aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorJohn McCall <rjmccall@apple.com>2011-06-15 21:21:53 +0000
committerJohn McCall <rjmccall@apple.com>2011-06-15 21:21:53 +0000
commit8246702d0cbecc3fd5748b58614ffed7ad9e04a5 (patch)
tree37ed38fcdf3a6134450ce89faf1b3b7c32b4c772
parentf069c9ede9ad5d19a190ca8398b5926e86a31a04 (diff)
The specification document for the new ObjC Automatic Reference Counting
feature. Implementation to follow. :) git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@133090 91177308-0d34-0410-b5e6-96231b3b80d8
-rw-r--r--docs/AutomaticReferenceCounting.html1350
1 files changed, 1350 insertions, 0 deletions
diff --git a/docs/AutomaticReferenceCounting.html b/docs/AutomaticReferenceCounting.html
new file mode 100644
index 0000000000..3d34ddb310
--- /dev/null
+++ b/docs/AutomaticReferenceCounting.html
@@ -0,0 +1,1350 @@
+<html>
+<head>
+<title>Objective-C Automatic Reference Counting (ARC)</title>
+<link type="text/css" rel="stylesheet" href="../menu.css" />
+<link type="text/css" rel="stylesheet" href="../content.css" />
+<style type="text/css">
+/* Collapse the items in the ToC to the left. */
+div#toc ul {
+ padding-left: 0
+}
+
+/* Rationales appear in italic. */
+div.rationale {
+ font-style: italic
+}
+
+div h1 { font-size: 2em; margin: .67em 0 }
+div div h1 { font-size: 1.5em; margin: .75em 0 }
+div div div h1 { font-size: 1.17em; margin: .83em 0 }
+div div div div h1 { margin: 1.12em 0 }
+
+span.term { font-style: italic; font-weight: bold }
+</style>
+
+<script lang="javascript">
+/// A little script to recursively build a table of contents.
+function buildTOC(div, toc, ancestry) {
+ var children = div.childNodes;
+ var len = children.length;
+
+ var childNumber = 0;
+
+ var list = null;
+ for (var i = 0; i < len; ++i) {
+ var child = children[i];
+ if (child.nodeName != "DIV") continue;
+ if (child.getAttribute("class") == "rationale") continue;
+ if (child.id == "toc") continue;
+
+ // Okay, we're actually going to build a list node.
+ if (list === null) list = document.createElement("ul");
+
+ var childAncestry = ancestry + ++childNumber + ".";
+
+ var headerNode = child.childNodes[1];
+ var title = headerNode.innerHTML;
+ headerNode.insertBefore(document.createTextNode(childAncestry + " "),
+ headerNode.firstChild);
+
+ var item = document.createElement("li");
+ item.appendChild(document.createTextNode(childAncestry + " "));
+
+ var anchor = document.createElement("a");
+ anchor.href = "#" + child.id;
+ anchor.innerHTML = title;
+ item.appendChild(anchor);
+
+ buildTOC(child, item, childAncestry);
+
+ list.appendChild(item);
+ }
+ if (list) toc.appendChild(list);
+}
+
+function onLoad() {
+ var toc = document.getElementById("toc");
+ var content = document.getElementById("content");
+ buildTOC(content, toc, "");
+}
+window.onload = onLoad;
+
+</script>
+</head>
+<body>
+
+<!--#include virtual="../menu.html.incl"-->
+
+<div id="content">
+<h1>Automatic Reference Counting</h1>
+
+<div id="toc">
+</div>
+
+<div id="meta">
+<h1>About this document</h1>
+
+<div id="meta.purpose">
+<h1>Purpose</h1>
+
+<p>The first and primary purpose of this document is to serve as a
+complete technical specification of Automatic Reference Counting.
+Given a core Objective-C compiler and runtime, it should be possible
+to write a compiler and runtime which implements these new
+semantics.</p>
+
+<p>The secondary purpose is to act as a rationale for why ARC was
+designed in this way. This should remain tightly focused on the
+technical design and should not stray into marketing speculation.</p>
+
+</div> <!-- meta.purpose -->
+
+<div id="meta.background">
+<h1>Background</h1>
+
+<p>This document assumes a basic familiarity with C.</p>
+
+<p><span class="term">Blocks</span> are a C language extension for
+creating anonymous functions. Users interact with and transfer block
+objects using <span class="term">block pointers</span>, which are
+represented like a normal pointer. A block may capture values from
+local variables; when this occurs, memory must be dynamically
+allocated. The initial allocation is done on the stack, but the
+runtime provides a <tt>Block_copy</tt> function which, given a block
+pointer, either copies the underlying block object to the heap,
+setting its reference count to 1 and returning the new block pointer,
+or (if the block object is already on the heap) increases its
+reference count by 1. The paired function is <tt>Block_release</tt>,
+which decreases the reference count by 1 and destroys the object if
+the count reaches zero and is on the heap.</p>
+
+<p>Objective-C is a set of language extensions, significant enough to
+be considered a different language. It is a strict superset of C.
+The extensions can also be imposed on C++, producing a language called
+Objective-C++. The primary feature is a single-inheritance object
+system; we briefly describe the modern dialect.</p>
+
+<p>Objective-C defines a new type kind, collectively called
+the <span class="term">object pointer types</span>. This kind has two
+notable builtin members, <tt>id</tt> and <tt>Class</tt>; <tt>id</tt>
+is the final supertype of all object pointers. The validity of
+conversions between object pointer types is not checked at runtime.
+Users may define <span class="term">classes</span>; each class is a
+type, and the pointer to that type is an object pointer type. A class
+may have a superclass; its pointer type is a subtype of its
+superclass's pointer type. A class has a set
+of <span class="term">ivars</span>, fields which appear on all
+instances of that class. For every class <i>T</i> there's an
+associated metaclass; it has no fields, its superclass is the
+metaclass of <i>T</i>'s superclass, and its metaclass is a global
+class. Every class has a global object whose class is the
+class's metaclass; metaclasses have no associated type, so pointers to
+this object have type <tt>Class</tt>.</p>
+
+<p>A class declaration (<tt>@interface</tt>) declares a set
+of <span class="term">methods</span>. A method has a return type, a
+list of argument types, and a <span class="term">selector</span>: a
+name like <tt>foo:bar:baz:</tt>, where the number of colons
+corresponds to the number of formal arguments. A method may be an
+instance method, in which case it can be invoked on objects of the
+class, or a class method, in which case it can be invoked on objects
+of the metaclass. A method may be invoked by providing an object
+(called the <span class="term">receiver</span>) and a list of formal
+arguments interspersed with the selector, like so:</p>
+
+<pre>[receiver foo: fooArg bar: barArg baz: bazArg]</pre>
+
+<p>This looks in the dynamic class of the receiver for a method with
+this name, then in that class's superclass, etc., until it finds
+something it can execute. The receiver <q>expression</q> may also be
+the name of a class, in which case the actual receiver is the class
+object for that class, or (within method definitions) it may
+be <tt>super</tt>, in which case the lookup algorithm starts with the
+static superclass instead of the dynamic class. The actual methods
+dynamically found in a class are not those declared in the
+<tt>@interface</tt>, but those defined in a separate
+<tt>@implementation</tt> declaration; however, when compiling a
+call, typechecking is done based on the methods declared in the
+<tt>@interface</tt>.</p>
+
+<p>Method declarations may also be grouped into
+<span class="term">protocols</span>, which are not inherently
+associated with any class, but which classes may claim to follow.
+Object pointer types may be qualified with additional protocols that
+the object is known to support.</p>
+
+<p><span class="term">Class extensions</span> are collections of ivars
+and methods, designed to allow a class's <tt>@interface</tt> to be
+split across multiple files; however, there is still a primary
+implementation file which must see the <tt>@interface</tt>s of all
+class extensions.
+<span class="term">Categories</span> allow methods (but not ivars) to
+be declared <i>post hoc</i> on an arbitrary class; the methods in the
+category's <tt>@implementation</tt> will be dynamically added to that
+class's method tables which the category is loaded at runtime,
+replacing those methods in case of a collision.</p>
+
+<p>In the standard environment, objects are allocated on the heap, and
+their lifetime is manually managed using a reference count. This is
+done using two instance methods which all classes are expected to
+implement: <tt>retain</tt> increases the object's reference count by
+1, whereas <tt>release</tt> decreases it by 1 and calls the instance
+method <tt>dealloc</tt> if the count reaches 0. To simplify certain
+operations, there is also an <span class="term">autorelease
+pool</span>, a thread-local list of objects to call <tt>release</tt>
+on later; an object can be added to this pool by
+calling <tt>autorelease</tt> on it.</p>
+
+<p>Block pointers may be converted to type <tt>id</tt>; block objects
+are laid out in a way that makes them compatible with Objective-C
+objects. There is a builtin class that all block objects are
+considered to be objects of; this class implements <tt>retain</tt> by
+adjusting the reference count, not by calling <tt>Block_copy</tt>.</p>
+
+</div> <!-- meta.background -->
+
+</div> <!-- meta -->
+
+<div id="general">
+<h1>General</h1>
+
+<p>Automatic Reference Counting implements automatic memory management
+for Objective-C objects and blocks, freeing the programmer from the
+need explicitly insert retains and releases. It does not provide a
+cycle collector; users must explicitly manage lifetime instead.</p>
+
+<p>ARC may be explicitly enabled with the compiler
+flag <tt>-fobjc-arc</tt>. It may also be explicitly disabled with the
+compiler flag <tt>-fno-objc-arc</tt>. The last of these two flags
+appearing on the compile line <q>wins</q>.</p>
+
+<p>If ARC is enabled, <tt>__has_feature(objc_arc)</tt> will expand to
+1 in the preprocessor. For more information about <tt>__has_feature</tt>,
+see the <a href="LanguageExtensions.html#__has_feature_extension">language
+extensions</a> document.</p>
+
+</div>
+
+<div id="objects">
+<h1>Retainable object pointers</h1>
+
+<p>This section describes retainable object pointers, their basic
+operations, and the restrictions imposed on their use under ARC. Note
+in particular that it covers the rules for pointer <em>values</em>
+(patterns of bits indicating the location of a pointed-to object), not
+pointer
+<em>objects</em> (locations in memory which store pointer values).
+The rules for objects are covered in the next section.</p>
+
+<p>A <span class="term">retainable object pointer</span>
+(or <q>retainable pointer</q>) is a value of
+a <span class="term">retainable object pointer type</span>
+(<q>retainable type</q>). There are three kinds of retainable object
+pointer types:</p>
+<ul>
+<li>block pointers (formed by applying the caret (<tt>^</tt>)
+declarator sigil to a function type)</li>
+<li>Objective-C object pointers (<tt>id</tt>, <tt>Class</tt>, <tt>NSFoo*</tt>, etc.)</li>
+<li>typedefs marked with <tt>__attribute__((NSObject))</tt></li>
+</ul>
+
+<p>Other pointer types, such as <tt>int*</tt> and <tt>CFStringRef</tt>,
+are not subject to ARC's semantics and restrictions.</p>
+
+<div class="rationale">
+
+<p>Rationale: We are not at liberty to require
+all code to be recompiled with ARC; therefore, ARC must interoperate
+with Objective-C code which manages retains and releases manually. In
+general, there are three requirements in order for a
+compiler-supported reference-count system to provide reliable
+interoperation:</p>
+
+<ul>
+<li>The type system must reliably identify which objects are to be
+managed. An <tt>int*</tt> might be a pointer to a <tt>malloc</tt>'ed
+array, or it might be a interior pointer to such an array, or it might
+point to some field or local variable. In contrast, values of the
+retainable object pointer types are never interior.</li>
+<li>The type system must reliably indicate how to
+manage objects of a type. This usually means that the type must imply
+a procedure for incrementing and decrementing retain counts.
+Supporting single-ownership objects requires a lot more explicit
+mediation in the language.</li>
+<li>There must be reliable conventions for whether and
+when <q>ownership</q> is passed between caller and callee, for both
+arguments and return values. Objective-C methods follow such a
+convention very reliably, at least for system libraries on Mac OS X,
+and functions always pass objects at +0. The C-based APIs for Core
+Foundation objects, on the other hand, have much more varied transfer
+semantics.</li>
+</ul>
+</div> <!-- rationale -->
+
+<p>The use of <tt>__attribute__((NSObject))</tt> typedefs is not
+recommended. If it's absolutely necessary to use this attribute, be
+very explicit about using the typedef, and do not assume that it will
+be preserved by language features like <tt>__typeof</tt> and C++
+template argument substitution.</p>
+
+<div class="rationale"><p>Rationale: any compiler operation which
+incidentally strips type <q>sugar</q> from a type will yield a type
+without the attribute, which may result in unexpected
+behavior.</p></div>
+
+<div id="objects.retains">
+<h1>Retain count semantics</h1>
+
+<p>A retainable object pointer is either a <span class="term">null
+pointer</span> or a pointer to a valid object. Furthermore, if it has
+block pointer type and is not <tt>null</tt> then it must actually be a
+pointer to a block object, and if it has <tt>Class</tt> type (possibly
+protocol-qualified) then it must actually be a pointer to a class
+object. Otherwise ARC does not enforce the Objective-C type system as
+long as the implementing methods follow the signature of the static
+type. It is undefined behavior if ARC is exposed to an invalid
+pointer.</p>
+
+<p>For ARC's purposes, a valid object is one with <q>well-behaved</q>
+retaining operations. Specifically, the object must be laid out such
+that the Objective-C message send machinery can successfully send it
+the following messages:</p>
+
+<ul>
+<li><tt>retain</tt>, taking no arguments and returning a pointer to
+the object.</li>
+<li><tt>release</tt>, taking no arguments and returning <tt>void</tt>.</li>
+<li><tt>autorelease</tt>, taking no arguments and returning a pointer
+to the object.</li>
+</ul>
+
+<p>The behavior of these methods is constrained in the following ways.
+The term <span class="term">high-level semantics</span> is an
+intentionally vague term; the intent is that programmers must
+implement these methods in a way such that the compiler, modifying
+code in ways it deems safe according to these constraints, will not
+violate their requirements. For example, if the user puts logging
+statements in <tt>retain</tt>, they should not be surprised if those
+statements are executed more or less often depending on optimization
+settings. These constraints are not exhaustive of the optimization
+opportunities: values held in local variables are subject to
+additional restrictions, described later in this document.</p>
+
+<p>It is undefined behavior if a computation history featuring a send
+of <tt>retain</tt> followed by a send of <tt>release</tt> to the same
+object, with no intervening <tt>release</tt> on that object, is not
+equivalent under the high-level semantics to a computation
+history in which these sends are removed. Note that this implies that
+these methods may not raise exceptions.</p>
+
+<p>It is undefined behavior if a computation history features any use
+whatsoever of an object following the completion of a send
+of <tt>release</tt> that is not preceded by a send of <tt>retain</tt>
+to the same object.</p>
+
+<p>The behavior of <tt>autorelease</tt> must be equivalent to sending
+<tt>release</tt> when one of the autorelease pools currently in scope
+is popped. It may not throw an exception.</p>
+
+<p>When the semantics call for performing one of these operations on a
+retainable object pointer, if that pointer is <tt>null</tt> then the
+effect is a no-op.</p>
+
+<p>All of the semantics described in this document are subject to
+additional <a href="#optimization">optimization rules</a> which permit
+the removal or optimization of operations based on local knowledge of
+data flow. The semantics describe the high-level behaviors that the
+compiler implements, not an exact sequence of operations that a
+program will be compiled into.</p>
+
+</div> <!-- objects.retains -->
+
+<div id="objects.operands">
+<h1>Retainable object pointers as operands and arguments</h1>
+
+<p>In general, ARC does not perform retain or release operations when
+simply using a retainable object pointer as an operand within an
+expression. This includes:</p>
+<ul>
+<li>loading a retainable pointer from an object with non-weak
+<a href="#ownership">ownership</a>,</li>
+<li>passing a retainable pointer as an argument to a function or
+method, and</li>
+<li>receiving a retainable pointer as the result of a function or
+method call.</li>
+</ul>
+
+<div class="rationale"><p>Rationale: while this might seem
+uncontroversial, it is actually unsafe when multiple expressions are
+evaluated in <q>parallel</q>, as with binary operators and calls,
+because (for example) one expression might load from an object while
+another writes to it. However, C and C++ already call this undefined
+behavior because the evaluations are unsequenced, and ARC simply
+exploits that here to avoid needing to retain arguments across a large
+number of calls.</p></div>
+
+<p>The remainder of this section describes exceptions to these rules,
+how those exceptions are detected, and what those exceptions imply
+semantically.</p>
+
+<div id="objects.operands.consumed">
+<h1>Consumed parameters</h1>
+
+<p>A function or method parameter of retainable object pointer type
+may be marked as <span class="term">consumed</span>, signifying that
+the callee expects to take ownership of a +1 retain count. This is
+done by adding the <tt>ns_consumed</tt> attribute to the parameter
+declaration, like so:</p>
+
+<pre>void foo(__attribute((ns_consumed)) id x);
+- (void) foo: (id) __attribute((ns_consumed)) x;</pre>
+
+<p>This attribute is part of the type of the function or method, not
+the type of the parameter. It controls only how the argument is
+passed and received.</p>
+
+<p>When passing such an argument, ARC retains the argument prior to
+making the call.</p>
+
+<p>When receiving such an argument, ARC releases the argument at the
+end of the function, subject to the usual optimizations for local
+values.</p>
+
+<div class="rationale"><p>Rationale: this formalizes direct transfers
+of ownership from a caller to a callee. The most common scenario here
+is passing the <tt>self</tt> parameter to <tt>init</tt>, but it is
+useful to generalize. Typically, local optimization will remove any
+extra retains and releases: on the caller side the retain will be
+merged with a +1 source, and on the callee side the release will be
+rolled into the initialization of the parameter.</p></div>
+
+<p>The implicit <tt>self</tt> parameter of a method may be marked as
+consumed by adding <tt>__attribute__((ns_consumes_self))</tt> to the
+method declaration. Methods in
+the <tt>init</tt> <a href="#family">family</a> are implicitly
+marked <tt>__attribute__((ns_consumes_self))</tt>.</p>
+
+<p>It is undefined behavior if an Objective-C message send of a method
+with <tt>ns_consumed</tt> parameters (other than self) is made to a
+null pointer.</p>
+
+<div class="rationale"><p>Rationale: in fact, it's probably a
+guaranteed leak.</p></div>
+
+</div>
+
+<div id="objects.operands.retained-returns">
+<h1>Retained return values</h1>
+
+<p>A function or method which returns a retainable object pointer type
+may be marked as returning a retained value, signifying that the
+caller expects to take ownership of a +1 retain count. This is done
+by adding the <tt>ns_returns_retained</tt> attribute to the function or
+method declaration, like so:</p>
+
+<pre>id foo(void) __attribute((ns_returns_retained));
+- (id) foo __attribute((ns_returns_retained));</pre>
+
+<p>This attribute is part of the type of the function or method.</p>
+
+<p>When returning from such a function or method, ARC retains the
+value at the point of evaluation of the return statement, before
+leaving all local scopes.</p>
+
+<p>When receiving a return result from such a function or method, ARC
+releases the value at the end of the full-expression it is contained
+within, subject to the usual optimizations for local values.</p>
+
+<div class="rationale"><p>Rationale: this formalizes direct transfers of
+ownership from a callee to a caller. The most common scenario this
+models is the retained return from <tt>init</tt>, <tt>alloc</tt>,
+<tt>new</tt>, and <tt>copy</tt> methods, but there are other cases in
+the frameworks. After optimization there are typically no extra
+retains and releases required.</p></div>
+
+<p>Methods in
+the <tt>alloc</tt>, <tt>copy</tt>, <tt>init</tt>, <tt>mutableCopy</tt>,
+and <tt>new</tt> <a href="#family">families</a> are implicitly marked
+<tt>__attribute__((ns_returns_retained))</tt>. This may be suppressed
+by explicitly marking the
+method <tt>__attribute__((ns_returns_not_retained))</tt>.</p>
+</div>
+
+<div id="objects.operands.other-returns">
+<h1>Unretained return values</h1>
+
+<p>A method or function which returns a retainable object type but
+does not return a retained value must ensure that the object is
+still valid across the return boundary.</p>
+
+<p>When returning from such a function or method, ARC retains the
+value at the point of evaluation of the return statement, then leaves
+all local scopes, and then balances out the retain while ensuring that
+the value lives across the call boundary. In the worst case, this may
+involve an <tt>autorelease</tt>, but callers must not assume that the
+value is actually in the autorelease pool.</p>
+
+<p>ARC performs no extra mandatory work on the caller side, although
+it may elect to do something to shorten the lifetime of the returned
+value.</p>
+
+<div class="rationale"><p>Rationale: it is common in non-ARC code to not
+return an autoreleased value; therefore the convention does not force
+either path. It is convenient to not be required to do unnecessary
+retains and autoreleases; this permits optimizations such as eliding
+retain/autoreleases when it can be shown that the original pointer
+will still be valid at the point of return.</p></div>
+
+<p>A method or function may be marked
+with <tt>__attribute__((ns_returns_autoreleased))</tt> to indicate
+that it returns a pointer which is guaranteed to be valid at least as
+long as the innermost autorelease pool. There are no additional
+semantics enforced in the definition of such a method; it merely
+enables optimizations in callers.</p>
+</div>
+
+<div id="objects.operands.casts">
+<h1>Bridged casts</h1>
+
+<p>A <span class="term">bridged cast</span> is a C-style cast
+annotated with one of three keywords:</p>
+
+<ul>
+<li><tt>(__bridge T) op</tt> casts the operand to the destination
+type <tt>T</tt>. If <tt>T</tt> is a retainable object pointer type,
+then <tt>op</tt> must have a non-retainable pointer type.
+If <tt>T</tt> is a non-retainable pointer type, then <tt>op</tt> must
+have a retainable object pointer type. Otherwise the cast is
+ill-formed. There is no transfer of ownership, and ARC inserts
+no retain operations.</li>
+
+<li><tt>(__bridge_retained T) op</tt> casts the operand, which must
+have retainable object pointer type, to the destination type, which
+must be a non-retainable pointer type. ARC retains the value, subject
+to the usual optimizations on local values, and the recipient is
+responsible for balancing that +1.</li>
+
+<li><tt>(__bridge_transfer T) op</tt> casts the operand, which must
+have non-retainable pointer type, to the destination type, which must
+be a retainable object pointer type. ARC will release the value at
+the end of the enclosing full-expression, subject to the usual
+optimizations on local values.</li>
+</ul>
+
+<p>These casts are required in order to transfer objects in and out of
+ARC control; see the rationale in the section
+on <a href="#objects.restrictions.conversion">conversion of retainable
+object pointers</a>.</p>
+
+<p>Using a <tt>__bridge_retained</tt> or <tt>__bridge_transfer</tt>
+cast purely to convince ARC to emit an unbalanced retain or release,
+respectively, is poor form.</p>
+
+</div>
+
+</div>
+
+<div id="objects.restrictions">
+<h1>Restrictions</h1>
+
+<div id="objects.restrictions.conversion">
+<h1>Conversion of retainable object pointers</h1>
+
+<p>In general, a program which attempts to implicitly or explicitly
+convert a value of retainable object pointer type to any
+non-retainable type, or vice-versa, is ill-formed. For example, an
+Objective-C object pointer shall not be converted to <tt>intptr_t</tt>
+or <tt>void*</tt>. The <a href="#objects.operands.casts">bridged
+casts</a> may be used to perform these conversions where
+necessary.</p>
+
+<div class="rationale"><p>Rationale: we cannot ensure the correct
+management of the lifetime of objects if they may be freely passed
+around as unmanaged types. The bridged casts are provided so that the
+programmer may explicitly describe whether the cast transfers control
+into or out of ARC.</p></div>
+</div>
+
+<p>An unbridged cast to a retainable object pointer type of the return
+value of a Objective-C message send which yields a non-retainable
+pointer is treated as a <tt>__bridge_transfer</tt> cast
+if:</p>
+
+<ul>
+<li>the method has the <tt>cf_returns_retained</tt> attribute, or if
+not that,</li>
+<li>the method does not have the <tt>cf_returns_not_retained</tt>
+attribute and</li>
+<li>the method's <a href="#family">selector family</a> would imply
+the <tt>ns_returns_retained</tt> attribute on a method which returned
+a retainable object pointer type.</li>
+</ul>
+
+<p>Otherwise the cast is treated as a <tt>__bridge</tt> cast.</p>
+
+</div>
+
+</div>
+
+<div id="ownership">
+<h1>Ownership qualification</h1>
+
+<p>This section describes the behavior of <em>objects</em> of
+retainable object pointer type; that is, locations in memory which
+store retainable object pointers.</p>
+
+<p>A type is a <span class="term">retainable object owner type</span>
+if it is a retainable object pointer type or an array type whose
+element type is a retainable object owner type.</p>
+
+<p>An <span class="term">ownership qualifier</span> is a type
+qualifier which applies only to retainable object owner types. A
+program is ill-formed if it attempts to apply an ownership qualifier
+to a type which is already ownership-qualified, even if it is the same
+qualifier. An array type is ownership-qualified according to its
+element type, and adding an ownership qualifier to an array type so
+qualifies its element type.</p>
+
+<p>Except as described under
+the <a href="#ownership.inference">inference rules</a>, a program is
+ill-formed if it attempts to form a pointer or reference type to a
+retainable object owner type which lacks an ownership qualifier.</p>
+
+<div class="rationale"><p>Rationale: these rules, together with the
+inference rules, ensure that all objects and lvalues of retainable
+object pointer type have an ownership qualifier.</p></div>
+
+<p>There are four ownership qualifiers:</p>
+
+<ul>
+<li><tt>__autoreleasing</tt></li>
+<li><tt>__strong</tt></li>
+<li><tt>__unsafe_unretained</tt></li>
+<li><tt>__weak</tt></li>
+</ul>
+
+<p>A type is <span class="term">nontrivially ownership-qualified</span>
+if it is qualified with <tt>__autoreleasing</tt>, <tt>__strong</tt>, or
+<tt>__weak</tt>.</p>
+
+<div id="ownership.spelling">
+<h1>Spelling</h1>
+
+<p>The names of the ownership qualifiers are reserved for the
+implementation. A program may not assume that they are or are not
+implemented with macros, or what those macros expand to.</p>
+
+<p>An ownership qualifier may be written anywhere that any other type
+qualifier may be written.</p>
+
+<p>If an ownership qualifier appears in
+the <i>declaration-specifiers</i>, the following rules apply:</p>
+
+<ul>
+<li>if the type specifier is a retainable object owner type, the
+qualifier applies to that type;</li>
+<li>if the outermost non-array part of the declarator is a pointer or
+block pointer, the qualifier applies to that type;</li>
+<li>otherwise the program is ill-formed.</li>
+</ul>
+
+<p>If an ownership qualifier appears on the declarator name, or on the
+declared object, it is applied to outermost pointer or block-pointer
+type.</p>
+
+<p>If an ownership qualifier appears anywhere else in a declarator, it
+applies to the type there.</p>
+
+</div> <!-- ownership.spelling -->
+
+<div id="ownership.semantics">
+<h1>Semantics</h1>
+
+<p>There are five <span class="term">managed operations</span> which
+may be performed on an object of retainable object pointer type. Each
+qualifier specifies different semantics for each of these operations.
+It is still undefined behavior to access an object outside of its
+lifetime.</p>
+
+<p>A load or store with <q>primitive semantics</q> has the same
+semantics as the respective operation would have on an <tt>void*</tt>
+lvalue with the same alignment and non-ownership qualification.</p>
+
+<p><span class="term">Reading</span> occurs when performing a
+lvalue-to-rvalue conversion on an object lvalue.
+
+<ul>
+<li>For <tt>__weak</tt> objects, the current pointee is retained and
+then released at the end of the current full-expression. This must
+execute atomically with respect to assignments and to the final
+release of the pointee.</li>
+<li>For all other objects, the lvalue is loaded with primitive
+semantics.</li>
+</ul>
+</p>
+
+<p><span class="term">Assignment</span> occurs when evaluating
+an assignment operator. The semantics vary based on the qualification:
+<ul>
+<li>For <tt>__strong</tt> objects, the new pointee is first retained;
+second, the lvalue is loaded with primitive semantics; third, the new
+pointee is stored into the lvalue with primitive semantics; and
+finally, the old pointee is released. This is not performed
+atomically; external synchronization must be used to make this safe in
+the face of concurrent loads and stores.</li>
+<li>For <tt>__weak</tt> objects, the lvalue is updated to point to the
+new pointee, unless that object is currently undergoing deallocation,
+in which case it the lvalue is updated to a null pointer. This must
+execute atomically with respect to other assignments to the object, to
+reads from the object, and to the final release of the new pointed-to
+value.</li>
+<li>For <tt>__unsafe_unretained</tt> objects, the new pointee is
+stored into the lvalue using primitive semantics.</li>
+<li>For <tt>__autoreleasing</tt> objects, the new pointee is retained,
+autoreleased, and stored into the lvalue using primitive semantics.</li>
+</ul>
+</p>
+
+<p><span class="term">Initialization</span> occurs when an object's
+lifetime begins, which depends on its storage duration.
+Initialization proceeds in two stages:
+<ol>
+<li>First, a null pointer is stored into the lvalue using primitive
+semantics. This step is skipped if the object
+is <tt>__unsafe_unretained</tt>.</li>
+<li>Second, if the object has an initializer, that expression is
+evaluated and then assigned into the object using the usual assignment
+semantics.</li>
+</ol>
+</p>
+
+<p><span class="term">Destruction</span> occurs when an object's
+lifetime ends. In all cases it is semantically equivalent to
+assigning a null pointer to the object, with the proviso that of
+course the object cannot be legally read after the object's lifetime
+ends.</p>
+
+<p><span class="term">Moving</span> occurs in specific situations
+where an lvalue is <q>moved from</q>, meaning that its current pointee
+will be used but the object may be left in a different (but still
+valid) state. This arises with <tt>__block</tt> variables and rvalue
+references in C++. For <tt>__strong</tt> lvalues, moving is equivalent
+to loading the lvalue with primitive semantics, writing a null pointer
+to it with primitive semantics, and then releasing the result of the
+load at the end of the current full-expression. For all other
+lvalues, moving is equivalent to reading the object.</p>
+
+</div> <!-- ownership.semantics -->
+
+<div id="ownership.restrictions">
+<h1>Restrictions</h1>
+
+<div id="ownership.restrictions.autoreleasing">
+<h1>Storage duration of<tt> __autoreleasing</tt> objects</h1>
+
+<p>A program is ill-formed if it declares an <tt>__autoreleasing</tt>
+object of non-automatic storage duration.</p>
+
+<div class="rationale"><p>Rationale: autorelease pools are tied to the
+current thread and scope by their nature. While it is possible to
+have temporary objects whose instance variables are filled with
+autoreleased objects, there is no way that ARC can provide any sort of
+safety guarantee there.</p></div>
+
+<p>It is undefined behavior if a non-null pointer is assigned to
+an <tt>__autoreleasing</tt> object while an autorelease pool is in
+scope and then that object is read after the autorelease pool's scope
+is left.</p>
+
+</div>
+
+<div id="ownership.restrictions.conversion.indirect">
+<h1>Conversion of pointers to ownership-qualified types</h1>
+
+<p>A program is ill-formed if an expression of type <tt>T*</tt> is
+converted, explicitly or implicitly, to the type <tt>U*</tt>,
+where <tt>T</tt> and <tt>U</tt> have different ownership
+qualification, unless:
+<ul>
+<li><tt>T</tt> is qualified with <tt>__strong</tt>,
+ <tt>__autoreleasing</tt>, or <tt>__unsafe_unretained</tt>, and
+ <tt>U</tt> is qualified with both <tt>const</tt> and
+ <tt>__unsafe_unretained</tt>; or</li>
+<li>either <tt>T</tt> or <tt>U</tt> is <tt>cv void</tt>, where
+<tt>cv</tt> is an optional sequence of non-ownership qualifiers; or</li>
+<li>the conversion is requested with a <tt>reinterpret_cast</tt> in
+ Objective-C++; or</li>
+<li>the conversion is a
+well-formed <a href="#ownership.restrictions.pass_by_writeback">pass-by-writeback</a>.</li>
+</ul>
+</p>
+
+<p>The analogous rule applies to <tt>T&</tt> and <tt>U&</tt> in
+Objective-C++.</p>
+
+<div class="rationale"><p>Rationale: these rules provide a reasonable
+level of type-safety for indirect pointers, as long as the underlying
+memory is not deallocated. The conversion to <tt>const
+__unsafe_unretained</tt> is permitted because the semantics of reads
+are equivalent across all these ownership semantics, and that's a very
+useful and common pattern. The interconversion with <tt>void*</tt> is
+useful for allocating memory or otherwise escaping the type system,
+but use it carefully. <tt>reinterpret_cast</tt> is considered to be
+an obvious enough sign of taking responsibility for any
+problems.</p></div>
+
+<p>It is undefined behavior to access an ownership-qualified object
+through an lvalue of a differently-qualified type, except that any
+non-<tt>__weak</tt> object may be read through
+an <tt>__unsafe_unretained</tt> lvalue.</p>
+
+<p>It is undefined behavior if a managed operation is performed on
+a <tt>__strong</tt> or <tt>__weak</tt> object without a guarantee that
+it contains a primitive zero bit-pattern, or if the storage for such
+an object is freed or reused without the object being first assigned a
+null pointer.</p>
+
+<div class="rationale"><p>Rationale: ARC cannot differentiate between
+an assignment operator which is intended to <q>initialize</q> dynamic
+memory and one which is intended to potentially replace a value.
+Therefore the object's pointer must be valid before letting ARC at it.
+Similarly, C and Objective-C do not provide any language hooks for
+destroying objects held in dynamic memory, so it is the programmer's
+responsibility to avoid leaks (<tt>__strong</tt> objects) and
+consistency errors (<tt>__weak</tt> objects).</p>
+
+<p>These requirements are followed automatically in Objective-C++ when
+creating objects of retainable object owner type with <tt>new</tt>
+or <tt>new[]</tt> and destroying them with <tt>delete</tt>,
+<tt>delete[]</tt>, or a pseudo-destructor expression. Note that
+arrays of nontrivially-ownership-qualified type are not ABI compatible
+with non-ARC code because the element type is non-POD: such arrays
+that are <tt>new[]</tt>'d in ARC translation units cannot
+be <tt>delete[]</tt>'d in non-ARC translation units and
+vice-versa.</p></div>
+
+</div>
+
+<div id="ownership.restrictions.pass_by_w