Blame - docs/InternalsManual.html - platform/external/clang

blob: 0353c5b274c7a8864f9096ac7fc8da61ed1df6dd [file] [log] [blame]

Ted Kremenek	17a295d	2008-06-11 06:19:49 +0000	[diff] [blame]	1	<html>
				2	<head>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	3	<title>"clang" CFE Internals Manual</title>
Ted Kremenek	17a295d	2008-06-11 06:19:49 +0000	[diff] [blame]	4	<link type="text/css" rel="stylesheet" href="../menu.css" />
				5	<link type="text/css" rel="stylesheet" href="../content.css" />
Sebastian Redl	6816856	2008-11-22 22:16:45 +0000	[diff] [blame]	6	<style type="text/css">
				7	td {
				8	vertical-align: top;
				9	}
				10	</style>
Ted Kremenek	17a295d	2008-06-11 06:19:49 +0000	[diff] [blame]	11	</head>
				12	<body>
				13
				14	<!--#include virtual="../menu.html.incl"-->
				15
				16	<div id="content">
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	17
				18	<h1>"clang" CFE Internals Manual</h1>
				19
				20	<ul>
				21	<li><a href="#intro">Introduction</a></li>
				22	<li><a href="#libsystem">LLVM System and Support Libraries</a></li>
				23	<li><a href="#libbasic">The clang 'Basic' Library</a>
				24	<ul>
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	25	<li><a href="#Diagnostics">The Diagnostics Subsystem</a></li>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	26	<li><a href="#SourceLocation">The SourceLocation and SourceManager
				27	classes</a></li>
				28	</ul>
				29	</li>
				30	<li><a href="#liblex">The Lexer and Preprocessor Library</a>
				31	<ul>
				32	<li><a href="#Token">The Token class</a></li>
				33	<li><a href="#Lexer">The Lexer class</a></li>
Chris Lattner	7928125	2008-03-09 02:27:26 +0000	[diff] [blame]	34	<li><a href="#TokenLexer">The TokenLexer class</a></li>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	35	<li><a href="#MultipleIncludeOpt">The MultipleIncludeOpt class</a></li>
				36	</ul>
				37	</li>
				38	<li><a href="#libparse">The Parser Library</a>
				39	<ul>
				40	</ul>
				41	</li>
				42	<li><a href="#libast">The AST Library</a>
				43	<ul>
				44	<li><a href="#Type">The Type class and its subclasses</a></li>
				45	<li><a href="#QualType">The QualType class</a></li>
Douglas Gregor	2e1cd42	2008-11-17 14:58:09 +0000	[diff] [blame]	46	<li><a href="#DeclarationName">Declaration names</a></li>
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	47	<li><a href="#CFG">The CFG class</a></li>
Chris Lattner	7bad199	2008-11-16 21:48:07 +0000	[diff] [blame]	48	<li><a href="#Constants">Constant Folding in the Clang AST</a></li>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	49	</ul>
				50	</li>
				51	</ul>
				52
				53
				54	<!-- ======================================================================= -->
				55	<h2 id="intro">Introduction</h2>
				56	<!-- ======================================================================= -->
				57
				58	<p>This document describes some of the more important APIs and internal design
				59	decisions made in the clang C front-end. The purpose of this document is to
				60	both capture some of this high level information and also describe some of the
				61	design decisions behind it. This is meant for people interested in hacking on
				62	clang, not for end-users. The description below is categorized by
				63	libraries, and does not describe any of the clients of the libraries.</p>
				64
				65	<!-- ======================================================================= -->
				66	<h2 id="libsystem">LLVM System and Support Libraries</h2>
				67	<!-- ======================================================================= -->
				68
				69	<p>The LLVM libsystem library provides the basic clang system abstraction layer,
				70	which is used for file system access. The LLVM libsupport library provides many
				71	underlying libraries and <a
				72	href="http://llvm.org/docs/ProgrammersManual.html">data-structures</a>,
				73	including command line option
				74	processing and various containers.</p>
				75
				76	<!-- ======================================================================= -->
				77	<h2 id="libbasic">The clang 'Basic' Library</h2>
				78	<!-- ======================================================================= -->
				79
				80	<p>This library certainly needs a better name. The 'basic' library contains a
				81	number of low-level utilities for tracking and manipulating source buffers,
				82	locations within the source buffers, diagnostics, tokens, target abstraction,
				83	and information about the subset of the language being compiled for.</p>
				84
				85	<p>Part of this infrastructure is specific to C (such as the TargetInfo class),
				86	other parts could be reused for other non-C-based languages (SourceLocation,
				87	SourceManager, Diagnostics, FileManager). When and if there is future demand
				88	we can figure out if it makes sense to introduce a new library, move the general
				89	classes somewhere else, or introduce some other solution.</p>
				90
				91	<p>We describe the roles of these classes in order of their dependencies.</p>
				92
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	93
				94	<!-- ======================================================================= -->
				95	<h3 id="Diagnostics">The Diagnostics Subsystem</h3>
				96	<!-- ======================================================================= -->
				97
				98	<p>The Clang Diagnostics subsystem is an important part of how the compiler
				99	communicates with the human. Diagnostics are the warnings and errors produced
				100	when the code is incorrect or dubious. In Clang, each diagnostic produced has
				101	(at the minimum) a unique ID, a <a href="#SourceLocation">SourceLocation</a> to
				102	"put the caret", an English translation associated with it, and a severity (e.g.
				103	<tt>WARNING</tt> or <tt>ERROR</tt>). They can also optionally include a number
				104	of arguments to the dianostic (which fill in "%0"'s in the string) as well as a
				105	number of source ranges that related to the diagnostic.</p>
				106
				107	<p>In this section, we'll be giving examples produced by the clang command line
				108	driver, but diagnostics can be <a href="#DiagnosticClient">rendered in many
				109	different ways</a> depending on how the DiagnosticClient interface is
				110	implemented. A representative example of a diagonstic is:</p>
				111
				112	<pre>
				113	t.c:38:15: error: invalid operands to binary expression ('int *' and '_Complex float')
				114	<font color="darkgreen">P = (P-42) + Gamma*4;</font>
				115	<font color="blue">~~~~~~ ^ ~~~~~~~</font>
				116	</pre>
				117
				118	<p>In this example, you can see the English translation, the severity (error),
				119	you can see the source location (the caret ("^") and file/line/column info),
				120	the source ranges "~~~~", arguments to the diagnostic ("int*" and "_Complex
				121	float"). You'll have to believe me that there is a unique ID backing the
				122	diagnostic :).</p>
				123
				124	<p>Getting all of this to happen has several steps and involves many moving
				125	pieces, this section describes them and talks about best practices when adding
				126	a new diagnostic.</p>
				127
				128	<!-- ============================ -->
				129	<h4>The DiagnosticKinds.def file</h4>
				130	<!-- ============================ -->
				131
				132	<p>Diagnostics are created by adding an entry to the <tt><a
				133	href="http://llvm.org/svn/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticKinds.def"
				134	>DiagnosticKinds.def</a></tt> file. This file encodes the unique ID of the
				135	diagnostic (as an enum, the first argument), the severity of the diagnostic
				136	(second argument) and the English translation + format string.</p>
				137
				138	<p>There is little sanity with the naming of the unique ID's right now. Some
				139	start with err_, warn_, ext_ to encode the severity into the name. Since the
				140	enum is referenced in the C++ code that produces the diagnostic, it is somewhat
				141	useful for it to be reasonably short.</p>
				142
				143	<p>The severity of the diagnostic comes from the set {<tt>NOTE</tt>,
				144	<tt>WARNING</tt>, <tt>EXTENSION</tt>, <tt>EXTWARN</tt>, <tt>ERROR</tt>}. The
				145	<tt>ERROR</tt> severity is used for diagnostics indicating the program is never
				146	acceptable under any circumstances. When an error is emitted, the AST for the
				147	input code may not be fully built. The <tt>EXTENSION</tt> and <tt>EXTWARN</tt>
				148	severities are used for extensions to the language that Clang accepts. This
				149	means that Clang fully understands and can represent them in the AST, but we
				150	produce diagnostics to tell the user their code is non-portable. The difference
				151	is that the former are ignored by default, and the later warn by default. The
				152	<tt>WARNING</tt> severity is used for constructs that are valid in the currently
				153	selected source language but that are dubious in some way. The <tt>NOTE</tt>
				154	level is used to staple more information onto a previous diagnostics.</p>
				155
				156	<p>These <em>severities</em> are mapped into a smaller set (the
				157	Diagnostic::Level enum, {<tt>Ignored</tt>, <tt>Note</tt>, <tt>Warning</tt>,
				158	<tt>Error</tt> }) of output <em>levels</em> by the diagnostics subsystem based
				159	on various configuration options. For example, if the user specifies
				160	<tt>-pedantic</tt>, <tt>EXTENSION</tt> maps to <tt>Warning</tt>, if they specify
				161	<tt>-pedantic-errors</tt>, it turns into <tt>Error</tt>. Clang also internally
				162	supports a fully fine grained mapping mechanism that allows you to map any
				163	diagnostic that doesn't have <tt>ERRROR</tt> severity to any output level that
				164	you want. This is used to implement options like <tt>-Wunused_macros</tt>,
				165	<tt>-Wundef</tt> etc.</p>
				166
				167	<!-- ================= -->
				168	<h4>The Format String</h4>
				169	<!-- ================= -->
				170
				171	<p>The format string for the diagnostic is very simple, but it has some power.
				172	It takes the form of a string in English with markers that indicate where and
				173	how arguments to the diagnostic are inserted and formatted. For example, here
				174	are some simple format strings:</p>
				175
				176	<pre>
				177	"binary integer literals are an extension"
				178	"format string contains '\\0' within the string body"
				179	"more '<b>%%</b>' conversions than data arguments"
				180	"invalid operands to binary expression ('<b>%0</b>' and '<b>%1</b>')"
				181	"overloaded '<b>%0</b>' must be a <b>%select{unary\|binary\|unary or binary}2</b> operator"
				182	" (has <b>%1</b> parameter<b>%s1</b>)"
				183	</pre>
				184
				185	<p>These examples show some important points of format strings. You can use any
				186	plain ASCII character in the diagnostic string except "%" without a problem,
				187	but these are C strings, so you have to use and be aware of all the C escape
				188	sequences (as in the second example). If you want to produce a "%" in the
				189	output, use the "%%" escape sequence, like the third diagnostic. Finally,
				190	clang uses the "%...[digit]" sequences to specify where and how arguments to
				191	the diagnostic are formatted.</p>
				192
				193	<p>Arguments to the diagnostic are numbered according to how they are specified
				194	by the C++ code that <a href="#producingdiag">produces them</a>, and are
				195	referenced by <tt>%0</tt> .. <tt>%9</tt>. If you have more than 10 arguments
				196	to your diagnostic, you are doing something wrong. :). Unlike printf, there
				197	is no requirement that arguments to the diagnostic end up in the output in
				198	the same order as they are specified, you could have a format string with
				199	<tt>"%1 %0"</tt> that swaps them, for example. The text in between the
				200	percent and digit are formatting instructions. If there are no instructions,
				201	the argument is just turned into a string and substituted in.</p>
				202
				203	<p>Here are some "best practices" for writing the English format string:</p>
				204
				205	<ul>
				206	<li>Keep the string short. It should ideally fit in the 80 column limit of the
				207	<tt>DiagnosticKinds.def</tt> file. This avoids the diagnostic wrapping when
				208	printed, and forces you to think about the important point you are conveying
				209	with the diagnostic.</li>
				210	<li>Take advantage of location information. The user will be able to see the
				211	line and location of the caret, so you don't need to tell them that the
				212	problem is with the 4th argument to the function: just point to it.</li>
				213	<li>Do not capitalize the diagnostic string, and do not end it with a
				214	period.</li>
				215	<li>If you need to quote something in the diagnostic string, use single
				216	quotes.</li>
				217	</ul>
				218
				219	<p>Diagnostics should never take random English strings as arguments: you
				220	shouldn't use <tt>"you have a problem with %0"</tt> and pass in things like
				221	<tt>"your argument"</tt> or <tt>"your return value"</tt> as arguments. Doing
				222	this prevents <a href="translation">translating</a> the Clang diagnostics to
				223	other languages (because they'll get random English words in their otherwise
				224	localized diagnostic). The exceptions to this are C/C++ language keywords
				225	(e.g. auto, const, mutable, etc) and C/C++ operators (<tt>/=</tt>). Note
				226	that things like "pointer" and "reference" are not keywords. On the other
				227	hand, you <em>can</em> include anything that comes from the user's source code,
				228	including variable names, types, labels, etc.</p>
				229
				230	<!-- ==================================== -->
				231	<h4>Formatting a Diagnostic Argument</a></h4>
				232	<!-- ==================================== -->
				233
				234	<p>Arguments to diagnostics are fully typed internally, and come from a couple
				235	different classes: integers, types, names, and random strings. Depending on
				236	the class of the argument, it can be optionally formatted in different ways.
				237	This gives the DiagnosticClient information about what the argument means
				238	without requiring it to use a specific presentation (consider this MVC for
				239	Clang :).</p>
				240
				241	<p>Here are the different diagnostic argument formats currently supported by
				242	Clang:</p>
				243
				244	<table>
				245	<tr><td colspan="2"><b>"s" format</b></td></tr>
				246	<tr><td>Example:</td><td><tt>"requires %1 parameter%s1"</tt></td></tr>
				247	<tr><td>Classes:</td><td>Integers</td></tr>
				248	<tr><td>Description:</td><td>This is a simple formatter for integers that is
				249	useful when producing English diagnostics. When the integer is 1, it prints
				250	as nothing. When the integer is not 1, it prints as "s". This allows some
				251	simple grammar to be to be handled correctly, and eliminates the need to use
Chris Lattner	cc54334	2008-11-22 23:50:47 +0000	[diff] [blame^]	252	gross things like <tt>"requires %1 parameter(s)"</tt>.</td></tr>
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	253
				254	<tr><td colspan="2"><b>"select" format</b></td></tr>
				255	<tr><td>Example:</td><td><tt>"must be a %select{unary\|binary\|unary or binary}2
				256	operator"</tt></td></tr>
				257	<tr><td>Classes:</td><td>Integers</td></tr>
Chris Lattner	cc54334	2008-11-22 23:50:47 +0000	[diff] [blame^]	258	<tr><td>Description:</td><td>This format specifier is used to merge multiple
				259	related diagnostics together into one common one, without requiring the
				260	different to be specified as an English string argument. Instead of
				261	specifying the string, the diagnostic gets an integer argument and the
				262	format string selects the numbered option. In this case, the "%2" value
				263	must be an integer in the range [0..2]. If it is 0, it prints 'unary', if
				264	it is 1 it prints 'binary' if it is 2, it prints 'unary or binary'. This
				265	allows other language translations to substitute reasonable words (or entire
				266	phrases) based on the semantics of the diagnostic instead of having to do
				267	things textually.</td></tr>
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	268
				269	<tr><td colspan="2"><b>"plural" format</b></td></tr>
Sebastian Redl	6816856	2008-11-22 22:16:45 +0000	[diff] [blame]	270	<tr><td>Example:</td><td><tt>"you have %1 %plural{1:mouse\|:mice}1 connected to
				271	your computer"</tt></td></tr>
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	272	<tr><td>Classes:</td><td>Integers</td></tr>
Sebastian Redl	6816856	2008-11-22 22:16:45 +0000	[diff] [blame]	273	<tr><td>Description:</td><td><p>This is a formatter for complex plural forms.
				274	It is designed to handle even the requirements of languages with very
				275	complex plural forms, as many Baltic languages have. The argument consists
				276	of a series of expression/form pairs, separated by ':', where the first form
				277	whose expression evaluates to true is the result of the modifier.</p>
				278	<p>An expression can be empty, in which case it is always true. See the
				279	example at the top. Otherwise, it is a series of one or more numeric
				280	conditions, separated by ','. If any condition matches, the expression
				281	matches. Each numeric condition can take one of three forms.</p>
				282	<ul>
				283	<li>number: A simple decimal number matches if the argument is the same
				284	as the number. Example: <tt>"{1:mouse\|:mice}"</tt></li>
				285	<li>range: A range in square brackets matches if the argument is within
				286	the range. Then range is inclusive both ends. Example:
				287	<tt>"{0:none\|1:one\|[2,5]:some\|:many}"</tt></li>
				288	<li>modulo: A modulo operator is followed by a number, and equals sign
				289	and either a number or a range. The tests are the same as for plain
				290	numbers and ranges, but the argument is taken modulo the number first.
				291	Example: <tt>"{%100=0:even hundred\|%100=[1,50]:lower half\|:everything
				292	else}"</tt></li>
				293	</ul>
				294	<p>The parser is very unforgiving. A syntax error, even whitespace, will
				295	abort, as will a failure to match the argument against any
				296	expression.</p></td></tr>
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	297
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	298	</table>
				299
Chris Lattner	cc54334	2008-11-22 23:50:47 +0000	[diff] [blame^]	300	<p>It is really easy to add format specifiers to the Clang diagnostics system,
				301	but they should be discussed before they are added. If you're creating a lot
				302	of repetitive diagnostics and/or have an idea for a useful formater, please
				303	bring it up on the cfe-dev mainling list.</p>
				304
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	305
				306
				307
				308	<!-- ===================================================== -->
				309	<h4><a name="#producingdiag">Producing the Diagnostic</a></h4>
				310	<!-- ===================================================== -->
				311
				312	<p>SemaExpr.cpp example</p>
				313
				314
				315	<!-- ============================================================= -->
				316	<h4><a name="DiagnosticClient">The DiagnosticClient Interface</a></h4>
				317	<!-- ============================================================= -->
				318
				319	<p>Clang command line, buffering, HTMLizing, etc.</p>
				320
				321	<!-- ====================================================== -->
				322	<h4><a name="translation">Adding Translations to Clang</a></h4>
				323	<!-- ====================================================== -->
				324
				325	<p>Not possible yet!</p>
				326
				327
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	328	<!-- ======================================================================= -->
				329	<h3 id="SourceLocation">The SourceLocation and SourceManager classes</h3>
				330	<!-- ======================================================================= -->
				331
				332	<p>Strangely enough, the SourceLocation class represents a location within the
				333	source code of the program. Important design points include:</p>
				334
				335	<ol>
				336	<li>sizeof(SourceLocation) must be extremely small, as these are embedded into
				337	many AST nodes and are passed around often. Currently it is 32 bits.</li>
				338	<li>SourceLocation must be a simple value object that can be efficiently
				339	copied.</li>
				340	<li>We should be able to represent a source location for any byte of any input
				341	file. This includes in the middle of tokens, in whitespace, in trigraphs,
				342	etc.</li>
				343	<li>A SourceLocation must encode the current #include stack that was active when
				344	the location was processed. For example, if the location corresponds to a
				345	token, it should contain the set of #includes active when the token was
				346	lexed. This allows us to print the #include stack for a diagnostic.</li>
				347	<li>SourceLocation must be able to describe macro expansions, capturing both
				348	the ultimate instantiation point and the source of the original character
				349	data.</li>
				350	</ol>
				351
				352	<p>In practice, the SourceLocation works together with the SourceManager class
				353	to encode two pieces of information about a location: it's physical location
				354	and it's virtual location. For most tokens, these will be the same. However,
				355	for a macro expansion (or tokens that came from a _Pragma directive) these will
				356	describe the location of the characters corresponding to the token and the
				357	location where the token was used (i.e. the macro instantiation point or the
				358	location of the _Pragma itself).</p>
				359
				360	<p>For efficiency, we only track one level of macro instantions: if a token was
				361	produced by multiple instantiations, we only track the source and ultimate
				362	destination. Though we could track the intermediate instantiation points, this
				363	would require extra bookkeeping and no known client would benefit substantially
				364	from this.</p>
				365
				366	<p>The clang front-end inherently depends on the location of a token being
				367	tracked correctly. If it is ever incorrect, the front-end may get confused and
				368	die. The reason for this is that the notion of the 'spelling' of a Token in
				369	clang depends on being able to find the original input characters for the token.
				370	This concept maps directly to the "physical" location for the token.</p>
				371
				372	<!-- ======================================================================= -->
				373	<h2 id="liblex">The Lexer and Preprocessor Library</h2>
				374	<!-- ======================================================================= -->
				375
				376	<p>The Lexer library contains several tightly-connected classes that are involved
				377	with the nasty process of lexing and preprocessing C source code. The main
				378	interface to this library for outside clients is the large <a
				379	href="#Preprocessor">Preprocessor</a> class.
				380	It contains the various pieces of state that are required to coherently read
				381	tokens out of a translation unit.</p>
				382
				383	<p>The core interface to the Preprocessor object (once it is set up) is the
				384	Preprocessor::Lex method, which returns the next <a href="#Token">Token</a> from
				385	the preprocessor stream. There are two types of token providers that the
				386	preprocessor is capable of reading from: a buffer lexer (provided by the <a
				387	href="#Lexer">Lexer</a> class) and a buffered token stream (provided by the <a
Chris Lattner	7928125	2008-03-09 02:27:26 +0000	[diff] [blame]	388	href="#TokenLexer">TokenLexer</a> class).
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	389
				390
				391	<!-- ======================================================================= -->
				392	<h3 id="Token">The Token class</h3>
				393	<!-- ======================================================================= -->
				394
				395	<p>The Token class is used to represent a single lexed token. Tokens are
				396	intended to be used by the lexer/preprocess and parser libraries, but are not
				397	intended to live beyond them (for example, they should not live in the ASTs).<p>
				398
				399	<p>Tokens most often live on the stack (or some other location that is efficient
				400	to access) as the parser is running, but occasionally do get buffered up. For
				401	example, macro definitions are stored as a series of tokens, and the C++
				402	front-end will eventually need to buffer tokens up for tentative parsing and
				403	various pieces of look-ahead. As such, the size of a Token matter. On a 32-bit
				404	system, sizeof(Token) is currently 16 bytes.</p>
				405
				406	<p>Tokens contain the following information:</p>
				407
				408	<ul>
				409	<li><b>A SourceLocation</b> - This indicates the location of the start of the
				410	token.</li>
				411
				412	<li><b>A length</b> - This stores the length of the token as stored in the
				413	SourceBuffer. For tokens that include them, this length includes trigraphs and
				414	escaped newlines which are ignored by later phases of the compiler. By pointing
				415	into the original source buffer, it is always possible to get the original
				416	spelling of a token completely accurately.</li>
				417
				418	<li><b>IdentifierInfo</b> - If a token takes the form of an identifier, and if
				419	identifier lookup was enabled when the token was lexed (e.g. the lexer was not
				420	reading in 'raw' mode) this contains a pointer to the unique hash value for the
				421	identifier. Because the lookup happens before keyword identification, this
				422	field is set even for language keywords like 'for'.</li>
				423
				424	<li><b>TokenKind</b> - This indicates the kind of token as classified by the
				425	lexer. This includes things like <tt>tok::starequal</tt> (for the "*="
				426	operator), <tt>tok::ampamp</tt> for the "&&" token, and keyword values
				427	(e.g. <tt>tok::kw_for</tt>) for identifiers that correspond to keywords. Note
				428	that some tokens can be spelled multiple ways. For example, C++ supports
				429	"operator keywords", where things like "and" are treated exactly like the
				430	"&&" operator. In these cases, the kind value is set to
				431	<tt>tok::ampamp</tt>, which is good for the parser, which doesn't have to
				432	consider both forms. For something that cares about which form is used (e.g.
				433	the preprocessor 'stringize' operator) the spelling indicates the original
				434	form.</li>
				435
				436	<li><b>Flags</b> - There are currently four flags tracked by the
				437	lexer/preprocessor system on a per-token basis:
				438
				439	<ol>
				440	<li><b>StartOfLine</b> - This was the first token that occurred on its input
				441	source line.</li>
				442	<li><b>LeadingSpace</b> - There was a space character either immediately
				443	before the token or transitively before the token as it was expanded
				444	through a macro. The definition of this flag is very closely defined by
				445	the stringizing requirements of the preprocessor.</li>
				446	<li><b>DisableExpand</b> - This flag is used internally to the preprocessor to
				447	represent identifier tokens which have macro expansion disabled. This
				448	prevents them from being considered as candidates for macro expansion ever
				449	in the future.</li>
				450	<li><b>NeedsCleaning</b> - This flag is set if the original spelling for the
				451	token includes a trigraph or escaped newline. Since this is uncommon,
				452	many pieces of code can fast-path on tokens that did not need cleaning.
				453	</p>
				454	</ol>
				455	</li>
				456	</ul>
				457
				458	<p>One interesting (and somewhat unusual) aspect of tokens is that they don't
				459	contain any semantic information about the lexed value. For example, if the
				460	token was a pp-number token, we do not represent the value of the number that
				461	was lexed (this is left for later pieces of code to decide). Additionally, the
				462	lexer library has no notion of typedef names vs variable names: both are
				463	returned as identifiers, and the parser is left to decide whether a specific
				464	identifier is a typedef or a variable (tracking this requires scope information
				465	among other things).</p>
				466
				467	<!-- ======================================================================= -->
				468	<h3 id="Lexer">The Lexer class</h3>
				469	<!-- ======================================================================= -->
				470
				471	<p>The Lexer class provides the mechanics of lexing tokens out of a source
				472	buffer and deciding what they mean. The Lexer is complicated by the fact that
				473	it operates on raw buffers that have not had spelling eliminated (this is a
				474	necessity to get decent performance), but this is countered with careful coding
				475	as well as standard performance techniques (for example, the comment handling
				476	code is vectorized on X86 and PowerPC hosts).</p>
				477
				478	<p>The lexer has a couple of interesting modal features:</p>
				479
				480	<ul>
				481	<li>The lexer can operate in 'raw' mode. This mode has several features that
				482	make it possible to quickly lex the file (e.g. it stops identifier lookup,
				483	doesn't specially handle preprocessor tokens, handles EOF differently, etc).
				484	This mode is used for lexing within an "<tt>#if 0</tt>" block, for
				485	example.</li>
				486	<li>The lexer can capture and return comments as tokens. This is required to
				487	support the -C preprocessor mode, which passes comments through, and is
				488	used by the diagnostic checker to identifier expect-error annotations.</li>
				489	<li>The lexer can be in ParsingFilename mode, which happens when preprocessing
Chris Lattner	8438624	2007-09-16 19:25:23 +0000	[diff] [blame]	490	after reading a #include directive. This mode changes the parsing of '<'
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	491	to return an "angled string" instead of a bunch of tokens for each thing
				492	within the filename.</li>
				493	<li>When parsing a preprocessor directive (after "<tt>#</tt>") the
				494	ParsingPreprocessorDirective mode is entered. This changes the parser to
				495	return EOM at a newline.</li>
				496	<li>The Lexer uses a LangOptions object to know whether trigraphs are enabled,
				497	whether C++ or ObjC keywords are recognized, etc.</li>
				498	</ul>
				499
				500	<p>In addition to these modes, the lexer keeps track of a couple of other
				501	features that are local to a lexed buffer, which change as the buffer is
				502	lexed:</p>
				503
				504	<ul>
				505	<li>The Lexer uses BufferPtr to keep track of the current character being
				506	lexed.</li>
				507	<li>The Lexer uses IsAtStartOfLine to keep track of whether the next lexed token
				508	will start with its "start of line" bit set.</li>
				509	<li>The Lexer keeps track of the current #if directives that are active (which
				510	can be nested).</li>
				511	<li>The Lexer keeps track of an <a href="#MultipleIncludeOpt">
				512	MultipleIncludeOpt</a> object, which is used to
				513	detect whether the buffer uses the standard "<tt>#ifndef XX</tt> /
				514	<tt>#define XX</tt>" idiom to prevent multiple inclusion. If a buffer does,
				515	subsequent includes can be ignored if the XX macro is defined.</li>
				516	</ul>
				517
				518	<!-- ======================================================================= -->
Chris Lattner	7928125	2008-03-09 02:27:26 +0000	[diff] [blame]	519	<h3 id="TokenLexer">The TokenLexer class</h3>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	520	<!-- ======================================================================= -->
				521
Chris Lattner	7928125	2008-03-09 02:27:26 +0000	[diff] [blame]	522	<p>The TokenLexer class is a token provider that returns tokens from a list
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	523	of tokens that came from somewhere else. It typically used for two things: 1)
				524	returning tokens from a macro definition as it is being expanded 2) returning
				525	tokens from an arbitrary buffer of tokens. The later use is used by _Pragma and
				526	will most likely be used to handle unbounded look-ahead for the C++ parser.</p>
				527
				528	<!-- ======================================================================= -->
				529	<h3 id="MultipleIncludeOpt">The MultipleIncludeOpt class</h3>
				530	<!-- ======================================================================= -->
				531
				532	<p>The MultipleIncludeOpt class implements a really simple little state machine
				533	that is used to detect the standard "<tt>#ifndef XX</tt> / <tt>#define XX</tt>"
				534	idiom that people typically use to prevent multiple inclusion of headers. If a
				535	buffer uses this idiom and is subsequently #include'd, the preprocessor can
				536	simply check to see whether the guarding condition is defined or not. If so,
				537	the preprocessor can completely ignore the include of the header.</p>
				538
				539
				540
				541	<!-- ======================================================================= -->
				542	<h2 id="libparse">The Parser Library</h2>
				543	<!-- ======================================================================= -->
				544
				545	<!-- ======================================================================= -->
				546	<h2 id="libast">The AST Library</h2>
				547	<!-- ======================================================================= -->
				548
				549	<!-- ======================================================================= -->
				550	<h3 id="Type">The Type class and its subclasses</h3>
				551	<!-- ======================================================================= -->
				552
				553	<p>The Type class (and its subclasses) are an important part of the AST. Types
				554	are accessed through the ASTContext class, which implicitly creates and uniques
				555	them as they are needed. Types have a couple of non-obvious features: 1) they
				556	do not capture type qualifiers like const or volatile (See
				557	<a href="#QualType">QualType</a>), and 2) they implicitly capture typedef
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	558	information. Once created, types are immutable (unlike decls).</p>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	559
				560	<p>Typedefs in C make semantic analysis a bit more complex than it would
				561	be without them. The issue is that we want to capture typedef information
				562	and represent it in the AST perfectly, but the semantics of operations need to
				563	"see through" typedefs. For example, consider this code:</p>
				564
				565	<code>
				566	void func() {<br>
Bill Wendling	30d1775	2007-10-06 01:56:01 +0000	[diff] [blame]	567	typedef int foo;<br>
				568	foo X, *Y;<br>
				569	typedef foo* bar;<br>
				570	bar Z;<br>
				571	*X; <i>// error</i><br>
				572	**Y; <i>// error</i><br>
				573	**Z; <i>// error</i><br>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	574	}<br>
				575	</code>
				576
				577	<p>The code above is illegal, and thus we expect there to be diagnostics emitted
				578	on the annotated lines. In this example, we expect to get:</p>
				579
				580	<pre>
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	581	<b>test.c:6:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	582	*X; // error
				583	<font color="blue">^~</font>
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	584	<b>test.c:7:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	585	**Y; // error
				586	<font color="blue">^~~</font>
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	587	<b>test.c:8:1: error: indirection requires pointer operand ('foo' invalid)</b>
				588	**Z; // error
				589	<font color="blue">^~~</font>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	590	</pre>
				591
				592	<p>While this example is somewhat silly, it illustrates the point: we want to
				593	retain typedef information where possible, so that we can emit errors about
				594	"<tt>std::string</tt>" instead of "<tt>std::basic_string<char, std:...</tt>".
				595	Doing this requires properly keeping typedef information (for example, the type
				596	of "X" is "foo", not "int"), and requires properly propagating it through the
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	597	various operators (for example, the type of *Y is "foo", not "int"). In order
				598	to retain this information, the type of these expressions is an instance of the
				599	TypedefType class, which indicates that the type of these expressions is a
				600	typedef for foo.
				601	</p>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	602
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	603	<p>Representing types like this is great for diagnostics, because the
				604	user-specified type is always immediately available. There are two problems
				605	with this: first, various semantic checks need to make judgements about the
Chris Lattner	33fc68a	2007-07-31 18:54:50 +0000	[diff] [blame]	606	<em>actual structure</em> of a type, ignoring typdefs. Second, we need an
				607	efficient way to query whether two types are structurally identical to each
				608	other, ignoring typedefs. The solution to both of these problems is the idea of
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	609	canonical types.</p>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	610
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	611	<!-- =============== -->
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	612	<h4>Canonical Types</h4>
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	613	<!-- =============== -->
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	614
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	615	<p>Every instance of the Type class contains a canonical type pointer. For
				616	simple types with no typedefs involved (e.g. "<tt>int</tt>", "<tt>int*</tt>",
				617	"<tt>int**</tt>"), the type just points to itself. For types that have a
				618	typedef somewhere in their structure (e.g. "<tt>foo</tt>", "<tt>foo*</tt>",
				619	"<tt>foo**</tt>", "<tt>bar</tt>"), the canonical type pointer points to their
				620	structurally equivalent type without any typedefs (e.g. "<tt>int</tt>",
				621	"<tt>int</tt>", "<tt>int</tt>", and "<tt>int</tt>" respectively).</p>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	622
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	623	<p>This design provides a constant time operation (dereferencing the canonical
				624	type pointer) that gives us access to the structure of types. For example,
				625	we can trivially tell that "bar" and "foo*" are the same type by dereferencing
				626	their canonical type pointers and doing a pointer comparison (they both point
				627	to the single "<tt>int*</tt>" type).</p>
				628
				629	<p>Canonical types and typedef types bring up some complexities that must be
				630	carefully managed. Specifically, the "isa/cast/dyncast" operators generally
				631	shouldn't be used in code that is inspecting the AST. For example, when type
				632	checking the indirection operator (unary '*' on a pointer), the type checker
				633	must verify that the operand has a pointer type. It would not be correct to
				634	check that with "<tt>isa<PointerType>(SubExpr->getType())</tt>",
				635	because this predicate would fail if the subexpression had a typedef type.</p>
				636
				637	<p>The solution to this problem are a set of helper methods on Type, used to
				638	check their properties. In this case, it would be correct to use
				639	"<tt>SubExpr->getType()->isPointerType()</tt>" to do the check. This
				640	predicate will return true if the <em>canonical type is a pointer</em>, which is
				641	true any time the type is structurally a pointer type. The only hard part here
				642	is remembering not to use the <tt>isa/cast/dyncast</tt> operations.</p>
				643
				644	<p>The second problem we face is how to get access to the pointer type once we
				645	know it exists. To continue the example, the result type of the indirection
				646	operator is the pointee type of the subexpression. In order to determine the
				647	type, we need to get the instance of PointerType that best captures the typedef
				648	information in the program. If the type of the expression is literally a
				649	PointerType, we can return that, otherwise we have to dig through the
				650	typedefs to find the pointer type. For example, if the subexpression had type
				651	"<tt>foo*</tt>", we could return that type as the result. If the subexpression
				652	had type "<tt>bar</tt>", we want to return "<tt>foo*</tt>" (note that we do
				653	<em>not</em> want "<tt>int*</tt>"). In order to provide all of this, Type has
Chris Lattner	11406c1	2007-07-31 16:50:51 +0000	[diff] [blame]	654	a getAsPointerType() method that checks whether the type is structurally a
Chris Lattner	8a2bc62	2007-07-31 06:37:39 +0000	[diff] [blame]	655	PointerType and, if so, returns the best one. If not, it returns a null
				656	pointer.</p>
				657
				658	<p>This structure is somewhat mystical, but after meditating on it, it will
				659	make sense to you :).</p>
Chris Lattner	86920d3	2007-07-31 05:42:17 +0000	[diff] [blame]	660
				661	<!-- ======================================================================= -->
				662	<h3 id="QualType">The QualType class</h3>
				663	<!-- ======================================================================= -->
				664
				665	<p>The QualType class is designed as a trivial value class that is small,
				666	passed by-value and is efficient to query. The idea of QualType is that it
				667	stores the type qualifiers (const, volatile, restrict) separately from the types
				668	themselves: QualType is conceptually a pair of "Type*" and bits for the type
				669	qualifiers.</p>
				670
				671	<p>By storing the type qualifiers as bits in the conceptual pair, it is
				672	extremely efficient to get the set of qualifiers on a QualType (just return the
				673	field of the pair), add a type qualifier (which is a trivial constant-time
				674	operation that sets a bit), and remove one or more type qualifiers (just return
				675	a QualType with the bitfield set to empty).</p>
				676
				677	<p>Further, because the bits are stored outside of the type itself, we do not
				678	need to create duplicates of types with different sets of qualifiers (i.e. there
				679	is only a single heap allocated "int" type: "const int" and "volatile const int"
				680	both point to the same heap allocated "int" type). This reduces the heap size
				681	used to represent bits and also means we do not have to consider qualifiers when
				682	uniquing types (<a href="#Type">Type</a> does not even contain qualifiers).</p>
				683
				684	<p>In practice, on hosts where it is safe, the 3 type qualifiers are stored in
				685	the low bit of the pointer to the Type object. This means that QualType is
				686	exactly the same size as a pointer, and this works fine on any system where
				687	malloc'd objects are at least 8 byte aligned.</p>
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	688
				689	<!-- ======================================================================= -->
Douglas Gregor	2e1cd42	2008-11-17 14:58:09 +0000	[diff] [blame]	690	<h3 id="DeclarationName">Declaration names</h3>
				691	<!-- ======================================================================= -->
				692
				693	<p>The <tt>DeclarationName</tt> class represents the name of a
				694	declaration in Clang. Declarations in the C family of languages can
				695	take several different forms. Most declarations are named by are
				696	simple identifiers, e.g., "<code>f</code>" and "<code>x</code>" in
				697	the function declaration <code>f(int x)</code>. In C++, declaration
				698	names can also name class constructors ("<code>Class</code>"
				699	in <code>struct Class { Class(); }</code>), class destructors
				700	("<code>~Class</code>"), overloaded operator names ("operator+"),
				701	and conversion functions ("<code>operator void const *</code>"). In
				702	Objective-C, declaration names can refer to the names of Objective-C
				703	methods, which involve the method name and the parameters,
				704	collectively called a <i>selector</i>, e.g..,
				705	"<code>setWidth:height:</code>". Since all of these kinds of
				706	entities--variables, functions, Objective-C methods, C++
				707	constructors, destructors, and operators---are represented as
				708	subclasses of Clang's common <code>NamedDecl</code>
				709	class, <code>DeclarationName</code> is designed to efficiently
				710	represent any kind of name.</p>
				711
				712	<p>Given
				713	a <code>DeclarationName</code> <code>N</code>, <code>N.getNameKind()</code>
Douglas Gregor	2def483	2008-11-17 20:34:05 +0000	[diff] [blame]	714	will produce a value that describes what kind of name <code>N</code>
Douglas Gregor	e94ca9e4	2008-11-18 14:39:36 +0000	[diff] [blame]	715	stores. There are 8 options (all of the names are inside
Douglas Gregor	2e1cd42	2008-11-17 14:58:09 +0000	[diff] [blame]	716	the <code>DeclarationName</code> class)</p>
				717	<dl>
				718	<dt>Identifier</dt>
				719	<dd>The name is a simple
				720	identifier. Use <code>N.getAsIdentifierInfo()</code> to retrieve the
				721	corresponding <code>IdentifierInfo*</code> pointing to the actual
				722	identifier. Note that C++ overloaded operators (e.g.,
				723	"<code>operator+</code>") are represented as special kinds of
				724	identifiers. Use <code>IdentifierInfo</code>'s <code>getOverloadedOperatorID</code>
				725	function to determine whether an identifier is an overloaded
				726	operator name.</dd>
				727
				728	<dt>ObjCZeroArgSelector, ObjCOneArgSelector,
				729	ObjCMultiArgSelector</dt>
				730	<dd>The name is an Objective-C selector, which can be retrieved as a
				731	<code>Selector</code> instance
				732	via <code>N.getObjCSelector()</code>. The three possible name
				733	kinds for Objective-C reflect an optimization within
				734	the <code>DeclarationName</code> class: both zero- and
				735	one-argument selectors are stored as a
				736	masked <code>IdentifierInfo</code> pointer, and therefore require
				737	very little space, since zero- and one-argument selectors are far
				738	more common than multi-argument selectors (which use a different
				739	structure).</dd>
				740
				741	<dt>CXXConstructorName</dt>
				742	<dd>The name is a C++ constructor
				743	name. Use <code>N.getCXXNameType()</code> to retrieve
				744	the <a href="#QualType">type</a> that this constructor is meant to
				745	construct. The type is always the canonical type, since all
				746	constructors for a given type have the same name.</dd>
				747
				748	<dt>CXXDestructorName</dt>
				749	<dd>The name is a C++ destructor
				750	name. Use <code>N.getCXXNameType()</code> to retrieve
				751	the <a href="#QualType">type</a> whose destructor is being
				752	named. This type is always a canonical type.</dd>
				753
				754	<dt>CXXConversionFunctionName</dt>
				755	<dd>The name is a C++ conversion function. Conversion functions are
				756	named according to the type they convert to, e.g., "<code>operator void
				757	const *</code>". Use <code>N.getCXXNameType()</code> to retrieve
				758	the type that this conversion function converts to. This type is
				759	always a canonical type.</dd>
Douglas Gregor	e94ca9e4	2008-11-18 14:39:36 +0000	[diff] [blame]	760
				761	<dt>CXXOperatorName</dt>
				762	<dd>The name is a C++ overloaded operator name. Overloaded operators
				763	are named according to their spelling, e.g.,
				764	"<code>operator+</code>" or "<code>operator new
				765	[]</code>". Use <code>N.getCXXOverloadedOperator()</code> to
				766	retrieve the overloaded operator (a value of
				767	type <code>OverloadedOperatorKind</code>).</dd>
Douglas Gregor	2e1cd42	2008-11-17 14:58:09 +0000	[diff] [blame]	768	</dl>
				769
				770	<p><code>DeclarationName</code>s are cheap to create, copy, and
				771	compare. They require only a single pointer's worth of storage in
Douglas Gregor	e94ca9e4	2008-11-18 14:39:36 +0000	[diff] [blame]	772	the common cases (identifiers, zero-
Douglas Gregor	2e1cd42	2008-11-17 14:58:09 +0000	[diff] [blame]	773	and one-argument Objective-C selectors) and use dense, uniqued
				774	storage for the other kinds of
				775	names. Two <code>DeclarationName</code>s can be compared for
				776	equality (<code>==</code>, <code>!=</code>) using a simple bitwise
				777	comparison, can be ordered
				778	with <code><</code>, <code>></code>, <code><=</code>,
				779	and <code>>=</code> (which provide a lexicographical ordering for
				780	normal identifiers but an unspecified ordering for other kinds of
				781	names), and can be placed into LLVM <code>DenseMap</code>s
				782	and <code>DenseSet</code>s.</p>
				783
				784	<p><code>DeclarationName</code> instances can be created in different
				785	ways depending on what kind of name the instance will store. Normal
Douglas Gregor	e94ca9e4	2008-11-18 14:39:36 +0000	[diff] [blame]	786	identifiers (<code>IdentifierInfo</code> pointers) and Objective-C selectors
Douglas Gregor	2e1cd42	2008-11-17 14:58:09 +0000	[diff] [blame]	787	(<code>Selector</code>) can be implicitly converted
				788	to <code>DeclarationName</code>s. Names for C++ constructors,
Douglas Gregor	e94ca9e4	2008-11-18 14:39:36 +0000	[diff] [blame]	789	destructors, conversion functions, and overloaded operators can be retrieved from
Douglas Gregor	2e1cd42	2008-11-17 14:58:09 +0000	[diff] [blame]	790	the <code>DeclarationNameTable</code>, an instance of which is
				791	available as <code>ASTContext::DeclarationNames</code>. The member
				792	functions <code>getCXXConstructorName</code>, <code>getCXXDestructorName</code>,
Douglas Gregor	e94ca9e4	2008-11-18 14:39:36 +0000	[diff] [blame]	793	<code>getCXXConversionFunctionName</code>, and <code>getCXXOperatorName</code>, respectively,
				794	return <code>DeclarationName</code> instances for the four kinds of
Douglas Gregor	2e1cd42	2008-11-17 14:58:09 +0000	[diff] [blame]	795	C++ special function names.</p>
				796
				797	<!-- ======================================================================= -->
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	798	<h3 id="CFG">The <tt>CFG</tt> class</h3>
				799	<!-- ======================================================================= -->
				800
				801	<p>The <tt>CFG</tt> class is designed to represent a source-level
				802	control-flow graph for a single statement (<tt>Stmt*</tt>). Typically
				803	instances of <tt>CFG</tt> are constructed for function bodies (usually
				804	an instance of <tt>CompoundStmt</tt>), but can also be instantiated to
				805	represent the control-flow of any class that subclasses <tt>Stmt</tt>,
				806	which includes simple expressions. Control-flow graphs are especially
				807	useful for performing
				808	<a href="http://en.wikipedia.org/wiki/Data_flow_analysis#Sensitivities">flow-
				809	or path-sensitive</a> program analyses on a given function.</p>
				810
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	811	<!-- ============ -->
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	812	<h4>Basic Blocks</h4>
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	813	<!-- ============ -->
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	814
				815	<p>Concretely, an instance of <tt>CFG</tt> is a collection of basic
				816	blocks. Each basic block is an instance of <tt>CFGBlock</tt>, which
				817	simply contains an ordered sequence of <tt>Stmt*</tt> (each referring
				818	to statements in the AST). The ordering of statements within a block
				819	indicates unconditional flow of control from one statement to the
				820	next. <a href="#ConditionalControlFlow">Conditional control-flow</a>
				821	is represented using edges between basic blocks. The statements
				822	within a given <tt>CFGBlock</tt> can be traversed using
				823	the <tt>CFGBlock::*iterator</tt> interface.</p>
				824
				825	<p>
Ted Kremenek	18e17e7	2007-10-18 22:50:52 +0000	[diff] [blame]	826	A <tt>CFG</tt> object owns the instances of <tt>CFGBlock</tt> within
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	827	the control-flow graph it represents. Each <tt>CFGBlock</tt> within a
				828	CFG is also uniquely numbered (accessible
				829	via <tt>CFGBlock::getBlockID()</tt>). Currently the number is
				830	based on the ordering the blocks were created, but no assumptions
				831	should be made on how <tt>CFGBlock</tt>s are numbered other than their
				832	numbers are unique and that they are numbered from 0..N-1 (where N is
				833	the number of basic blocks in the CFG).</p>
				834
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	835	<!-- ===================== -->
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	836	<h4>Entry and Exit Blocks</h4>
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	837	<!-- ===================== -->
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	838
				839	Each instance of <tt>CFG</tt> contains two special blocks:
				840	an <i>entry</i> block (accessible via <tt>CFG::getEntry()</tt>), which
				841	has no incoming edges, and an <i>exit</i> block (accessible
				842	via <tt>CFG::getExit()</tt>), which has no outgoing edges. Neither
				843	block contains any statements, and they serve the role of providing a
				844	clear entrance and exit for a body of code such as a function body.
				845	The presence of these empty blocks greatly simplifies the
				846	implementation of many analyses built on top of CFGs.
				847
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	848	<!-- ===================================================== -->
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	849	<h4 id ="ConditionalControlFlow">Conditional Control-Flow</h4>
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	850	<!-- ===================================================== -->
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	851
				852	<p>Conditional control-flow (such as those induced by if-statements
				853	and loops) is represented as edges between <tt>CFGBlock</tt>s.
				854	Because different C language constructs can induce control-flow,
				855	each <tt>CFGBlock</tt> also records an extra <tt>Stmt*</tt> that
				856	represents the <i>terminator</i> of the block. A terminator is simply
				857	the statement that caused the control-flow, and is used to identify
				858	the nature of the conditional control-flow between blocks. For
				859	example, in the case of an if-statement, the terminator refers to
				860	the <tt>IfStmt</tt> object in the AST that represented the given
				861	branch.</p>
				862
				863	<p>To illustrate, consider the following code example:</p>
				864
				865	<code>
				866	int foo(int x) {<br>
				867	x = x + 1;<br>
				868	<br>
				869	if (x > 2) x++;<br>
				870	else {<br>
				871	x += 2;<br>
				872	x *= 2;<br>
				873	}<br>
				874	<br>
				875	return x;<br>
				876	}
				877	</code>
				878
				879	<p>After invoking the parser+semantic analyzer on this code fragment,
				880	the AST of the body of <tt>foo</tt> is referenced by a
				881	single <tt>Stmt*</tt>. We can then construct an instance
				882	of <tt>CFG</tt> representing the control-flow graph of this function
				883	body by single call to a static class method:</p>
				884
				885	<code>
				886	Stmt* FooBody = ...<br>
				887	CFG* FooCFG = <b>CFG::buildCFG</b>(FooBody);
				888	</code>
				889
				890	<p>It is the responsibility of the caller of <tt>CFG::buildCFG</tt>
				891	to <tt>delete</tt> the returned <tt>CFG*</tt> when the CFG is no
				892	longer needed.</p>
				893
				894	<p>Along with providing an interface to iterate over
				895	its <tt>CFGBlock</tt>s, the <tt>CFG</tt> class also provides methods
				896	that are useful for debugging and visualizing CFGs. For example, the
				897	method
				898	<tt>CFG::dump()</tt> dumps a pretty-printed version of the CFG to
				899	standard error. This is especially useful when one is using a
				900	debugger such as gdb. For example, here is the output
				901	of <tt>FooCFG->dump()</tt>:</p>
				902
				903	<code>
				904	[ B5 (ENTRY) ]<br>
				905	Predecessors (0):<br>
				906	Successors (1): B4<br>
				907	<br>
				908	[ B4 ]<br>
				909	1: x = x + 1<br>
				910	2: (x > 2)<br>
				911	<b>T: if [B4.2]</b><br>
				912	Predecessors (1): B5<br>
				913	Successors (2): B3 B2<br>
				914	<br>
				915	[ B3 ]<br>
				916	1: x++<br>
				917	Predecessors (1): B4<br>
				918	Successors (1): B1<br>
				919	<br>
				920	[ B2 ]<br>
				921	1: x += 2<br>
				922	2: x *= 2<br>
				923	Predecessors (1): B4<br>
				924	Successors (1): B1<br>
				925	<br>
				926	[ B1 ]<br>
				927	1: return x;<br>
				928	Predecessors (2): B2 B3<br>
				929	Successors (1): B0<br>
				930	<br>
				931	[ B0 (EXIT) ]<br>
				932	Predecessors (1): B1<br>
				933	Successors (0):
				934	</code>
				935
				936	<p>For each block, the pretty-printed output displays for each block
				937	the number of <i>predecessor</i> blocks (blocks that have outgoing
				938	control-flow to the given block) and <i>successor</i> blocks (blocks
				939	that have control-flow that have incoming control-flow from the given
				940	block). We can also clearly see the special entry and exit blocks at
				941	the beginning and end of the pretty-printed output. For the entry
				942	block (block B5), the number of predecessor blocks is 0, while for the
				943	exit block (block B0) the number of successor blocks is 0.</p>
				944
				945	<p>The most interesting block here is B4, whose outgoing control-flow
				946	represents the branching caused by the sole if-statement
				947	in <tt>foo</tt>. Of particular interest is the second statement in
				948	the block, <b><tt>(x > 2)</tt></b>, and the terminator, printed
				949	as <b><tt>if [B4.2]</tt></b>. The second statement represents the
				950	evaluation of the condition of the if-statement, which occurs before
				951	the actual branching of control-flow. Within the <tt>CFGBlock</tt>
				952	for B4, the <tt>Stmt*</tt> for the second statement refers to the
				953	actual expression in the AST for <b><tt>(x > 2)</tt></b>. Thus
				954	pointers to subclasses of <tt>Expr</tt> can appear in the list of
				955	statements in a block, and not just subclasses of <tt>Stmt</tt> that
				956	refer to proper C statements.</p>
				957
				958	<p>The terminator of block B4 is a pointer to the <tt>IfStmt</tt>
				959	object in the AST. The pretty-printer outputs <b><tt>if
				960	[B4.2]</tt></b> because the condition expression of the if-statement
				961	has an actual place in the basic block, and thus the terminator is
				962	essentially
				963	<i>referring</i> to the expression that is the second statement of
				964	block B4 (i.e., B4.2). In this manner, conditions for control-flow
				965	(which also includes conditions for loops and switch statements) are
				966	hoisted into the actual basic block.</p>
				967
Chris Lattner	62fd278	2008-11-22 21:41:31 +0000	[diff] [blame]	968	<!-- ===================== -->
				969	<!-- <h4>Implicit Control-Flow</h4> -->
				970	<!-- ===================== -->
Ted Kremenek	8bc0571	2007-10-10 23:01:43 +0000	[diff] [blame]	971
				972	<!--
				973	<p>A key design principle of the <tt>CFG</tt> class was to not require
				974	any transformations to the AST in order to represent control-flow.
				975	Thus the <tt>CFG</tt> does not perform any "lowering" of the
				976	statements in an AST: loops are not transformed into guarded gotos,
				977	short-circuit operations are not converted to a set of if-statements,
				978	and so on.</p>
				979	-->
Ted Kremenek	17a295d	2008-06-11 06:19:49 +0000	[diff] [blame]	980
Chris Lattner	7bad199	2008-11-16 21:48:07 +0000	[diff] [blame]	981
				982	<!-- ======================================================================= -->
				983	<h3 id="Constants">Constant Folding in the Clang AST</h3>
				984	<!-- ======================================================================= -->
				985
				986	<p>There are several places where constants and constant folding matter a lot to
				987	the Clang front-end. First, in general, we prefer the AST to retain the source
				988	code as close to how the user wrote it as possible. This means that if they
				989	wrote "5+4", we want to keep the addition and two constants in the AST, we don't
				990	want to fold to "9". This means that constant folding in various ways turns
				991	into a tree walk that needs to handle the various cases.</p>
				992
				993	<p>However, there are places in both C and C++ that require constants to be
				994	folded. For example, the C standard defines what an "integer constant
				995	expression" (i-c-e) is with very precise and specific requirements. The
				996	language then requires i-c-e's in a lot of places (for example, the size of a
				997	bitfield, the value for a case statement, etc). For these, we have to be able
				998	to constant fold the constants, to do semantic checks (e.g. verify bitfield size
				999	is non-negative and that case statements aren't duplicated). We aim for Clang
				1000	to be very pedantic about this, diagnosing cases when the code does not use an
				1001	i-c-e where one is required, but accepting the code unless running with
				1002	<tt>-pedantic-errors</tt>.</p>
				1003
				1004	<p>Things get a little bit more tricky when it comes to compatibility with
				1005	real-world source code. Specifically, GCC has historically accepted a huge
				1006	superset of expressions as i-c-e's, and a lot of real world code depends on this
				1007	unfortuate accident of history (including, e.g., the glibc system headers). GCC
				1008	accepts anything its "fold" optimizer is capable of reducing to an integer
				1009	constant, which means that the definition of what it accepts changes as its
				1010	optimizer does. One example is that GCC accepts things like "case X-X:" even
				1011	when X is a variable, because it can fold this to 0.</p>
				1012
				1013	<p>Another issue are how constants interact with the extensions we support, such
				1014	as __builtin_constant_p, __builtin_inf, __extension__ and many others. C99
				1015	obviously does not specify the semantics of any of these extensions, and the
				1016	definition of i-c-e does not include them. However, these extensions are often
				1017	used in real code, and we have to have a way to reason about them.</p>
				1018
				1019	<p>Finally, this is not just a problem for semantic analysis. The code
				1020	generator and other clients have to be able to fold constants (e.g. to
				1021	initialize global variables) and has to handle a superset of what C99 allows.
				1022	Further, these clients can benefit from extended information. For example, we
				1023	know that "foo()\|\|1" always evaluates to true, but we can't replace the
				1024	expression with true because it has side effects.</p>
				1025
				1026	<!-- ======================= -->
				1027	<h4>Implementation Approach</h4>
				1028	<!-- ======================= -->
				1029
				1030	<p>After trying several different approaches, we've finally converged on a
				1031	design (Note, at the time of this writing, not all of this has been implemented,
				1032	consider this a design goal!). Our basic approach is to define a single
				1033	recursive method evaluation method (<tt>Expr::Evaluate</tt>), which is
				1034	implemented in <tt>AST/ExprConstant.cpp</tt>. Given an expression with 'scalar'
				1035	type (integer, fp, complex, or pointer) this method returns the following
				1036	information:</p>
				1037
				1038	<ul>
				1039	<li>Whether the expression is an integer constant expression, a general
				1040	constant that was folded but has no side effects, a general constant that
				1041	was folded but that does have side effects, or an uncomputable/unfoldable
				1042	value.
				1043	</li>
				1044	<li>If the expression was computable in any way, this method returns the APValue
				1045	for the result of the expression.</li>
				1046	<li>If the expression is not evaluatable at all, this method returns
				1047	information on one of the problems with the expression. This includes a
				1048	SourceLocation for where the problem is, and a diagnostic ID that explains
				1049	the problem. The diagnostic should be have ERROR type.</li>
				1050	<li>If the expression is not an integer constant expression, this method returns
				1051	information on one of the problems with the expression. This includes a
				1052	SourceLocation for where the problem is, and a diagnostic ID that explains
				1053	the problem. The diagnostic should be have EXTENSION type.</li>
				1054	</ul>
				1055
				1056	<p>This information gives various clients the flexibility that they want, and we
				1057	will eventually have some helper methods for various extensions. For example,
				1058	Sema should have a <tt>Sema::VerifyIntegerConstantExpression</tt> method, which
				1059	calls Evaluate on the expression. If the expression is not foldable, the error
				1060	is emitted, and it would return true. If the expression is not an i-c-e, the
				1061	EXTENSION diagnostic is emitted. Finally it would return false to indicate that
				1062	the AST is ok.</p>
				1063
				1064	<p>Other clients can use the information in other ways, for example, codegen can
				1065	just use expressions that are foldable in any way.</p>
				1066
				1067	<!-- ========== -->
				1068	<h4>Extensions</h4>
				1069	<!-- ========== -->
				1070
				1071	<p>This section describes how some of the various extensions clang supports
				1072	interacts with constant evaluation:</p>
				1073
				1074	<ul>
				1075	<li><b><tt>__extension__</tt></b>: The expression form of this extension causes
				1076	any evaluatable subexpression to be accepted as an integer constant
				1077	expression.</li>
				1078	<li><b><tt>__builtin_constant_p</tt></b>: This returns true (as a integer
				1079	constant expression) if the operand is any evaluatable constant.</li>
				1080	<li><b><tt>__builtin_choose_expr</tt></b>: The condition is required to be an
				1081	integer constant expression, but we accept any constant as an "extension of
				1082	an extension". This only evaluates one operand depending on which way the
				1083	condition evaluates.</li>
				1084	<li><b><tt>__builtin_classify_type</tt></b>: This always returns an integer
				1085	constant expression.</li>
				1086	<li><b><tt>__builtin_inf,nan,..</tt></b>: These are treated just like a
				1087	floating-point literal.</li>
				1088	<li><b><tt>__builtin_abs,copysign,..</tt></b>: These are constant folded as
				1089	general constant expressions.</li>
				1090	</ul>
				1091
				1092
				1093
				1094
Ted Kremenek	17a295d	2008-06-11 06:19:49 +0000	[diff] [blame]	1095	</div>
				1096	</body>
Douglas Gregor	2e1cd42	2008-11-17 14:58:09 +0000	[diff] [blame]	1097	</html>