_huge_page_alloc(size_t size, size_t page_alignment,
       _mi_os_decommit(decommit_start, decommit_size, os_tld->stats);
     }
   }
-  #endif
   // for huge pages we initialize the xblock_size as we may
   // overallocate to accommodate large alignments.
   size_t psize;
diff --git a/test/main-override.cpp b/test/main-override.cpp
index e0dba5a3..b205dc85 100644
--- a/test/main-override.cpp
+++ b/test/main-override.cpp
@@ -38,6 +38,8 @@ static void tsan_numa_test();         // issue #414
 static void strdup_test();            // issue #445 
 static void bench_alloc_large(void);  // issue #xxx
 
+static void test_stl_allocators();
+
 int main() {
   mi_stats_reset();  // ignore earlier allocations
 
@@ -50,6 +52,8 @@ int main() {
    tsan_numa_test();
    strdup_test();
 
+  test_stl_allocators();
+
   test_mt_shutdown();
   //fail_aslr();
   bench_alloc_large();
@@ -127,6 +131,43 @@ static bool test_stl_allocator2() {
   return vec.size() == 0;
 }
 
+static bool test_stl_allocator3() {
+  std::vector<int, mi_heap_stl_allocator<int> > vec;
+  vec.push_back(1);
+  vec.pop_back();
+  return vec.size() == 0;
+}
+
+static bool test_stl_allocator4() {
+  std::vector<some_struct, mi_heap_stl_allocator<some_struct> > vec;
+  vec.push_back(some_struct());
+  vec.pop_back();
+  return vec.size() == 0;
+}
+
+static bool test_stl_allocator5() {
+  std::vector<int, mi_heap_destroy_stl_allocator<int> > vec;
+  vec.push_back(1);
+  vec.pop_back();
+  return vec.size() == 0;
+}
+
+static bool test_stl_allocator6() {
+  std::vector<some_struct, mi_heap_destroy_stl_allocator<some_struct> > vec;
+  vec.push_back(some_struct());
+  vec.pop_back();
+  return vec.size() == 0;
+}
+
+static void test_stl_allocators() {
+  test_stl_allocator1();
+  test_stl_allocator2();
+  test_stl_allocator3();
+  test_stl_allocator4();
+  test_stl_allocator5();
+  test_stl_allocator6();
+}
+
 // issue 445
 static void strdup_test() {
 #ifdef _MSC_VER