diff --git a/src/libarena/lib.rs b/src/libarena/lib.rs
index 24e7a65e02a..ec9d4eaed9e 100644
--- a/src/libarena/lib.rs
+++ b/src/libarena/lib.rs
@@ -33,6 +33,7 @@ use std::cmp;
 use std::intrinsics::{TyDesc, get_tydesc};
 use std::intrinsics;
 use std::mem;
+use std::mem::min_align_of;
 use std::num;
 use std::ptr::read;
 use std::rc::Rc;
@@ -204,7 +205,7 @@ impl Arena {
     #[inline]
     fn alloc_copy<'a, T>(&'a mut self, op: || -> T) -> &'a T {
         unsafe {
-            let ptr = self.alloc_copy_inner(mem::size_of::<T>(), mem::min_align_of::<T>());
+            let ptr = self.alloc_copy_inner(mem::size_of::<T>(), min_align_of::<T>());
             let ptr: *mut T = transmute(ptr);
             mem::move_val_init(&mut (*ptr), op());
             return transmute(ptr);
@@ -261,7 +262,7 @@ impl Arena {
         unsafe {
             let tydesc = get_tydesc::<T>();
             let (ty_ptr, ptr) =
-                self.alloc_noncopy_inner(mem::size_of::<T>(), mem::min_align_of::<T>());
+                self.alloc_noncopy_inner(mem::size_of::<T>(), min_align_of::<T>());
             let ty_ptr: *mut uint = transmute(ty_ptr);
             let ptr: *mut T = transmute(ptr);
             // Write in our tydesc along with a bit indicating that it
@@ -353,7 +354,29 @@ struct TypedArenaChunk<T> {
 }
 
 impl<T> TypedArenaChunk<T> {
+    #[cfg(stage0)]
     #[inline]
+    fn new(next: Option<Box<TypedArenaChunk<T>>>, capacity: uint)
+           -> Box<TypedArenaChunk<T>> {
+        let mut size = mem::size_of::<TypedArenaChunk<T>>();
+        size = round_up(size, min_align_of::<T>());
+        let elem_size = mem::size_of::<T>();
+        let elems_size = elem_size.checked_mul(&capacity).unwrap();
+        size = size.checked_add(&elems_size).unwrap();
+
+        let mut chunk = unsafe {
+            let chunk = global_heap::exchange_malloc(size);
+            let mut chunk: Box<TypedArenaChunk<T>> = cast::transmute(chunk);
+            mem::move_val_init(&mut chunk.next, next);
+            chunk
+        };
+
+        chunk.capacity = capacity;
+        chunk
+    }
+
+    #[inline]
+    #[cfg(not(stage0))]
     fn new(next: Option<Box<TypedArenaChunk<T>>>, capacity: uint)
            -> Box<TypedArenaChunk<T>> {
         let mut size = mem::size_of::<TypedArenaChunk<T>>();
@@ -363,7 +386,7 @@ impl<T> TypedArenaChunk<T> {
         size = size.checked_add(&elems_size).unwrap();
 
         let mut chunk = unsafe {
-            let chunk = global_heap::exchange_malloc(size);
+            let chunk = global_heap::exchange_malloc(size, min_align_of::<TypedArenaChunk<T>>());
             let mut chunk: Box<TypedArenaChunk<T>> = cast::transmute(chunk);
             mem::move_val_init(&mut chunk.next, next);
             chunk
@@ -402,7 +425,7 @@ impl<T> TypedArenaChunk<T> {
     fn start(&self) -> *u8 {
         let this: *TypedArenaChunk<T> = self;
         unsafe {
-            cast::transmute(round_up(this.offset(1) as uint, mem::min_align_of::<T>()))
+            cast::transmute(round_up(this.offset(1) as uint, min_align_of::<T>()))
         }
     }
 
diff --git a/src/libcore/should_not_exist.rs b/src/libcore/should_not_exist.rs
index 0e6baaa518a..74bebc921e3 100644
--- a/src/libcore/should_not_exist.rs
+++ b/src/libcore/should_not_exist.rs
@@ -29,13 +29,29 @@ use str::StrSlice;
 
 #[allow(ctypes)]
 extern {
-    fn malloc(size: uint) -> *u8;
-    fn free(ptr: *u8);
+    #[cfg(stage0)]
+    fn rust_malloc(size: uint) -> *u8;
+    #[cfg(not(stage0))]
+    fn rust_malloc(size: uint, align: uint) -> *u8;
+    fn rust_free(ptr: *u8);
 }
 
+#[cfg(stage0)]
 unsafe fn alloc(cap: uint) -> *mut Vec<()> {
     let cap = cap.checked_add(&mem::size_of::<Vec<()>>()).unwrap();
-    let ret = malloc(cap) as *mut Vec<()>;
+    let ret = rust_malloc(cap) as *mut Vec<()>;
+    if ret.is_null() {
+        intrinsics::abort();
+    }
+    (*ret).fill = 0;
+    (*ret).alloc = cap;
+    ret
+}
+
+#[cfg(not(stage0))]
+unsafe fn alloc(cap: uint) -> *mut Vec<()> {
+    let cap = cap.checked_add(&mem::size_of::<Vec<()>>()).unwrap();
+    let ret = rust_malloc(cap, 8) as *mut Vec<()>;
     if ret.is_null() {
         intrinsics::abort();
     }
@@ -102,7 +118,7 @@ impl FromIterator<char> for ~str {
                     ptr::copy_nonoverlapping_memory(&mut (*ptr2).data,
                                                     &(*ptr).data,
                                                     len);
-                    free(ptr as *u8);
+                    rust_free(ptr as *u8);
                     cast::forget(ret);
                     ret = cast::transmute(ptr2);
                     ptr = ptr2;
@@ -172,7 +188,7 @@ impl<A: Clone> Clone for ~[A] {
                     for j in range(0, *i as int) {
                         ptr::read(&*p.offset(j));
                     }
-                    free(ret as *u8);
+                    rust_free(ret as *u8);
                 });
             cast::transmute(ret)
         }
diff --git a/src/librustc/middle/trans/base.rs b/src/librustc/middle/trans/base.rs
index 6c4566c09aa..8f878f099fc 100644
--- a/src/librustc/middle/trans/base.rs
+++ b/src/librustc/middle/trans/base.rs
@@ -345,7 +345,8 @@ fn require_alloc_fn(bcx: &Block, info_ty: ty::t, it: LangItem) -> ast::DefId {
 
 pub fn malloc_raw_dyn<'a>(bcx: &'a Block<'a>,
                           ptr_ty: ty::t,
-                          size: ValueRef)
+                          size: ValueRef,
+                          align: ValueRef)
                           -> Result<'a> {
     let _icx = push_ctxt("malloc_raw_exchange");
     let ccx = bcx.ccx();
@@ -353,7 +354,7 @@ pub fn malloc_raw_dyn<'a>(bcx: &'a Block<'a>,
     // Allocate space:
     let r = callee::trans_lang_call(bcx,
         require_alloc_fn(bcx, ptr_ty, ExchangeMallocFnLangItem),
-        [size],
+        [size, align],
         None);
 
     let llty_ptr = type_of::type_of(ccx, ptr_ty);
diff --git a/src/librustc/middle/trans/expr.rs b/src/librustc/middle/trans/expr.rs
index 806fb3e125d..463aa67ac90 100644
--- a/src/librustc/middle/trans/expr.rs
+++ b/src/librustc/middle/trans/expr.rs
@@ -67,7 +67,7 @@ use middle::typeck::MethodCall;
 use util::common::indenter;
 use util::ppaux::Repr;
 use util::nodemap::NodeMap;
-use middle::trans::machine::{llsize_of, llsize_of_alloc};
+use middle::trans::machine::{llalign_of_min, llsize_of, llsize_of_alloc};
 use middle::trans::type_::Type;
 
 use syntax::ast;
@@ -1170,10 +1170,11 @@ fn trans_uniq_expr<'a>(bcx: &'a Block<'a>,
     let fcx = bcx.fcx;
     let llty = type_of::type_of(bcx.ccx(), contents_ty);
     let size = llsize_of(bcx.ccx(), llty);
+    let align = C_uint(bcx.ccx(), llalign_of_min(bcx.ccx(), llty) as uint);
     // We need to a make a pointer type because box_ty is ty_bot
     // if content_ty is, e.g. box fail!().
     let real_box_ty = ty::mk_uniq(bcx.tcx(), contents_ty);
-    let Result { bcx, val } = malloc_raw_dyn(bcx, real_box_ty, size);
+    let Result { bcx, val } = malloc_raw_dyn(bcx, real_box_ty, size, align);
     // Unique boxes do not allocate for zero-size types. The standard library
     // may assume that `free` is never called on the pointer returned for
     // `Box<ZeroSizeType>`.
diff --git a/src/librustc/middle/trans/tvec.rs b/src/librustc/middle/trans/tvec.rs
index e210437a370..48179abc052 100644
--- a/src/librustc/middle/trans/tvec.rs
+++ b/src/librustc/middle/trans/tvec.rs
@@ -278,7 +278,9 @@ pub fn trans_uniq_vstore<'a>(bcx: &'a Block<'a>,
 
     let vecsize = Add(bcx, alloc, llsize_of(ccx, ccx.opaque_vec_type));
 
-    let Result { bcx: bcx, val: val } = malloc_raw_dyn(bcx, vec_ty, vecsize);
+    // ~[T] is not going to be changed to support alignment, since it's obsolete.
+    let align = C_uint(ccx, 8);
+    let Result { bcx: bcx, val: val } = malloc_raw_dyn(bcx, vec_ty, vecsize, align);
     Store(bcx, fill, GEPi(bcx, val, [0u, abi::vec_elt_fill]));
     Store(bcx, alloc, GEPi(bcx, val, [0u, abi::vec_elt_alloc]));
 
diff --git a/src/libstd/rt/global_heap.rs b/src/libstd/rt/global_heap.rs
index 7d54c3faf42..c8808b6e821 100644
--- a/src/libstd/rt/global_heap.rs
+++ b/src/libstd/rt/global_heap.rs
@@ -68,7 +68,7 @@ pub unsafe fn realloc_raw(ptr: *mut u8, size: uint) -> *mut u8 {
 }
 
 /// The allocator for unique pointers without contained managed pointers.
-#[cfg(not(test))]
+#[cfg(not(test), stage0)]
 #[lang="exchange_malloc"]
 #[inline]
 pub unsafe fn exchange_malloc(size: uint) -> *mut u8 {
@@ -85,6 +85,23 @@ pub unsafe fn exchange_malloc(size: uint) -> *mut u8 {
     }
 }
 
+/// The allocator for unique pointers without contained managed pointers.
+#[cfg(not(test), not(stage0))]
+#[lang="exchange_malloc"]
+#[inline]
+pub unsafe fn exchange_malloc(size: uint, _align: uint) -> *mut u8 {
+    // The compiler never calls `exchange_free` on ~ZeroSizeType, so zero-size
+    // allocations can point to this `static`. It would be incorrect to use a null
+    // pointer, due to enums assuming types like unique pointers are never null.
+    static EMPTY: () = ();
+
+    if size == 0 {
+        &EMPTY as *() as *mut u8
+    } else {
+        malloc_raw(size)
+    }
+}
+
 // FIXME: #7496
 #[cfg(not(test))]
 #[lang="closure_exchange_malloc"]
@@ -118,6 +135,32 @@ pub unsafe fn exchange_free(ptr: *u8) {
     free(ptr as *mut c_void);
 }
 
+// hack for libcore
+#[no_mangle]
+#[doc(hidden)]
+#[deprecated]
+#[cfg(stage0)]
+pub extern "C" fn rust_malloc(size: uint) -> *mut u8 {
+    unsafe { exchange_malloc(size) }
+}
+
+// hack for libcore
+#[no_mangle]
+#[doc(hidden)]
+#[deprecated]
+#[cfg(not(stage0))]
+pub extern "C" fn rust_malloc(size: uint, align: uint) -> *mut u8 {
+    unsafe { exchange_malloc(size, align) }
+}
+
+// hack for libcore
+#[no_mangle]
+#[doc(hidden)]
+#[deprecated]
+pub extern "C" fn rust_free(ptr: *u8) {
+    unsafe { exchange_free(ptr) }
+}
+
 #[cfg(test)]
 mod bench {
     extern crate test;