Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / clang / eb666737bae8a5af8fb6b485cd15e2e62caa5caf / . / docs / InternalsManual.html
blob: 0df1e32c77499e0473296566b26650cc37881cf5 [file] [log] [blame]
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]1<html>
2<head>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]3<title>"Clang" CFE Internals Manual</title>
Ted Kremenek17a295d2008-06-11 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 Redl68168562008-11-22 22:16:45 +0000[diff] [blame]6<style type="text/css">
7td {
8 vertical-align: top;
9}
10</style>
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]11</head>
12<body>
13
14<!--#include virtual="../menu.html.incl"-->
15
16<div id="content">
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]17
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]18<h1>"Clang" CFE Internals Manual</h1>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]19
20<ul>
21<li><a href="#intro">Introduction</a></li>
Peter Collingbourne967c1182011-10-15 16:59:24 +0000[diff] [blame]22<li><a href="#libsupport">LLVM Support Library</a></li>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]23<li><a href="#libbasic">The Clang 'Basic' Library</a>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]24 <ul>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]25 <li><a href="#Diagnostics">The Diagnostics Subsystem</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]26 <li><a href="#SourceLocation">The SourceLocation and SourceManager
27 classes</a></li>
Douglas Gregor715c92a2010-10-27 16:02:28 +0000[diff] [blame]28 <li><a href="#SourceRange">SourceRange and CharSourceRange</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]29 </ul>
30</li>
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000[diff] [blame]31<li><a href="#libdriver">The Driver Library</a>
32 <ul>
33 </ul>
34</li>
Douglas Gregor32110df2009-05-20 00:16:32 +0000[diff] [blame]35<li><a href="#pch">Precompiled Headers</a>
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000[diff] [blame]36<li><a href="#libfrontend">The Frontend Library</a>
37 <ul>
38 </ul>
39</li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]40<li><a href="#liblex">The Lexer and Preprocessor Library</a>
41 <ul>
42 <li><a href="#Token">The Token class</a></li>
43 <li><a href="#Lexer">The Lexer class</a></li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]44 <li><a href="#AnnotationToken">Annotation Tokens</a></li>
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]45 <li><a href="#TokenLexer">The TokenLexer class</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]46 <li><a href="#MultipleIncludeOpt">The MultipleIncludeOpt class</a></li>
47 </ul>
48</li>
49<li><a href="#libparse">The Parser Library</a>
50 <ul>
51 </ul>
52</li>
53<li><a href="#libast">The AST Library</a>
54 <ul>
55 <li><a href="#Type">The Type class and its subclasses</a></li>
56 <li><a href="#QualType">The QualType class</a></li>
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]57 <li><a href="#DeclarationName">Declaration names</a></li>
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]58 <li><a href="#DeclContext">Declaration contexts</a>
59 <ul>
60 <li><a href="#Redeclarations">Redeclarations and Overloads</a></li>
61 <li><a href="#LexicalAndSemanticContexts">Lexical and Semantic
62 Contexts</a></li>
63 <li><a href="#TransparentContexts">Transparent Declaration Contexts</a></li>
64 <li><a href="#MultiDeclContext">Multiply-Defined Declaration Contexts</a></li>
65 </ul>
66 </li>
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]67 <li><a href="#CFG">The CFG class</a></li>
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]68 <li><a href="#Constants">Constant Folding in the Clang AST</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]69 </ul>
70</li>
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +0000[diff] [blame]71<li><a href="#Howtos">Howto guides</a>
72 <ul>
73 <li><a href="#AddingAttributes">How to add an attribute</a></li>
Douglas Gregor1f634c62011-09-30 21:32:37 +0000[diff] [blame]74 <li><a href="#AddingExprStmt">How to add a new expression or statement</a></li>
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +0000[diff] [blame]75 </ul>
76</li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]77</ul>
78
79
80<!-- ======================================================================= -->
81<h2 id="intro">Introduction</h2>
82<!-- ======================================================================= -->
83
84<p>This document describes some of the more important APIs and internal design
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]85decisions made in the Clang C front-end. The purpose of this document is to
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]86both capture some of this high level information and also describe some of the
87design decisions behind it. This is meant for people interested in hacking on
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]88Clang, not for end-users. The description below is categorized by
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]89libraries, and does not describe any of the clients of the libraries.</p>
90
91<!-- ======================================================================= -->
Peter Collingbourne967c1182011-10-15 16:59:24 +0000[diff] [blame]92<h2 id="libsupport">LLVM Support Library</h2>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]93<!-- ======================================================================= -->
94
Peter Collingbourne967c1182011-10-15 16:59:24 +0000[diff] [blame]95<p>The LLVM libsupport library provides many underlying libraries and
96<a href="http://llvm.org/docs/ProgrammersManual.html">data-structures</a>,
97including command line option processing, various containers and a system
98abstraction layer, which is used for file system access.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]99
100<!-- ======================================================================= -->
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]101<h2 id="libbasic">The Clang 'Basic' Library</h2>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]102<!-- ======================================================================= -->
103
104<p>This library certainly needs a better name. The 'basic' library contains a
105number of low-level utilities for tracking and manipulating source buffers,
106locations within the source buffers, diagnostics, tokens, target abstraction,
107and information about the subset of the language being compiled for.</p>
108
109<p>Part of this infrastructure is specific to C (such as the TargetInfo class),
110other parts could be reused for other non-C-based languages (SourceLocation,
111SourceManager, Diagnostics, FileManager). When and if there is future demand
112we can figure out if it makes sense to introduce a new library, move the general
113classes somewhere else, or introduce some other solution.</p>
114
115<p>We describe the roles of these classes in order of their dependencies.</p>
116
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]117
118<!-- ======================================================================= -->
119<h3 id="Diagnostics">The Diagnostics Subsystem</h3>
120<!-- ======================================================================= -->
121
122<p>The Clang Diagnostics subsystem is an important part of how the compiler
123communicates with the human. Diagnostics are the warnings and errors produced
124when the code is incorrect or dubious. In Clang, each diagnostic produced has
Sebastian Redl9bc2a992010-07-07 23:42:27 +0000[diff] [blame]125(at the minimum) a unique ID, an English translation associated with it, a <a
126href="#SourceLocation">SourceLocation</a> to "put the caret", and a severity (e.g.
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]127<tt>WARNING</tt> or <tt>ERROR</tt>). They can also optionally include a number
128of arguments to the dianostic (which fill in "%0"'s in the string) as well as a
129number of source ranges that related to the diagnostic.</p>
130
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]131<p>In this section, we'll be giving examples produced by the Clang command line
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]132driver, but diagnostics can be <a href="#DiagnosticClient">rendered in many
133different ways</a> depending on how the DiagnosticClient interface is
Sebastian Redl9bc2a992010-07-07 23:42:27 +0000[diff] [blame]134implemented. A representative example of a diagnostic is:</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]135
136<pre>
137t.c:38:15: error: invalid operands to binary expression ('int *' and '_Complex float')
138 <font color="darkgreen">P = (P-42) + Gamma*4;</font>
139 <font color="blue">~~~~~~ ^ ~~~~~~~</font>
140</pre>
141
142<p>In this example, you can see the English translation, the severity (error),
143you can see the source location (the caret ("^") and file/line/column info),
144the source ranges "~~~~", arguments to the diagnostic ("int*" and "_Complex
145float"). You'll have to believe me that there is a unique ID backing the
146diagnostic :).</p>
147
148<p>Getting all of this to happen has several steps and involves many moving
149pieces, this section describes them and talks about best practices when adding
150a new diagnostic.</p>
151
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000[diff] [blame]152<!-- ============================= -->
153<h4>The Diagnostic*Kinds.td files</h4>
154<!-- ============================= -->
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]155
Chris Lattner4c50b692010-05-01 17:35:19 +0000[diff] [blame]156<p>Diagnostics are created by adding an entry to one of the <tt>
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000[diff] [blame]157clang/Basic/Diagnostic*Kinds.td</tt> files, depending on what library will
158be using it. From this file, tblgen generates the unique ID of the diagnostic,
159the severity of the diagnostic and the English translation + format string.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]160
161<p>There is little sanity with the naming of the unique ID's right now. Some
162start with err_, warn_, ext_ to encode the severity into the name. Since the
163enum is referenced in the C++ code that produces the diagnostic, it is somewhat
164useful for it to be reasonably short.</p>
165
166<p>The severity of the diagnostic comes from the set {<tt>NOTE</tt>,
167<tt>WARNING</tt>, <tt>EXTENSION</tt>, <tt>EXTWARN</tt>, <tt>ERROR</tt>}. The
168<tt>ERROR</tt> severity is used for diagnostics indicating the program is never
169acceptable under any circumstances. When an error is emitted, the AST for the
170input code may not be fully built. The <tt>EXTENSION</tt> and <tt>EXTWARN</tt>
171severities are used for extensions to the language that Clang accepts. This
172means that Clang fully understands and can represent them in the AST, but we
173produce diagnostics to tell the user their code is non-portable. The difference
174is that the former are ignored by default, and the later warn by default. The
175<tt>WARNING</tt> severity is used for constructs that are valid in the currently
176selected source language but that are dubious in some way. The <tt>NOTE</tt>
Daniel Dunbar426b8632009-02-17 15:49:03 +0000[diff] [blame]177level is used to staple more information onto previous diagnostics.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]178
179<p>These <em>severities</em> are mapped into a smaller set (the
180Diagnostic::Level enum, {<tt>Ignored</tt>, <tt>Note</tt>, <tt>Warning</tt>,
Chris Lattner0aad2972009-02-05 22:49:08 +0000[diff] [blame]181<tt>Error</tt>, <tt>Fatal</tt> }) of output <em>levels</em> by the diagnostics
Chris Lattnera180fdd2009-02-17 07:07:29 +0000[diff] [blame]182subsystem based on various configuration options. Clang internally supports a
183fully fine grained mapping mechanism that allows you to map almost any
184diagnostic to the output level that you want. The only diagnostics that cannot
185be mapped are <tt>NOTE</tt>s, which always follow the severity of the previously
186emitted diagnostic and <tt>ERROR</tt>s, which can only be mapped to
187<tt>Fatal</tt> (it is not possible to turn an error into a warning,
188for example).</p>
189
190<p>Diagnostic mappings are used in many ways. For example, if the user
191specifies <tt>-pedantic</tt>, <tt>EXTENSION</tt> maps to <tt>Warning</tt>, if
192they specify <tt>-pedantic-errors</tt>, it turns into <tt>Error</tt>. This is
193used to implement options like <tt>-Wunused_macros</tt>, <tt>-Wundef</tt> etc.
194</p>
195
196<p>
197Mapping to <tt>Fatal</tt> should only be used for diagnostics that are
198considered so severe that error recovery won't be able to recover sensibly from
199them (thus spewing a ton of bogus errors). One example of this class of error
200are failure to #include a file.
201</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]202
203<!-- ================= -->
204<h4>The Format String</h4>
205<!-- ================= -->
206
207<p>The format string for the diagnostic is very simple, but it has some power.
208It takes the form of a string in English with markers that indicate where and
209how arguments to the diagnostic are inserted and formatted. For example, here
210are some simple format strings:</p>
211
212<pre>
213 "binary integer literals are an extension"
214 "format string contains '\\0' within the string body"
215 "more '<b>%%</b>' conversions than data arguments"
Chris Lattner545b3682008-11-23 20:27:13 +0000[diff] [blame]216 "invalid operands to binary expression (<b>%0</b> and <b>%1</b>)"
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]217 "overloaded '<b>%0</b>' must be a <b>%select{unary|binary|unary or binary}2</b> operator"
218 " (has <b>%1</b> parameter<b>%s1</b>)"
219</pre>
220
221<p>These examples show some important points of format strings. You can use any
222 plain ASCII character in the diagnostic string except "%" without a problem,
223 but these are C strings, so you have to use and be aware of all the C escape
224 sequences (as in the second example). If you want to produce a "%" in the
225 output, use the "%%" escape sequence, like the third diagnostic. Finally,
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]226 Clang uses the "%...[digit]" sequences to specify where and how arguments to
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]227 the diagnostic are formatted.</p>
228
229<p>Arguments to the diagnostic are numbered according to how they are specified
230 by the C++ code that <a href="#producingdiag">produces them</a>, and are
231 referenced by <tt>%0</tt> .. <tt>%9</tt>. If you have more than 10 arguments
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]232 to your diagnostic, you are doing something wrong :). Unlike printf, there
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]233 is no requirement that arguments to the diagnostic end up in the output in
234 the same order as they are specified, you could have a format string with
235 <tt>"%1 %0"</tt> that swaps them, for example. The text in between the
236 percent and digit are formatting instructions. If there are no instructions,
237 the argument is just turned into a string and substituted in.</p>
238
239<p>Here are some "best practices" for writing the English format string:</p>
240
241<ul>
242<li>Keep the string short. It should ideally fit in the 80 column limit of the
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000[diff] [blame]243 <tt>DiagnosticKinds.td</tt> file. This avoids the diagnostic wrapping when
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]244 printed, and forces you to think about the important point you are conveying
245 with the diagnostic.</li>
246<li>Take advantage of location information. The user will be able to see the
247 line and location of the caret, so you don't need to tell them that the
248 problem is with the 4th argument to the function: just point to it.</li>
249<li>Do not capitalize the diagnostic string, and do not end it with a
250 period.</li>
251<li>If you need to quote something in the diagnostic string, use single
252 quotes.</li>
253</ul>
254
255<p>Diagnostics should never take random English strings as arguments: you
256shouldn't use <tt>"you have a problem with %0"</tt> and pass in things like
257<tt>"your argument"</tt> or <tt>"your return value"</tt> as arguments. Doing
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000[diff] [blame]258this prevents <a href="#translation">translating</a> the Clang diagnostics to
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]259other languages (because they'll get random English words in their otherwise
260localized diagnostic). The exceptions to this are C/C++ language keywords
261(e.g. auto, const, mutable, etc) and C/C++ operators (<tt>/=</tt>). Note
262that things like "pointer" and "reference" are not keywords. On the other
263hand, you <em>can</em> include anything that comes from the user's source code,
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]264including variable names, types, labels, etc. The 'select' format can be
265used to achieve this sort of thing in a localizable way, see below.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]266
267<!-- ==================================== -->
268<h4>Formatting a Diagnostic Argument</a></h4>
269<!-- ==================================== -->
270
271<p>Arguments to diagnostics are fully typed internally, and come from a couple
272different classes: integers, types, names, and random strings. Depending on
273the class of the argument, it can be optionally formatted in different ways.
274This gives the DiagnosticClient information about what the argument means
275without requiring it to use a specific presentation (consider this MVC for
276Clang :).</p>
277
278<p>Here are the different diagnostic argument formats currently supported by
279Clang:</p>
280
281<table>
282<tr><td colspan="2"><b>"s" format</b></td></tr>
283<tr><td>Example:</td><td><tt>"requires %1 parameter%s1"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]284<tr><td>Class:</td><td>Integers</td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]285<tr><td>Description:</td><td>This is a simple formatter for integers that is
286 useful when producing English diagnostics. When the integer is 1, it prints
287 as nothing. When the integer is not 1, it prints as "s". This allows some
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]288 simple grammatical forms to be to be handled correctly, and eliminates the
289 need to use gross things like <tt>"requires %1 parameter(s)"</tt>.</td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]290
291<tr><td colspan="2"><b>"select" format</b></td></tr>
292<tr><td>Example:</td><td><tt>"must be a %select{unary|binary|unary or binary}2
293 operator"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]294<tr><td>Class:</td><td>Integers</td></tr>
John McCall3a47e232010-01-14 19:12:17 +0000[diff] [blame]295<tr><td>Description:</td><td><p>This format specifier is used to merge multiple
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]296 related diagnostics together into one common one, without requiring the
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]297 difference to be specified as an English string argument. Instead of
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]298 specifying the string, the diagnostic gets an integer argument and the
299 format string selects the numbered option. In this case, the "%2" value
300 must be an integer in the range [0..2]. If it is 0, it prints 'unary', if
301 it is 1 it prints 'binary' if it is 2, it prints 'unary or binary'. This
302 allows other language translations to substitute reasonable words (or entire
303 phrases) based on the semantics of the diagnostic instead of having to do
John McCall3a47e232010-01-14 19:12:17 +0000[diff] [blame]304 things textually.</p>
305 <p>The selected string does undergo formatting.</p></td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]306
307<tr><td colspan="2"><b>"plural" format</b></td></tr>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]308<tr><td>Example:</td><td><tt>"you have %1 %plural{1:mouse|:mice}1 connected to
309 your computer"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]310<tr><td>Class:</td><td>Integers</td></tr>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]311<tr><td>Description:</td><td><p>This is a formatter for complex plural forms.
312 It is designed to handle even the requirements of languages with very
313 complex plural forms, as many Baltic languages have. The argument consists
314 of a series of expression/form pairs, separated by ':', where the first form
315 whose expression evaluates to true is the result of the modifier.</p>
316 <p>An expression can be empty, in which case it is always true. See the
317 example at the top. Otherwise, it is a series of one or more numeric
318 conditions, separated by ','. If any condition matches, the expression
319 matches. Each numeric condition can take one of three forms.</p>
320 <ul>
321 <li>number: A simple decimal number matches if the argument is the same
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]322 as the number. Example: <tt>"%plural{1:mouse|:mice}4"</tt></li>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]323 <li>range: A range in square brackets matches if the argument is within
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]324 the range. Then range is inclusive on both ends. Example:
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]325 <tt>"%plural{0:none|1:one|[2,5]:some|:many}2"</tt></li>
326 <li>modulo: A modulo operator is followed by a number, and
327 equals sign and either a number or a range. The tests are the
328 same as for plain
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]329 numbers and ranges, but the argument is taken modulo the number first.
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]330 Example: <tt>"%plural{%100=0:even hundred|%100=[1,50]:lower half|:everything
331 else}1"</tt></li>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]332 </ul>
333 <p>The parser is very unforgiving. A syntax error, even whitespace, will
334 abort, as will a failure to match the argument against any
335 expression.</p></td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]336
John McCall3a47e232010-01-14 19:12:17 +0000[diff] [blame]337<tr><td colspan="2"><b>"ordinal" format</b></td></tr>
338<tr><td>Example:</td><td><tt>"ambiguity in %ordinal0 argument"</tt></td></tr>
339<tr><td>Class:</td><td>Integers</td></tr>
340<tr><td>Description:</td><td><p>This is a formatter which represents the
341 argument number as an ordinal: the value <tt>1</tt> becomes <tt>1st</tt>,
342 <tt>3</tt> becomes <tt>3rd</tt>, and so on. Values less than <tt>1</tt>
343 are not supported.</p>
344 <p>This formatter is currently hard-coded to use English ordinals.</p></td></tr>
345
Chris Lattner077bf5e2008-11-24 03:33:13 +0000[diff] [blame]346<tr><td colspan="2"><b>"objcclass" format</b></td></tr>
347<tr><td>Example:</td><td><tt>"method %objcclass0 not found"</tt></td></tr>
348<tr><td>Class:</td><td>DeclarationName</td></tr>
349<tr><td>Description:</td><td><p>This is a simple formatter that indicates the
350 DeclarationName corresponds to an Objective-C class method selector. As
351 such, it prints the selector with a leading '+'.</p></td></tr>
352
353<tr><td colspan="2"><b>"objcinstance" format</b></td></tr>
354<tr><td>Example:</td><td><tt>"method %objcinstance0 not found"</tt></td></tr>
355<tr><td>Class:</td><td>DeclarationName</td></tr>
356<tr><td>Description:</td><td><p>This is a simple formatter that indicates the
357 DeclarationName corresponds to an Objective-C instance method selector. As
358 such, it prints the selector with a leading '-'.</p></td></tr>
359
Douglas Gregor47b9a1c2009-02-04 17:27:36 +0000[diff] [blame]360<tr><td colspan="2"><b>"q" format</b></td></tr>
361<tr><td>Example:</td><td><tt>"candidate found by name lookup is %q0"</tt></td></tr>
362<tr><td>Class:</td><td>NamedDecl*</td></tr>
363<tr><td>Description</td><td><p>This formatter indicates that the fully-qualified name of the declaration should be printed, e.g., "std::vector" rather than "vector".</p></td></tr>
364
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]365</table>
366
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]367<p>It is really easy to add format specifiers to the Clang diagnostics system,
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]368but they should be discussed before they are added. If you are creating a lot
369of repetitive diagnostics and/or have an idea for a useful formatter, please
370bring it up on the cfe-dev mailing list.</p>
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]371
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]372<!-- ===================================================== -->
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000[diff] [blame]373<h4 id="producingdiag">Producing the Diagnostic</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]374<!-- ===================================================== -->
375
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000[diff] [blame]376<p>Now that you've created the diagnostic in the DiagnosticKinds.td file, you
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]377need to write the code that detects the condition in question and emits the
378new diagnostic. Various components of Clang (e.g. the preprocessor, Sema,
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]379etc) provide a helper function named "Diag". It creates a diagnostic and
380accepts the arguments, ranges, and other information that goes along with
381it.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]382
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]383<p>For example, the binary expression error comes from code like this:</p>
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]384
385<pre>
386 if (various things that are bad)
387 Diag(Loc, diag::err_typecheck_invalid_operands)
388 &lt;&lt; lex-&gt;getType() &lt;&lt; rex-&gt;getType()
389 &lt;&lt; lex-&gt;getSourceRange() &lt;&lt; rex-&gt;getSourceRange();
390</pre>
391
392<p>This shows that use of the Diag method: they take a location (a <a
393href="#SourceLocation">SourceLocation</a> object) and a diagnostic enum value
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000[diff] [blame]394(which matches the name from DiagnosticKinds.td). If the diagnostic takes
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]395arguments, they are specified with the &lt;&lt; operator: the first argument
396becomes %0, the second becomes %1, etc. The diagnostic interface allows you to
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]397specify arguments of many different types, including <tt>int</tt> and
398<tt>unsigned</tt> for integer arguments, <tt>const char*</tt> and
399<tt>std::string</tt> for string arguments, <tt>DeclarationName</tt> and
400<tt>const IdentifierInfo*</tt> for names, <tt>QualType</tt> for types, etc.
401SourceRanges are also specified with the &lt;&lt; operator, but do not have a
402specific ordering requirement.</p>
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]403
404<p>As you can see, adding and producing a diagnostic is pretty straightforward.
405The hard part is deciding exactly what you need to say to help the user, picking
406a suitable wording, and providing the information needed to format it correctly.
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]407The good news is that the call site that issues a diagnostic should be
408completely independent of how the diagnostic is formatted and in what language
409it is rendered.
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]410</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]411
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]412<!-- ==================================================== -->
Peter Collingbourne38448d32011-03-21 01:45:18 +0000[diff] [blame]413<h4 id="fix-it-hints">Fix-It Hints</h4>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]414<!-- ==================================================== -->
415
416<p>In some cases, the front end emits diagnostics when it is clear
417that some small change to the source code would fix the problem. For
418example, a missing semicolon at the end of a statement or a use of
Chris Lattner34c05332009-02-27 19:31:12 +0000[diff] [blame]419deprecated syntax that is easily rewritten into a more modern form.
420Clang tries very hard to emit the diagnostic and recover gracefully
421in these and other cases.</p>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]422
Peter Collingbourne38448d32011-03-21 01:45:18 +0000[diff] [blame]423<p>However, for these cases where the fix is obvious, the diagnostic
424can be annotated with a hint (referred to as a "fix-it hint") that
425describes how to change the code referenced by the diagnostic to fix
426the problem. For example, it might add the missing semicolon at the
427end of the statement or rewrite the use of a deprecated construct
428into something more palatable. Here is one such example from the C++
429front end, where we warn about the right-shift operator changing
David Blaikie5090e9f2011-10-18 05:49:30 +0000[diff] [blame]430meaning from C++98 to C++11:</p>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]431
432<pre>
David Blaikie5090e9f2011-10-18 05:49:30 +0000[diff] [blame]433test.cpp:3:7: warning: use of right-shift operator ('&gt;&gt;') in template argument will require parentheses in C++11
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]434A&lt;100 &gt;&gt; 2&gt; *a;
435 ^
436 ( )
437</pre>
438
Peter Collingbourne38448d32011-03-21 01:45:18 +0000[diff] [blame]439<p>Here, the fix-it hint is suggesting that parentheses be added,
440and showing exactly where those parentheses would be inserted into the
441source code. The fix-it hints themselves describe what changes to make
442to the source code in an abstract manner, which the text diagnostic
443printer renders as a line of "insertions" below the caret line. <a
444href="#DiagnosticClient">Other diagnostic clients</a> might choose
445to render the code differently (e.g., as markup inline) or even give
446the user the ability to automatically fix the problem.</p>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]447
Peter Collingbourne38448d32011-03-21 01:45:18 +0000[diff] [blame]448<p>All fix-it hints are described by the <code>FixItHint</code> class,
449instances of which should be attached to the diagnostic using the
450&lt;&lt; operator in the same way that highlighted source ranges and
451arguments are passed to the diagnostic. Fix-it hints can be created
452with one of three constructors:</p>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]453
454<dl>
Peter Collingbourne38448d32011-03-21 01:45:18 +0000[diff] [blame]455 <dt><code>FixItHint::CreateInsertion(Loc, Code)</code></dt>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]456 <dd>Specifies that the given <code>Code</code> (a string) should be inserted
457 before the source location <code>Loc</code>.</dd>
458
Peter Collingbourne38448d32011-03-21 01:45:18 +0000[diff] [blame]459 <dt><code>FixItHint::CreateRemoval(Range)</code></dt>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]460 <dd>Specifies that the code in the given source <code>Range</code>
461 should be removed.</dd>
462
Peter Collingbourne38448d32011-03-21 01:45:18 +0000[diff] [blame]463 <dt><code>FixItHint::CreateReplacement(Range, Code)</code></dt>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]464 <dd>Specifies that the code in the given source <code>Range</code>
465 should be removed, and replaced with the given <code>Code</code> string.</dd>
466</dl>
467
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]468<!-- ============================================================= -->
469<h4><a name="DiagnosticClient">The DiagnosticClient Interface</a></h4>
470<!-- ============================================================= -->
471
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]472<p>Once code generates a diagnostic with all of the arguments and the rest of
473the relevant information, Clang needs to know what to do with it. As previously
474mentioned, the diagnostic machinery goes through some filtering to map a
475severity onto a diagnostic level, then (assuming the diagnostic is not mapped to
476"<tt>Ignore</tt>") it invokes an object that implements the DiagnosticClient
477interface with the information.</p>
478
479<p>It is possible to implement this interface in many different ways. For
480example, the normal Clang DiagnosticClient (named 'TextDiagnosticPrinter') turns
481the arguments into strings (according to the various formatting rules), prints
482out the file/line/column information and the string, then prints out the line of
483code, the source ranges, and the caret. However, this behavior isn't required.
484</p>
485
486<p>Another implementation of the DiagnosticClient interface is the
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]487'TextDiagnosticBuffer' class, which is used when Clang is in -verify mode.
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]488Instead of formatting and printing out the diagnostics, this implementation just
489captures and remembers the diagnostics as they fly by. Then -verify compares
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]490the list of produced diagnostics to the list of expected ones. If they disagree,
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]491it prints out its own output.
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]492</p>
493
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]494<p>There are many other possible implementations of this interface, and this is
495why we prefer diagnostics to pass down rich structured information in arguments.
496For example, an HTML output might want declaration names be linkified to where
497they come from in the source. Another example is that a GUI might let you click
498on typedefs to expand them. This application would want to pass significantly
499more information about types through to the GUI than a simple flat string. The
500interface allows this to happen.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]501
502<!-- ====================================================== -->
503<h4><a name="translation">Adding Translations to Clang</a></h4>
504<!-- ====================================================== -->
505
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]506<p>Not possible yet! Diagnostic strings should be written in UTF-8, the client
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]507can translate to the relevant code page if needed. Each translation completely
508replaces the format string for the diagnostic.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]509
510
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]511<!-- ======================================================================= -->
512<h3 id="SourceLocation">The SourceLocation and SourceManager classes</h3>
513<!-- ======================================================================= -->
514
515<p>Strangely enough, the SourceLocation class represents a location within the
516source code of the program. Important design points include:</p>
517
518<ol>
519<li>sizeof(SourceLocation) must be extremely small, as these are embedded into
520 many AST nodes and are passed around often. Currently it is 32 bits.</li>
521<li>SourceLocation must be a simple value object that can be efficiently
522 copied.</li>
523<li>We should be able to represent a source location for any byte of any input
524 file. This includes in the middle of tokens, in whitespace, in trigraphs,
525 etc.</li>
526<li>A SourceLocation must encode the current #include stack that was active when
527 the location was processed. For example, if the location corresponds to a
528 token, it should contain the set of #includes active when the token was
529 lexed. This allows us to print the #include stack for a diagnostic.</li>
530<li>SourceLocation must be able to describe macro expansions, capturing both
531 the ultimate instantiation point and the source of the original character
532 data.</li>
533</ol>
534
535<p>In practice, the SourceLocation works together with the SourceManager class
Nick Lewycky77561e52010-05-26 21:48:10 +0000[diff] [blame]536to encode two pieces of information about a location: its spelling location
537and its instantiation location. For most tokens, these will be the same.
538However, for a macro expansion (or tokens that came from a _Pragma directive)
539these will describe the location of the characters corresponding to the token
540and the location where the token was used (i.e. the macro instantiation point
541or the location of the _Pragma itself).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]542
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]543<p>The Clang front-end inherently depends on the location of a token being
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]544tracked correctly. If it is ever incorrect, the front-end may get confused and
545die. The reason for this is that the notion of the 'spelling' of a Token in
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]546Clang depends on being able to find the original input characters for the token.
Chris Lattner18376dd2009-01-16 07:00:50 +0000[diff] [blame]547This concept maps directly to the "spelling location" for the token.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]548
Douglas Gregor715c92a2010-10-27 16:02:28 +0000[diff] [blame]549
550<!-- ======================================================================= -->
551<h3 id="SourceRange">SourceRange and CharSourceRange</h3>
552<!-- ======================================================================= -->
553<!-- mostly taken from
554 http://lists.cs.uiuc.edu/pipermail/cfe-dev/2010-August/010595.html -->
555
556<p>Clang represents most source ranges by [first, last], where first and last
557each point to the beginning of their respective tokens. For example
558consider the SourceRange of the following statement:</p>
559<pre>
560x = foo + bar;
561^first ^last
562</pre>
563
564<p>To map from this representation to a character-based
565representation, the 'last' location needs to be adjusted to point to
566(or past) the end of that token with either
567<code>Lexer::MeasureTokenLength()</code> or
Chris Lattner7ef5c272010-11-17 07:05:50 +0000[diff] [blame]568<code>Lexer::getLocForEndOfToken()</code>. For the rare cases
Douglas Gregor715c92a2010-10-27 16:02:28 +0000[diff] [blame]569where character-level source ranges information is needed we use
570the <code>CharSourceRange</code> class.</p>
571
572
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]573<!-- ======================================================================= -->
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000[diff] [blame]574<h2 id="libdriver">The Driver Library</h2>
575<!-- ======================================================================= -->
576
Ted Kremenekcfa8d572009-04-09 18:08:18 +0000[diff] [blame]577<p>The clang Driver and library are documented <a
578href="DriverInternals.html">here<a>.<p>
579
580<!-- ======================================================================= -->
Douglas Gregor32110df2009-05-20 00:16:32 +0000[diff] [blame]581<h2 id="pch">Precompiled Headers</h2>
Ted Kremenekcfa8d572009-04-09 18:08:18 +0000[diff] [blame]582<!-- ======================================================================= -->
583
Douglas Gregor32110df2009-05-20 00:16:32 +0000[diff] [blame]584<p>Clang supports two implementations of precompiled headers. The
585 default implementation, precompiled headers (<a
586 href="PCHInternals.html">PCH</a>) uses a serialized representation
587 of Clang's internal data structures, encoded with the <a
588 href="http://llvm.org/docs/BitCodeFormat.html">LLVM bitstream
589 format</a>. Pretokenized headers (<a
590 href="PTHInternals.html">PTH</a>), on the other hand, contain a
591 serialized representation of the tokens encountered when
592 preprocessing a header (and anything that header includes).</p>
593
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000[diff] [blame]594
595<!-- ======================================================================= -->
596<h2 id="libfrontend">The Frontend Library</h2>
597<!-- ======================================================================= -->
598
599<p>The Frontend library contains functionality useful for building
600tools on top of the clang libraries, for example several methods for
601outputting diagnostics.</p>
602
603<!-- ======================================================================= -->
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]604<h2 id="liblex">The Lexer and Preprocessor Library</h2>
605<!-- ======================================================================= -->
606
607<p>The Lexer library contains several tightly-connected classes that are involved
608with the nasty process of lexing and preprocessing C source code. The main
609interface to this library for outside clients is the large <a
610href="#Preprocessor">Preprocessor</a> class.
611It contains the various pieces of state that are required to coherently read
612tokens out of a translation unit.</p>
613
614<p>The core interface to the Preprocessor object (once it is set up) is the
615Preprocessor::Lex method, which returns the next <a href="#Token">Token</a> from
616the preprocessor stream. There are two types of token providers that the
617preprocessor is capable of reading from: a buffer lexer (provided by the <a
618href="#Lexer">Lexer</a> class) and a buffered token stream (provided by the <a
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]619href="#TokenLexer">TokenLexer</a> class).
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]620
621
622<!-- ======================================================================= -->
623<h3 id="Token">The Token class</h3>
624<!-- ======================================================================= -->
625
626<p>The Token class is used to represent a single lexed token. Tokens are
627intended to be used by the lexer/preprocess and parser libraries, but are not
628intended to live beyond them (for example, they should not live in the ASTs).<p>
629
630<p>Tokens most often live on the stack (or some other location that is efficient
631to access) as the parser is running, but occasionally do get buffered up. For
632example, macro definitions are stored as a series of tokens, and the C++
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]633front-end periodically needs to buffer tokens up for tentative parsing and
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]634various pieces of look-ahead. As such, the size of a Token matter. On a 32-bit
635system, sizeof(Token) is currently 16 bytes.</p>
636
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]637<p>Tokens occur in two forms: "<a href="#AnnotationToken">Annotation
638Tokens</a>" and normal tokens. Normal tokens are those returned by the lexer,
639annotation tokens represent semantic information and are produced by the parser,
640replacing normal tokens in the token stream. Normal tokens contain the
641following information:</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]642
643<ul>
644<li><b>A SourceLocation</b> - This indicates the location of the start of the
645token.</li>
646
647<li><b>A length</b> - This stores the length of the token as stored in the
648SourceBuffer. For tokens that include them, this length includes trigraphs and
649escaped newlines which are ignored by later phases of the compiler. By pointing
650into the original source buffer, it is always possible to get the original
651spelling of a token completely accurately.</li>
652
653<li><b>IdentifierInfo</b> - If a token takes the form of an identifier, and if
654identifier lookup was enabled when the token was lexed (e.g. the lexer was not
655reading in 'raw' mode) this contains a pointer to the unique hash value for the
656identifier. Because the lookup happens before keyword identification, this
657field is set even for language keywords like 'for'.</li>
658
659<li><b>TokenKind</b> - This indicates the kind of token as classified by the
660lexer. This includes things like <tt>tok::starequal</tt> (for the "*="
661operator), <tt>tok::ampamp</tt> for the "&amp;&amp;" token, and keyword values
662(e.g. <tt>tok::kw_for</tt>) for identifiers that correspond to keywords. Note
663that some tokens can be spelled multiple ways. For example, C++ supports
664"operator keywords", where things like "and" are treated exactly like the
665"&amp;&amp;" operator. In these cases, the kind value is set to
666<tt>tok::ampamp</tt>, which is good for the parser, which doesn't have to
667consider both forms. For something that cares about which form is used (e.g.
668the preprocessor 'stringize' operator) the spelling indicates the original
669form.</li>
670
671<li><b>Flags</b> - There are currently four flags tracked by the
672lexer/preprocessor system on a per-token basis:
673
674 <ol>
675 <li><b>StartOfLine</b> - This was the first token that occurred on its input
676 source line.</li>
677 <li><b>LeadingSpace</b> - There was a space character either immediately
678 before the token or transitively before the token as it was expanded
679 through a macro. The definition of this flag is very closely defined by
680 the stringizing requirements of the preprocessor.</li>
681 <li><b>DisableExpand</b> - This flag is used internally to the preprocessor to
682 represent identifier tokens which have macro expansion disabled. This
683 prevents them from being considered as candidates for macro expansion ever
684 in the future.</li>
685 <li><b>NeedsCleaning</b> - This flag is set if the original spelling for the
686 token includes a trigraph or escaped newline. Since this is uncommon,
687 many pieces of code can fast-path on tokens that did not need cleaning.
688 </p>
689 </ol>
690</li>
691</ul>
692
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]693<p>One interesting (and somewhat unusual) aspect of normal tokens is that they
694don't contain any semantic information about the lexed value. For example, if
695the token was a pp-number token, we do not represent the value of the number
696that was lexed (this is left for later pieces of code to decide). Additionally,
697the lexer library has no notion of typedef names vs variable names: both are
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]698returned as identifiers, and the parser is left to decide whether a specific
699identifier is a typedef or a variable (tracking this requires scope information
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]700among other things). The parser can do this translation by replacing tokens
701returned by the preprocessor with "Annotation Tokens".</p>
702
703<!-- ======================================================================= -->
704<h3 id="AnnotationToken">Annotation Tokens</h3>
705<!-- ======================================================================= -->
706
707<p>Annotation Tokens are tokens that are synthesized by the parser and injected
708into the preprocessor's token stream (replacing existing tokens) to record
709semantic information found by the parser. For example, if "foo" is found to be
710a typedef, the "foo" <tt>tok::identifier</tt> token is replaced with an
711<tt>tok::annot_typename</tt>. This is useful for a couple of reasons: 1) this
712makes it easy to handle qualified type names (e.g. "foo::bar::baz&lt;42&gt;::t")
713in C++ as a single "token" in the parser. 2) if the parser backtracks, the
714reparse does not need to redo semantic analysis to determine whether a token
715sequence is a variable, type, template, etc.</p>
716
717<p>Annotation Tokens are created by the parser and reinjected into the parser's
718token stream (when backtracking is enabled). Because they can only exist in
719tokens that the preprocessor-proper is done with, it doesn't need to keep around
720flags like "start of line" that the preprocessor uses to do its job.
721Additionally, an annotation token may "cover" a sequence of preprocessor tokens
722(e.g. <tt>a::b::c</tt> is five preprocessor tokens). As such, the valid fields
723of an annotation token are different than the fields for a normal token (but
724they are multiplexed into the normal Token fields):</p>
725
726<ul>
727<li><b>SourceLocation "Location"</b> - The SourceLocation for the annotation
728token indicates the first token replaced by the annotation token. In the example
729above, it would be the location of the "a" identifier.</li>
730
731<li><b>SourceLocation "AnnotationEndLoc"</b> - This holds the location of the
732last token replaced with the annotation token. In the example above, it would
733be the location of the "c" identifier.</li>
734
John McCall027ac442010-09-03 05:07:55 +0000[diff] [blame]735<li><b>void* "AnnotationValue"</b> - This contains an opaque object
736that the parser gets from Sema. The parser merely preserves the
737information for Sema to later interpret based on the annotation token
738kind.</li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]739
740<li><b>TokenKind "Kind"</b> - This indicates the kind of Annotation token this
741is. See below for the different valid kinds.</li>
742</ul>
743
744<p>Annotation tokens currently come in three kinds:</p>
745
746<ol>
747<li><b>tok::annot_typename</b>: This annotation token represents a
John McCall027ac442010-09-03 05:07:55 +0000[diff] [blame]748resolved typename token that is potentially qualified. The
749AnnotationValue field contains the <tt>QualType</tt> returned by
750Sema::getTypeName(), possibly with source location information
751attached.</li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]752
John McCall027ac442010-09-03 05:07:55 +0000[diff] [blame]753<li><b>tok::annot_cxxscope</b>: This annotation token represents a C++
754scope specifier, such as "A::B::". This corresponds to the grammar
755productions "::" and ":: [opt] nested-name-specifier". The
756AnnotationValue pointer is a <tt>NestedNameSpecifier*</tt> returned by
757the Sema::ActOnCXXGlobalScopeSpecifier and
758Sema::ActOnCXXNestedNameSpecifier callbacks.</li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]759
Douglas Gregor39a8de12009-02-25 19:37:18 +0000[diff] [blame]760<li><b>tok::annot_template_id</b>: This annotation token represents a
761C++ template-id such as "foo&lt;int, 4&gt;", where "foo" is the name
762of a template. The AnnotationValue pointer is a pointer to a malloc'd
John McCall027ac442010-09-03 05:07:55 +0000[diff] [blame]763TemplateIdAnnotation object. Depending on the context, a parsed
764template-id that names a type might become a typename annotation token
765(if all we care about is the named type, e.g., because it occurs in a
766type specifier) or might remain a template-id token (if we want to
767retain more source location information or produce a new type, e.g.,
768in a declaration of a class template specialization). template-id
769annotation tokens that refer to a type can be "upgraded" to typename
770annotation tokens by the parser.</li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]771
772</ol>
773
Cedric Venetda76b282009-01-06 16:22:54 +0000[diff] [blame]774<p>As mentioned above, annotation tokens are not returned by the preprocessor,
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]775they are formed on demand by the parser. This means that the parser has to be
776aware of cases where an annotation could occur and form it where appropriate.
777This is somewhat similar to how the parser handles Translation Phase 6 of C99:
778String Concatenation (see C99 5.1.1.2). In the case of string concatenation,
779the preprocessor just returns distinct tok::string_literal and
780tok::wide_string_literal tokens and the parser eats a sequence of them wherever
781the grammar indicates that a string literal can occur.</p>
782
783<p>In order to do this, whenever the parser expects a tok::identifier or
784tok::coloncolon, it should call the TryAnnotateTypeOrScopeToken or
785TryAnnotateCXXScopeToken methods to form the annotation token. These methods
786will maximally form the specified annotation tokens and replace the current
787token with them, if applicable. If the current tokens is not valid for an
788annotation token, it will remain an identifier or :: token.</p>
789
790
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]791
792<!-- ======================================================================= -->
793<h3 id="Lexer">The Lexer class</h3>
794<!-- ======================================================================= -->
795
796<p>The Lexer class provides the mechanics of lexing tokens out of a source
797buffer and deciding what they mean. The Lexer is complicated by the fact that
798it operates on raw buffers that have not had spelling eliminated (this is a
799necessity to get decent performance), but this is countered with careful coding
800as well as standard performance techniques (for example, the comment handling
801code is vectorized on X86 and PowerPC hosts).</p>
802
803<p>The lexer has a couple of interesting modal features:</p>
804
805<ul>
806<li>The lexer can operate in 'raw' mode. This mode has several features that
807 make it possible to quickly lex the file (e.g. it stops identifier lookup,
808 doesn't specially handle preprocessor tokens, handles EOF differently, etc).
809 This mode is used for lexing within an "<tt>#if 0</tt>" block, for
810 example.</li>
811<li>The lexer can capture and return comments as tokens. This is required to
812 support the -C preprocessor mode, which passes comments through, and is
813 used by the diagnostic checker to identifier expect-error annotations.</li>
814<li>The lexer can be in ParsingFilename mode, which happens when preprocessing
Chris Lattner84386242007-09-16 19:25:23 +0000[diff] [blame]815 after reading a #include directive. This mode changes the parsing of '&lt;'
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]816 to return an "angled string" instead of a bunch of tokens for each thing
817 within the filename.</li>
818<li>When parsing a preprocessor directive (after "<tt>#</tt>") the
819 ParsingPreprocessorDirective mode is entered. This changes the parser to
Peter Collingbourne84021552011-02-28 02:37:51 +0000[diff] [blame]820 return EOD at a newline.</li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]821<li>The Lexer uses a LangOptions object to know whether trigraphs are enabled,
822 whether C++ or ObjC keywords are recognized, etc.</li>
823</ul>
824
825<p>In addition to these modes, the lexer keeps track of a couple of other
826 features that are local to a lexed buffer, which change as the buffer is
827 lexed:</p>
828
829<ul>
830<li>The Lexer uses BufferPtr to keep track of the current character being
831 lexed.</li>
832<li>The Lexer uses IsAtStartOfLine to keep track of whether the next lexed token
833 will start with its "start of line" bit set.</li>
834<li>The Lexer keeps track of the current #if directives that are active (which
835 can be nested).</li>
836<li>The Lexer keeps track of an <a href="#MultipleIncludeOpt">
837 MultipleIncludeOpt</a> object, which is used to
838 detect whether the buffer uses the standard "<tt>#ifndef XX</tt> /
839 <tt>#define XX</tt>" idiom to prevent multiple inclusion. If a buffer does,
840 subsequent includes can be ignored if the XX macro is defined.</li>
841</ul>
842
843<!-- ======================================================================= -->
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]844<h3 id="TokenLexer">The TokenLexer class</h3>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]845<!-- ======================================================================= -->
846
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]847<p>The TokenLexer class is a token provider that returns tokens from a list
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]848of tokens that came from somewhere else. It typically used for two things: 1)
849returning tokens from a macro definition as it is being expanded 2) returning
850tokens from an arbitrary buffer of tokens. The later use is used by _Pragma and
851will most likely be used to handle unbounded look-ahead for the C++ parser.</p>
852
853<!-- ======================================================================= -->
854<h3 id="MultipleIncludeOpt">The MultipleIncludeOpt class</h3>
855<!-- ======================================================================= -->
856
857<p>The MultipleIncludeOpt class implements a really simple little state machine
858that is used to detect the standard "<tt>#ifndef XX</tt> / <tt>#define XX</tt>"
859idiom that people typically use to prevent multiple inclusion of headers. If a
860buffer uses this idiom and is subsequently #include'd, the preprocessor can
861simply check to see whether the guarding condition is defined or not. If so,
862the preprocessor can completely ignore the include of the header.</p>
863
864
865
866<!-- ======================================================================= -->
867<h2 id="libparse">The Parser Library</h2>
868<!-- ======================================================================= -->
869
870<!-- ======================================================================= -->
871<h2 id="libast">The AST Library</h2>
872<!-- ======================================================================= -->
873
874<!-- ======================================================================= -->
875<h3 id="Type">The Type class and its subclasses</h3>
876<!-- ======================================================================= -->
877
878<p>The Type class (and its subclasses) are an important part of the AST. Types
879are accessed through the ASTContext class, which implicitly creates and uniques
880them as they are needed. Types have a couple of non-obvious features: 1) they
881do not capture type qualifiers like const or volatile (See
882<a href="#QualType">QualType</a>), and 2) they implicitly capture typedef
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]883information. Once created, types are immutable (unlike decls).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]884
885<p>Typedefs in C make semantic analysis a bit more complex than it would
886be without them. The issue is that we want to capture typedef information
887and represent it in the AST perfectly, but the semantics of operations need to
888"see through" typedefs. For example, consider this code:</p>
889
890<code>
891void func() {<br>
Bill Wendling30d17752007-10-06 01:56:01 +0000[diff] [blame]892&nbsp;&nbsp;typedef int foo;<br>
893&nbsp;&nbsp;foo X, *Y;<br>
894&nbsp;&nbsp;typedef foo* bar;<br>
895&nbsp;&nbsp;bar Z;<br>
896&nbsp;&nbsp;*X; <i>// error</i><br>
897&nbsp;&nbsp;**Y; <i>// error</i><br>
898&nbsp;&nbsp;**Z; <i>// error</i><br>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]899}<br>
900</code>
901
902<p>The code above is illegal, and thus we expect there to be diagnostics emitted
903on the annotated lines. In this example, we expect to get:</p>
904
905<pre>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]906<b>test.c:6:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]907*X; // error
908<font color="blue">^~</font>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]909<b>test.c:7:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]910**Y; // error
911<font color="blue">^~~</font>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]912<b>test.c:8:1: error: indirection requires pointer operand ('foo' invalid)</b>
913**Z; // error
914<font color="blue">^~~</font>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]915</pre>
916
917<p>While this example is somewhat silly, it illustrates the point: we want to
918retain typedef information where possible, so that we can emit errors about
919"<tt>std::string</tt>" instead of "<tt>std::basic_string&lt;char, std:...</tt>".
920Doing this requires properly keeping typedef information (for example, the type
921of "X" is "foo", not "int"), and requires properly propagating it through the
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]922various operators (for example, the type of *Y is "foo", not "int"). In order
923to retain this information, the type of these expressions is an instance of the
924TypedefType class, which indicates that the type of these expressions is a
925typedef for foo.
926</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]927
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]928<p>Representing types like this is great for diagnostics, because the
929user-specified type is always immediately available. There are two problems
930with this: first, various semantic checks need to make judgements about the
Douglas Gregor2d1e21a2011-12-19 19:50:23 +0000[diff] [blame]931<em>actual structure</em> of a type, ignoring typedefs. Second, we need an
Chris Lattner33fc68a2007-07-31 18:54:50 +0000[diff] [blame]932efficient way to query whether two types are structurally identical to each
933other, ignoring typedefs. The solution to both of these problems is the idea of
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]934canonical types.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]935
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]936<!-- =============== -->
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]937<h4>Canonical Types</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]938<!-- =============== -->
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]939
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]940<p>Every instance of the Type class contains a canonical type pointer. For
941simple types with no typedefs involved (e.g. "<tt>int</tt>", "<tt>int*</tt>",
942"<tt>int**</tt>"), the type just points to itself. For types that have a
943typedef somewhere in their structure (e.g. "<tt>foo</tt>", "<tt>foo*</tt>",
944"<tt>foo**</tt>", "<tt>bar</tt>"), the canonical type pointer points to their
945structurally equivalent type without any typedefs (e.g. "<tt>int</tt>",
946"<tt>int*</tt>", "<tt>int**</tt>", and "<tt>int*</tt>" respectively).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]947
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]948<p>This design provides a constant time operation (dereferencing the canonical
949type pointer) that gives us access to the structure of types. For example,
950we can trivially tell that "bar" and "foo*" are the same type by dereferencing
951their canonical type pointers and doing a pointer comparison (they both point
952to the single "<tt>int*</tt>" type).</p>
953
954<p>Canonical types and typedef types bring up some complexities that must be
955carefully managed. Specifically, the "isa/cast/dyncast" operators generally
956shouldn't be used in code that is inspecting the AST. For example, when type
957checking the indirection operator (unary '*' on a pointer), the type checker
958must verify that the operand has a pointer type. It would not be correct to
959check that with "<tt>isa&lt;PointerType&gt;(SubExpr-&gt;getType())</tt>",
960because this predicate would fail if the subexpression had a typedef type.</p>
961
962<p>The solution to this problem are a set of helper methods on Type, used to
963check their properties. In this case, it would be correct to use
964"<tt>SubExpr-&gt;getType()-&gt;isPointerType()</tt>" to do the check. This
965predicate will return true if the <em>canonical type is a pointer</em>, which is
966true any time the type is structurally a pointer type. The only hard part here
967is remembering not to use the <tt>isa/cast/dyncast</tt> operations.</p>
968
969<p>The second problem we face is how to get access to the pointer type once we
970know it exists. To continue the example, the result type of the indirection
971operator is the pointee type of the subexpression. In order to determine the
972type, we need to get the instance of PointerType that best captures the typedef
973information in the program. If the type of the expression is literally a
974PointerType, we can return that, otherwise we have to dig through the
975typedefs to find the pointer type. For example, if the subexpression had type
976"<tt>foo*</tt>", we could return that type as the result. If the subexpression
977had type "<tt>bar</tt>", we want to return "<tt>foo*</tt>" (note that we do
978<em>not</em> want "<tt>int*</tt>"). In order to provide all of this, Type has
Chris Lattner11406c12007-07-31 16:50:51 +0000[diff] [blame]979a getAsPointerType() method that checks whether the type is structurally a
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]980PointerType and, if so, returns the best one. If not, it returns a null
981pointer.</p>
982
983<p>This structure is somewhat mystical, but after meditating on it, it will
984make sense to you :).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]985
986<!-- ======================================================================= -->
987<h3 id="QualType">The QualType class</h3>
988<!-- ======================================================================= -->
989
John McCall027ac442010-09-03 05:07:55 +0000[diff] [blame]990<p>The QualType class is designed as a trivial value class that is
991small, passed by-value and is efficient to query. The idea of
992QualType is that it stores the type qualifiers (const, volatile,
993restrict, plus some extended qualifiers required by language
994extensions) separately from the types themselves. QualType is
995conceptually a pair of "Type*" and the bits for these type qualifiers.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]996
997<p>By storing the type qualifiers as bits in the conceptual pair, it is
998extremely efficient to get the set of qualifiers on a QualType (just return the
999field of the pair), add a type qualifier (which is a trivial constant-time
1000operation that sets a bit), and remove one or more type qualifiers (just return
1001a QualType with the bitfield set to empty).</p>
1002
1003<p>Further, because the bits are stored outside of the type itself, we do not
1004need to create duplicates of types with different sets of qualifiers (i.e. there
1005is only a single heap allocated "int" type: "const int" and "volatile const int"
1006both point to the same heap allocated "int" type). This reduces the heap size
1007used to represent bits and also means we do not have to consider qualifiers when
1008uniquing types (<a href="#Type">Type</a> does not even contain qualifiers).</p>
1009
John McCall027ac442010-09-03 05:07:55 +0000[diff] [blame]1010<p>In practice, the two most common type qualifiers (const and
1011restrict) are stored in the low bits of the pointer to the Type
1012object, together with a flag indicating whether extended qualifiers
1013are present (which must be heap-allocated). This means that QualType
1014is exactly the same size as a pointer.</p>
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1015
1016<!-- ======================================================================= -->
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1017<h3 id="DeclarationName">Declaration names</h3>
1018<!-- ======================================================================= -->
1019
1020<p>The <tt>DeclarationName</tt> class represents the name of a
1021 declaration in Clang. Declarations in the C family of languages can
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]1022 take several different forms. Most declarations are named by
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1023 simple identifiers, e.g., "<code>f</code>" and "<code>x</code>" in
1024 the function declaration <code>f(int x)</code>. In C++, declaration
1025 names can also name class constructors ("<code>Class</code>"
1026 in <code>struct Class { Class(); }</code>), class destructors
1027 ("<code>~Class</code>"), overloaded operator names ("operator+"),
1028 and conversion functions ("<code>operator void const *</code>"). In
1029 Objective-C, declaration names can refer to the names of Objective-C
1030 methods, which involve the method name and the parameters,
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]1031 collectively called a <i>selector</i>, e.g.,
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1032 "<code>setWidth:height:</code>". Since all of these kinds of
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]1033 entities - variables, functions, Objective-C methods, C++
1034 constructors, destructors, and operators - are represented as
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1035 subclasses of Clang's common <code>NamedDecl</code>
1036 class, <code>DeclarationName</code> is designed to efficiently
1037 represent any kind of name.</p>
1038
1039<p>Given
1040 a <code>DeclarationName</code> <code>N</code>, <code>N.getNameKind()</code>
Douglas Gregor2def4832008-11-17 20:34:05 +0000[diff] [blame]1041 will produce a value that describes what kind of name <code>N</code>
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1042 stores. There are 8 options (all of the names are inside
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1043 the <code>DeclarationName</code> class)</p>
1044<dl>
1045 <dt>Identifier</dt>
1046 <dd>The name is a simple
1047 identifier. Use <code>N.getAsIdentifierInfo()</code> to retrieve the
1048 corresponding <code>IdentifierInfo*</code> pointing to the actual
1049 identifier. Note that C++ overloaded operators (e.g.,
1050 "<code>operator+</code>") are represented as special kinds of
1051 identifiers. Use <code>IdentifierInfo</code>'s <code>getOverloadedOperatorID</code>
1052 function to determine whether an identifier is an overloaded
1053 operator name.</dd>
1054
1055 <dt>ObjCZeroArgSelector, ObjCOneArgSelector,
1056 ObjCMultiArgSelector</dt>
1057 <dd>The name is an Objective-C selector, which can be retrieved as a
1058 <code>Selector</code> instance
1059 via <code>N.getObjCSelector()</code>. The three possible name
1060 kinds for Objective-C reflect an optimization within
1061 the <code>DeclarationName</code> class: both zero- and
1062 one-argument selectors are stored as a
1063 masked <code>IdentifierInfo</code> pointer, and therefore require
1064 very little space, since zero- and one-argument selectors are far
1065 more common than multi-argument selectors (which use a different
1066 structure).</dd>
1067
1068 <dt>CXXConstructorName</dt>
1069 <dd>The name is a C++ constructor
1070 name. Use <code>N.getCXXNameType()</code> to retrieve
1071 the <a href="#QualType">type</a> that this constructor is meant to
1072 construct. The type is always the canonical type, since all
1073 constructors for a given type have the same name.</dd>
1074
1075 <dt>CXXDestructorName</dt>
1076 <dd>The name is a C++ destructor
1077 name. Use <code>N.getCXXNameType()</code> to retrieve
1078 the <a href="#QualType">type</a> whose destructor is being
1079 named. This type is always a canonical type.</dd>
1080
1081 <dt>CXXConversionFunctionName</dt>
1082 <dd>The name is a C++ conversion function. Conversion functions are
1083 named according to the type they convert to, e.g., "<code>operator void
1084 const *</code>". Use <code>N.getCXXNameType()</code> to retrieve
1085 the type that this conversion function converts to. This type is
1086 always a canonical type.</dd>
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1087
1088 <dt>CXXOperatorName</dt>
1089 <dd>The name is a C++ overloaded operator name. Overloaded operators
1090 are named according to their spelling, e.g.,
1091 "<code>operator+</code>" or "<code>operator new
1092 []</code>". Use <code>N.getCXXOverloadedOperator()</code> to
1093 retrieve the overloaded operator (a value of
1094 type <code>OverloadedOperatorKind</code>).</dd>
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1095</dl>
1096
1097<p><code>DeclarationName</code>s are cheap to create, copy, and
1098 compare. They require only a single pointer's worth of storage in
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1099 the common cases (identifiers, zero-
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1100 and one-argument Objective-C selectors) and use dense, uniqued
1101 storage for the other kinds of
1102 names. Two <code>DeclarationName</code>s can be compared for
1103 equality (<code>==</code>, <code>!=</code>) using a simple bitwise
1104 comparison, can be ordered
1105 with <code>&lt;</code>, <code>&gt;</code>, <code>&lt;=</code>,
1106 and <code>&gt;=</code> (which provide a lexicographical ordering for
1107 normal identifiers but an unspecified ordering for other kinds of
1108 names), and can be placed into LLVM <code>DenseMap</code>s
1109 and <code>DenseSet</code>s.</p>
1110
1111<p><code>DeclarationName</code> instances can be created in different
1112 ways depending on what kind of name the instance will store. Normal
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1113 identifiers (<code>IdentifierInfo</code> pointers) and Objective-C selectors
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1114 (<code>Selector</code>) can be implicitly converted
1115 to <code>DeclarationName</code>s. Names for C++ constructors,
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1116 destructors, conversion functions, and overloaded operators can be retrieved from
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1117 the <code>DeclarationNameTable</code>, an instance of which is
1118 available as <code>ASTContext::DeclarationNames</code>. The member
1119 functions <code>getCXXConstructorName</code>, <code>getCXXDestructorName</code>,
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1120 <code>getCXXConversionFunctionName</code>, and <code>getCXXOperatorName</code>, respectively,
1121 return <code>DeclarationName</code> instances for the four kinds of
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1122 C++ special function names.</p>
1123
1124<!-- ======================================================================= -->
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1125<h3 id="DeclContext">Declaration contexts</h3>
1126<!-- ======================================================================= -->
1127<p>Every declaration in a program exists within some <i>declaration
1128 context</i>, such as a translation unit, namespace, class, or
1129 function. Declaration contexts in Clang are represented by
1130 the <code>DeclContext</code> class, from which the various
1131 declaration-context AST nodes
1132 (<code>TranslationUnitDecl</code>, <code>NamespaceDecl</code>, <code>RecordDecl</code>, <code>FunctionDecl</code>,
1133 etc.) will derive. The <code>DeclContext</code> class provides
1134 several facilities common to each declaration context:</p>
1135<dl>
1136 <dt>Source-centric vs. Semantics-centric View of Declarations</dt>
1137 <dd><code>DeclContext</code> provides two views of the declarations
1138 stored within a declaration context. The source-centric view
1139 accurately represents the program source code as written, including
1140 multiple declarations of entities where present (see the
1141 section <a href="#Redeclarations">Redeclarations and
1142 Overloads</a>), while the semantics-centric view represents the
1143 program semantics. The two views are kept synchronized by semantic
1144 analysis while the ASTs are being constructed.</dd>
1145
1146 <dt>Storage of declarations within that context</dt>
1147 <dd>Every declaration context can contain some number of
1148 declarations. For example, a C++ class (represented
1149 by <code>RecordDecl</code>) contains various member functions,
1150 fields, nested types, and so on. All of these declarations will be
1151 stored within the <code>DeclContext</code>, and one can iterate
1152 over the declarations via
1153 [<code>DeclContext::decls_begin()</code>,
1154 <code>DeclContext::decls_end()</code>). This mechanism provides
1155 the source-centric view of declarations in the context.</dd>
1156
1157 <dt>Lookup of declarations within that context</dt>
1158 <dd>The <code>DeclContext</code> structure provides efficient name
1159 lookup for names within that declaration context. For example,
1160 if <code>N</code> is a namespace we can look for the
1161 name <code>N::f</code>
1162 using <code>DeclContext::lookup</code>. The lookup itself is
1163 based on a lazily-constructed array (for declaration contexts
1164 with a small number of declarations) or hash table (for
1165 declaration contexts with more declarations). The lookup
1166 operation provides the semantics-centric view of the declarations
1167 in the context.</dd>
1168
1169 <dt>Ownership of declarations</dt>
1170 <dd>The <code>DeclContext</code> owns all of the declarations that
1171 were declared within its declaration context, and is responsible
1172 for the management of their memory as well as their
1173 (de-)serialization.</dd>
1174</dl>
1175
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1176<p>All declarations are stored within a declaration context, and one
1177 can query
1178 information about the context in which each declaration lives. One
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1179 can retrieve the <code>DeclContext</code> that contains a
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1180 particular <code>Decl</code>
1181 using <code>Decl::getDeclContext</code>. However, see the
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1182 section <a href="#LexicalAndSemanticContexts">Lexical and Semantic
1183 Contexts</a> for more information about how to interpret this
1184 context information.</p>
1185
1186<h4 id="Redeclarations">Redeclarations and Overloads</h4>
1187<p>Within a translation unit, it is common for an entity to be
1188declared several times. For example, we might declare a function "f"
1189 and then later re-declare it as part of an inlined definition:</p>
1190
1191<pre>
1192void f(int x, int y, int z = 1);
1193
1194inline void f(int x, int y, int z) { /* ... */ }
1195</pre>
1196
1197<p>The representation of "f" differs in the source-centric and
1198 semantics-centric views of a declaration context. In the
1199 source-centric view, all redeclarations will be present, in the
1200 order they occurred in the source code, making
1201 this view suitable for clients that wish to see the structure of
1202 the source code. In the semantics-centric view, only the most recent "f"
1203 will be found by the lookup, since it effectively replaces the first
1204 declaration of "f".</p>
1205
1206<p>In the semantics-centric view, overloading of functions is
1207 represented explicitly. For example, given two declarations of a
1208 function "g" that are overloaded, e.g.,</p>
1209<pre>
1210void g();
1211void g(int);
1212</pre>
1213<p>the <code>DeclContext::lookup</code> operation will return
Jonathan D. Turnerd3224292011-07-06 18:12:36 +0000[diff] [blame]1214 a <code>DeclContext::lookup_result</code> that contains a range of iterators
1215 over declarations of "g". Clients that perform semantic analysis on a
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1216 program that is not concerned with the actual source code will
1217 primarily use this semantics-centric view.</p>
1218
1219<h4 id="LexicalAndSemanticContexts">Lexical and Semantic Contexts</h4>
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1220<p>Each declaration has two potentially different
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1221 declaration contexts: a <i>lexical</i> context, which corresponds to
1222 the source-centric view of the declaration context, and
1223 a <i>semantic</i> context, which corresponds to the
1224 semantics-centric view. The lexical context is accessible
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1225 via <code>Decl::getLexicalDeclContext</code> while the
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1226 semantic context is accessible
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1227 via <code>Decl::getDeclContext</code>, both of which return
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1228 <code>DeclContext</code> pointers. For most declarations, the two
1229 contexts are identical. For example:</p>
1230
1231<pre>
1232class X {
1233public:
1234 void f(int x);
1235};
1236</pre>
1237
1238<p>Here, the semantic and lexical contexts of <code>X::f</code> are
1239 the <code>DeclContext</code> associated with the
1240 class <code>X</code> (itself stored as a <code>RecordDecl</code> AST
1241 node). However, we can now define <code>X::f</code> out-of-line:</p>
1242
1243<pre>
1244void X::f(int x = 17) { /* ... */ }
1245</pre>
1246
1247<p>This definition of has different lexical and semantic
1248 contexts. The lexical context corresponds to the declaration
1249 context in which the actual declaration occurred in the source
1250 code, e.g., the translation unit containing <code>X</code>. Thus,
1251 this declaration of <code>X::f</code> can be found by traversing
1252 the declarations provided by
1253 [<code>decls_begin()</code>, <code>decls_end()</code>) in the
1254 translation unit.</p>
1255
1256<p>The semantic context of <code>X::f</code> corresponds to the
1257 class <code>X</code>, since this member function is (semantically) a
1258 member of <code>X</code>. Lookup of the name <code>f</code> into
1259 the <code>DeclContext</code> associated with <code>X</code> will
1260 then return the definition of <code>X::f</code> (including
1261 information about the default argument).</p>
1262
1263<h4 id="TransparentContexts">Transparent Declaration Contexts</h4>
1264<p>In C and C++, there are several contexts in which names that are
1265 logically declared inside another declaration will actually "leak"
1266 out into the enclosing scope from the perspective of name
1267 lookup. The most obvious instance of this behavior is in
1268 enumeration types, e.g.,</p>
1269<pre>
1270enum Color {
1271 Red,
1272 Green,
1273 Blue
1274};
1275</pre>
1276
1277<p>Here, <code>Color</code> is an enumeration, which is a declaration
1278 context that contains the
1279 enumerators <code>Red</code>, <code>Green</code>,
1280 and <code>Blue</code>. Thus, traversing the list of declarations
1281 contained in the enumeration <code>Color</code> will
1282 yield <code>Red</code>, <code>Green</code>,
1283 and <code>Blue</code>. However, outside of the scope
1284 of <code>Color</code> one can name the enumerator <code>Red</code>
1285 without qualifying the name, e.g.,</p>
1286
1287<pre>
1288Color c = Red;
1289</pre>
1290
1291<p>There are other entities in C++ that provide similar behavior. For
1292 example, linkage specifications that use curly braces:</p>
1293
1294<pre>
1295extern "C" {
1296 void f(int);
1297 void g(int);
1298}
1299// f and g are visible here
1300</pre>
1301
1302<p>For source-level accuracy, we treat the linkage specification and
1303 enumeration type as a
1304 declaration context in which its enclosed declarations ("Red",
1305 "Green", and "Blue"; "f" and "g")
1306 are declared. However, these declarations are visible outside of the
1307 scope of the declaration context.</p>
1308
1309<p>These language features (and several others, described below) have
1310 roughly the same set of
1311 requirements: declarations are declared within a particular lexical
1312 context, but the declarations are also found via name lookup in
1313 scopes enclosing the declaration itself. This feature is implemented
1314 via <i>transparent</i> declaration contexts
1315 (see <code>DeclContext::isTransparentContext()</code>), whose
1316 declarations are visible in the nearest enclosing non-transparent
1317 declaration context. This means that the lexical context of the
1318 declaration (e.g., an enumerator) will be the
1319 transparent <code>DeclContext</code> itself, as will the semantic
1320 context, but the declaration will be visible in every outer context
1321 up to and including the first non-transparent declaration context (since
1322 transparent declaration contexts can be nested).</p>
1323
1324<p>The transparent <code>DeclContexts</code> are:</p>
1325<ul>
David Blaikie5090e9f2011-10-18 05:49:30 +0000[diff] [blame]1326 <li>Enumerations (but not C++11 "scoped enumerations"):
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1327 <pre>
1328enum Color {
1329 Red,
1330 Green,
1331 Blue
1332};
1333// Red, Green, and Blue are in scope
1334 </pre></li>
1335 <li>C++ linkage specifications:
1336 <pre>
1337extern "C" {
1338 void f(int);
1339 void g(int);
1340}
1341// f and g are in scope
1342 </pre></li>
1343 <li>Anonymous unions and structs:
1344 <pre>
1345struct LookupTable {
1346 bool IsVector;
1347 union {
1348 std::vector&lt;Item&gt; *Vector;
1349 std::set&lt;Item&gt; *Set;
1350 };
1351};
1352
1353LookupTable LT;
1354LT.Vector = 0; // Okay: finds Vector inside the unnamed union
1355 </pre>
1356 </li>
David Blaikie5090e9f2011-10-18 05:49:30 +0000[diff] [blame]1357 <li>C++11 inline namespaces:
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1358<pre>
1359namespace mylib {
1360 inline namespace debug {
1361 class X;
1362 }
1363}
1364mylib::X *xp; // okay: mylib::X refers to mylib::debug::X
1365</pre>
1366</li>
1367</ul>
1368
1369
1370<h4 id="MultiDeclContext">Multiply-Defined Declaration Contexts</h4>
1371<p>C++ namespaces have the interesting--and, so far, unique--property that
1372the namespace can be defined multiple times, and the declarations
1373provided by each namespace definition are effectively merged (from
1374the semantic point of view). For example, the following two code
1375snippets are semantically indistinguishable:</p>
1376<pre>
1377// Snippet #1:
1378namespace N {
1379 void f();
1380}
1381namespace N {
1382 void f(int);
1383}
1384
1385// Snippet #2:
1386namespace N {
1387 void f();
1388 void f(int);
1389}
1390</pre>
1391
1392<p>In Clang's representation, the source-centric view of declaration
1393 contexts will actually have two separate <code>NamespaceDecl</code>
1394 nodes in Snippet #1, each of which is a declaration context that
1395 contains a single declaration of "f". However, the semantics-centric
1396 view provided by name lookup into the namespace <code>N</code> for
Jonathan D. Turnerd3224292011-07-06 18:12:36 +0000[diff] [blame]1397 "f" will return a <code>DeclContext::lookup_result</code> that contains
1398 a range of iterators over declarations of "f".</p>
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1399
1400<p><code>DeclContext</code> manages multiply-defined declaration
1401 contexts internally. The
1402 function <code>DeclContext::getPrimaryContext</code> retrieves the
1403 "primary" context for a given <code>DeclContext</code> instance,
1404 which is the <code>DeclContext</code> responsible for maintaining
1405 the lookup table used for the semantics-centric view. Given the
1406 primary context, one can follow the chain
1407 of <code>DeclContext</code> nodes that define additional
1408 declarations via <code>DeclContext::getNextContext</code>. Note that
1409 these functions are used internally within the lookup and insertion
1410 methods of the <code>DeclContext</code>, so the vast majority of
1411 clients can ignore them.</p>
1412
1413<!-- ======================================================================= -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1414<h3 id="CFG">The <tt>CFG</tt> class</h3>
1415<!-- ======================================================================= -->
1416
1417<p>The <tt>CFG</tt> class is designed to represent a source-level
1418control-flow graph for a single statement (<tt>Stmt*</tt>). Typically
1419instances of <tt>CFG</tt> are constructed for function bodies (usually
1420an instance of <tt>CompoundStmt</tt>), but can also be instantiated to
1421represent the control-flow of any class that subclasses <tt>Stmt</tt>,
1422which includes simple expressions. Control-flow graphs are especially
1423useful for performing
1424<a href="http://en.wikipedia.org/wiki/Data_flow_analysis#Sensitivities">flow-
1425or path-sensitive</a> program analyses on a given function.</p>
1426
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1427<!-- ============ -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1428<h4>Basic Blocks</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1429<!-- ============ -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1430
1431<p>Concretely, an instance of <tt>CFG</tt> is a collection of basic
1432blocks. Each basic block is an instance of <tt>CFGBlock</tt>, which
1433simply contains an ordered sequence of <tt>Stmt*</tt> (each referring
1434to statements in the AST). The ordering of statements within a block
1435indicates unconditional flow of control from one statement to the
1436next. <a href="#ConditionalControlFlow">Conditional control-flow</a>
1437is represented using edges between basic blocks. The statements
1438within a given <tt>CFGBlock</tt> can be traversed using
1439the <tt>CFGBlock::*iterator</tt> interface.</p>
1440
1441<p>
Ted Kremenek18e17e72007-10-18 22:50:52 +0000[diff] [blame]1442A <tt>CFG</tt> object owns the instances of <tt>CFGBlock</tt> within
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1443the control-flow graph it represents. Each <tt>CFGBlock</tt> within a
1444CFG is also uniquely numbered (accessible
1445via <tt>CFGBlock::getBlockID()</tt>). Currently the number is
1446based on the ordering the blocks were created, but no assumptions
1447should be made on how <tt>CFGBlock</tt>s are numbered other than their
1448numbers are unique and that they are numbered from 0..N-1 (where N is
1449the number of basic blocks in the CFG).</p>
1450
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1451<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1452<h4>Entry and Exit Blocks</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1453<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1454
1455Each instance of <tt>CFG</tt> contains two special blocks:
1456an <i>entry</i> block (accessible via <tt>CFG::getEntry()</tt>), which
1457has no incoming edges, and an <i>exit</i> block (accessible
1458via <tt>CFG::getExit()</tt>), which has no outgoing edges. Neither
1459block contains any statements, and they serve the role of providing a
1460clear entrance and exit for a body of code such as a function body.
1461The presence of these empty blocks greatly simplifies the
1462implementation of many analyses built on top of CFGs.
1463
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1464<!-- ===================================================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1465<h4 id ="ConditionalControlFlow">Conditional Control-Flow</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1466<!-- ===================================================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1467
1468<p>Conditional control-flow (such as those induced by if-statements
1469and loops) is represented as edges between <tt>CFGBlock</tt>s.
1470Because different C language constructs can induce control-flow,
1471each <tt>CFGBlock</tt> also records an extra <tt>Stmt*</tt> that
1472represents the <i>terminator</i> of the block. A terminator is simply
1473the statement that caused the control-flow, and is used to identify
1474the nature of the conditional control-flow between blocks. For
1475example, in the case of an if-statement, the terminator refers to
1476the <tt>IfStmt</tt> object in the AST that represented the given
1477branch.</p>
1478
1479<p>To illustrate, consider the following code example:</p>
1480
1481<code>
1482int foo(int x) {<br>
1483&nbsp;&nbsp;x = x + 1;<br>
1484<br>
1485&nbsp;&nbsp;if (x > 2) x++;<br>
1486&nbsp;&nbsp;else {<br>
1487&nbsp;&nbsp;&nbsp;&nbsp;x += 2;<br>
1488&nbsp;&nbsp;&nbsp;&nbsp;x *= 2;<br>
1489&nbsp;&nbsp;}<br>
1490<br>
1491&nbsp;&nbsp;return x;<br>
1492}
1493</code>
1494
1495<p>After invoking the parser+semantic analyzer on this code fragment,
1496the AST of the body of <tt>foo</tt> is referenced by a
1497single <tt>Stmt*</tt>. We can then construct an instance
1498of <tt>CFG</tt> representing the control-flow graph of this function
1499body by single call to a static class method:</p>
1500
1501<code>
1502&nbsp;&nbsp;Stmt* FooBody = ...<br>
1503&nbsp;&nbsp;CFG* FooCFG = <b>CFG::buildCFG</b>(FooBody);
1504</code>
1505
1506<p>It is the responsibility of the caller of <tt>CFG::buildCFG</tt>
1507to <tt>delete</tt> the returned <tt>CFG*</tt> when the CFG is no
1508longer needed.</p>
1509
1510<p>Along with providing an interface to iterate over
1511its <tt>CFGBlock</tt>s, the <tt>CFG</tt> class also provides methods
1512that are useful for debugging and visualizing CFGs. For example, the
1513method
1514<tt>CFG::dump()</tt> dumps a pretty-printed version of the CFG to
1515standard error. This is especially useful when one is using a
1516debugger such as gdb. For example, here is the output
1517of <tt>FooCFG->dump()</tt>:</p>
1518
1519<code>
1520&nbsp;[ B5 (ENTRY) ]<br>
1521&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (0):<br>
1522&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B4<br>
1523<br>
1524&nbsp;[ B4 ]<br>
1525&nbsp;&nbsp;&nbsp;&nbsp;1: x = x + 1<br>
1526&nbsp;&nbsp;&nbsp;&nbsp;2: (x > 2)<br>
1527&nbsp;&nbsp;&nbsp;&nbsp;<b>T: if [B4.2]</b><br>
1528&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B5<br>
1529&nbsp;&nbsp;&nbsp;&nbsp;Successors (2): B3 B2<br>
1530<br>
1531&nbsp;[ B3 ]<br>
1532&nbsp;&nbsp;&nbsp;&nbsp;1: x++<br>
1533&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B4<br>
1534&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B1<br>
1535<br>
1536&nbsp;[ B2 ]<br>
1537&nbsp;&nbsp;&nbsp;&nbsp;1: x += 2<br>
1538&nbsp;&nbsp;&nbsp;&nbsp;2: x *= 2<br>
1539&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B4<br>
1540&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B1<br>
1541<br>
1542&nbsp;[ B1 ]<br>
1543&nbsp;&nbsp;&nbsp;&nbsp;1: return x;<br>
1544&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (2): B2 B3<br>
1545&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B0<br>
1546<br>
1547&nbsp;[ B0 (EXIT) ]<br>
1548&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B1<br>
1549&nbsp;&nbsp;&nbsp;&nbsp;Successors (0):
1550</code>
1551
1552<p>For each block, the pretty-printed output displays for each block
1553the number of <i>predecessor</i> blocks (blocks that have outgoing
1554control-flow to the given block) and <i>successor</i> blocks (blocks
1555that have control-flow that have incoming control-flow from the given
1556block). We can also clearly see the special entry and exit blocks at
1557the beginning and end of the pretty-printed output. For the entry
1558block (block B5), the number of predecessor blocks is 0, while for the
1559exit block (block B0) the number of successor blocks is 0.</p>
1560
1561<p>The most interesting block here is B4, whose outgoing control-flow
1562represents the branching caused by the sole if-statement
1563in <tt>foo</tt>. Of particular interest is the second statement in
1564the block, <b><tt>(x > 2)</tt></b>, and the terminator, printed
1565as <b><tt>if [B4.2]</tt></b>. The second statement represents the
1566evaluation of the condition of the if-statement, which occurs before
1567the actual branching of control-flow. Within the <tt>CFGBlock</tt>
1568for B4, the <tt>Stmt*</tt> for the second statement refers to the
1569actual expression in the AST for <b><tt>(x > 2)</tt></b>. Thus
1570pointers to subclasses of <tt>Expr</tt> can appear in the list of
1571statements in a block, and not just subclasses of <tt>Stmt</tt> that
1572refer to proper C statements.</p>
1573
1574<p>The terminator of block B4 is a pointer to the <tt>IfStmt</tt>
1575object in the AST. The pretty-printer outputs <b><tt>if
1576[B4.2]</tt></b> because the condition expression of the if-statement
1577has an actual place in the basic block, and thus the terminator is
1578essentially
1579<i>referring</i> to the expression that is the second statement of
1580block B4 (i.e., B4.2). In this manner, conditions for control-flow
1581(which also includes conditions for loops and switch statements) are
1582hoisted into the actual basic block.</p>
1583
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1584<!-- ===================== -->
1585<!-- <h4>Implicit Control-Flow</h4> -->
1586<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1587
1588<!--
1589<p>A key design principle of the <tt>CFG</tt> class was to not require
1590any transformations to the AST in order to represent control-flow.
1591Thus the <tt>CFG</tt> does not perform any "lowering" of the
1592statements in an AST: loops are not transformed into guarded gotos,
1593short-circuit operations are not converted to a set of if-statements,
1594and so on.</p>
1595-->
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]1596
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1597
1598<!-- ======================================================================= -->
1599<h3 id="Constants">Constant Folding in the Clang AST</h3>
1600<!-- ======================================================================= -->
1601
1602<p>There are several places where constants and constant folding matter a lot to
1603the Clang front-end. First, in general, we prefer the AST to retain the source
1604code as close to how the user wrote it as possible. This means that if they
1605wrote "5+4", we want to keep the addition and two constants in the AST, we don't
1606want to fold to "9". This means that constant folding in various ways turns
1607into a tree walk that needs to handle the various cases.</p>
1608
1609<p>However, there are places in both C and C++ that require constants to be
1610folded. For example, the C standard defines what an "integer constant
1611expression" (i-c-e) is with very precise and specific requirements. The
1612language then requires i-c-e's in a lot of places (for example, the size of a
1613bitfield, the value for a case statement, etc). For these, we have to be able
1614to constant fold the constants, to do semantic checks (e.g. verify bitfield size
1615is non-negative and that case statements aren't duplicated). We aim for Clang
1616to be very pedantic about this, diagnosing cases when the code does not use an
1617i-c-e where one is required, but accepting the code unless running with
1618<tt>-pedantic-errors</tt>.</p>
1619
1620<p>Things get a little bit more tricky when it comes to compatibility with
1621real-world source code. Specifically, GCC has historically accepted a huge
1622superset of expressions as i-c-e's, and a lot of real world code depends on this
1623unfortuate accident of history (including, e.g., the glibc system headers). GCC
1624accepts anything its "fold" optimizer is capable of reducing to an integer
1625constant, which means that the definition of what it accepts changes as its
1626optimizer does. One example is that GCC accepts things like "case X-X:" even
1627when X is a variable, because it can fold this to 0.</p>
1628
1629<p>Another issue are how constants interact with the extensions we support, such
1630as __builtin_constant_p, __builtin_inf, __extension__ and many others. C99
1631obviously does not specify the semantics of any of these extensions, and the
1632definition of i-c-e does not include them. However, these extensions are often
1633used in real code, and we have to have a way to reason about them.</p>
1634
1635<p>Finally, this is not just a problem for semantic analysis. The code
1636generator and other clients have to be able to fold constants (e.g. to
1637initialize global variables) and has to handle a superset of what C99 allows.
1638Further, these clients can benefit from extended information. For example, we
1639know that "foo()||1" always evaluates to true, but we can't replace the
1640expression with true because it has side effects.</p>
1641
1642<!-- ======================= -->
1643<h4>Implementation Approach</h4>
1644<!-- ======================= -->
1645
1646<p>After trying several different approaches, we've finally converged on a
1647design (Note, at the time of this writing, not all of this has been implemented,
1648consider this a design goal!). Our basic approach is to define a single
1649recursive method evaluation method (<tt>Expr::Evaluate</tt>), which is
1650implemented in <tt>AST/ExprConstant.cpp</tt>. Given an expression with 'scalar'
1651type (integer, fp, complex, or pointer) this method returns the following
1652information:</p>
1653
1654<ul>
1655<li>Whether the expression is an integer constant expression, a general
1656 constant that was folded but has no side effects, a general constant that
1657 was folded but that does have side effects, or an uncomputable/unfoldable
1658 value.
1659</li>
1660<li>If the expression was computable in any way, this method returns the APValue
1661 for the result of the expression.</li>
1662<li>If the expression is not evaluatable at all, this method returns
1663 information on one of the problems with the expression. This includes a
1664 SourceLocation for where the problem is, and a diagnostic ID that explains
1665 the problem. The diagnostic should be have ERROR type.</li>
1666<li>If the expression is not an integer constant expression, this method returns
1667 information on one of the problems with the expression. This includes a
1668 SourceLocation for where the problem is, and a diagnostic ID that explains
1669 the problem. The diagnostic should be have EXTENSION type.</li>
1670</ul>
1671
1672<p>This information gives various clients the flexibility that they want, and we
1673will eventually have some helper methods for various extensions. For example,
1674Sema should have a <tt>Sema::VerifyIntegerConstantExpression</tt> method, which
1675calls Evaluate on the expression. If the expression is not foldable, the error
1676is emitted, and it would return true. If the expression is not an i-c-e, the
1677EXTENSION diagnostic is emitted. Finally it would return false to indicate that
1678the AST is ok.</p>
1679
1680<p>Other clients can use the information in other ways, for example, codegen can
1681just use expressions that are foldable in any way.</p>
1682
1683<!-- ========== -->
1684<h4>Extensions</h4>
1685<!-- ========== -->
1686
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]1687<p>This section describes how some of the various extensions Clang supports
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1688interacts with constant evaluation:</p>
1689
1690<ul>
1691<li><b><tt>__extension__</tt></b>: The expression form of this extension causes
1692 any evaluatable subexpression to be accepted as an integer constant
1693 expression.</li>
1694<li><b><tt>__builtin_constant_p</tt></b>: This returns true (as a integer
Richard Smith8a0f1552011-12-09 03:40:28 +0000[diff] [blame]1695 constant expression) if the operand evaluates to either a numeric value
1696 (that is, not a pointer cast to integral type) of integral, enumeration,
1697 floating or complex type, or if it evaluates to the address of the first
1698 character of a string literal (possibly cast to some other type). As a
Chris Lattner28daa532008-12-12 06:55:44 +0000[diff] [blame]1699 special case, if <tt>__builtin_constant_p</tt> is the (potentially
1700 parenthesized) condition of a conditional operator expression ("?:"), only
Chris Lattner42b83dd2008-12-12 18:00:51 +0000[diff] [blame]1701 the true side of the conditional operator is considered, and it is evaluated
1702 with full constant folding.</li>
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1703<li><b><tt>__builtin_choose_expr</tt></b>: The condition is required to be an
1704 integer constant expression, but we accept any constant as an "extension of
1705 an extension". This only evaluates one operand depending on which way the
1706 condition evaluates.</li>
1707<li><b><tt>__builtin_classify_type</tt></b>: This always returns an integer
1708 constant expression.</li>
1709<li><b><tt>__builtin_inf,nan,..</tt></b>: These are treated just like a
1710 floating-point literal.</li>
1711<li><b><tt>__builtin_abs,copysign,..</tt></b>: These are constant folded as
1712 general constant expressions.</li>
Richard Smith8a0f1552011-12-09 03:40:28 +0000[diff] [blame]1713<li><b><tt>__builtin_strlen</tt></b> and <b><tt>strlen</tt></b>: These are
1714 constant folded as integer constant expressions if the argument is a string
1715 literal.</li>
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1716</ul>
1717
1718
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +0000[diff] [blame]1719<!-- ======================================================================= -->
1720<h2 id="Howtos">How to change Clang</h2>
1721<!-- ======================================================================= -->
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1722
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +0000[diff] [blame]1723<!-- ======================================================================= -->
1724<h3 id="AddingAttributes">How to add an attribute</h3>
1725<!-- ======================================================================= -->
1726
1727<p>To add an attribute, you'll have to add it to the list of attributes, add it
1728to the parsing phase, and look for it in the AST scan.
1729<a href="http://llvm.org/viewvc/llvm-project?view=rev&revision=124217">r124217</a>
1730has a good example of adding a warning attribute.</p>
1731
1732<p>(Beware that this hasn't been reviewed/fixed by the people who designed the
1733attributes system yet.)</p>
1734
1735<h4><a
1736href="http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/Attr.td?view=markup">include/clang/Basic/Attr.td</a></h4>
1737
1738<p>Each attribute gets a <tt>def</tt> inheriting from <tt>Attr</tt> or one of
1739its subclasses. <tt>InheritableAttr</tt> means that the attribute also applies
1740to subsequent declarations of the same name.</p>
1741
1742<p><tt>Spellings</tt> lists the strings that can appear in
1743<tt>__attribute__((here))</tt> or <tt>[[here]]</tt>. All such strings
David Blaikie5090e9f2011-10-18 05:49:30 +0000[diff] [blame]1744will be synonymous. If you want to allow the <tt>[[]]</tt> C++11
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +0000[diff] [blame]1745syntax, you have to define a list of <tt>Namespaces</tt>, which will
1746let users write <tt>[[namespace:spelling]]</tt>. Using the empty
1747string for a namespace will allow users to write just the spelling
1748with no "<tt>:</tt>".</p>
1749
1750<p><tt>Subjects</tt> restricts what kinds of AST node to which this attribute
1751can appertain (roughly, attach).</p>
1752
1753<p><tt>Args</tt> names the arguments the attribute takes, in order. If
1754<tt>Args</tt> is <tt>[StringArgument&lt;"Arg1">, IntArgument&lt;"Arg2">]</tt>
1755then <tt>__attribute__((myattribute("Hello", 3)))</tt> will be a valid use.</p>
1756
1757<h4>Boilerplate</h4>
1758
1759<p>Add an element to the <tt>AttributeList::Kind</tt> enum in <a
1760href="http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Sema/AttributeList.h?view=markup">include/clang/Sema/AttributeList.h</a>
1761named <tt>AT_lower_with_underscores</tt>. That is, a CamelCased
1762<tt>AttributeName</tt> in <tt>Attr.td</tt> name should become
1763<tt>AT_attribute_name</tt>.</p>
1764
1765<p>Add a case to the <tt>StringSwitch</tt> in <tt>AttributeList::getKind()</tt>
1766in <a
1767href="http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/AttributeList.cpp?view=markup">lib/Sema/AttributeList.cpp</a>
1768for each spelling of your attribute. Less common attributes should come toward
1769the end of that list.</p>
1770
1771<p>Write a new <tt>HandleYourAttr()</tt> function in <a
1772href="http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaDeclAttr.cpp?view=markup">lib/Sema/SemaDeclAttr.cpp</a>,
1773and add a case to the switch in <tt>ProcessNonInheritableDeclAttr()</tt> or
1774<tt>ProcessInheritableDeclAttr()</tt> forwarding to it.</p>
1775
1776<p>If your attribute causes extra warnings to fire, define a <tt>DiagGroup</tt>
1777in <a
1778href="http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticGroups.td?view=markup">include/clang/Basic/DiagnosticGroups.td</a>
1779named after the attribute's <tt>Spelling</tt> with "_"s replaced by "-"s. If
1780you're only defining one diagnostic, you can skip <tt>DiagnosticGroups.td</tt>
1781and use <tt>InGroup&lt;DiagGroup&lt;"your-attribute">></tt> directly in <a
1782href="http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td?view=markup">DiagnosticSemaKinds.td</a></p>
1783
1784<h4>The meat of your attribute</h4>
1785
1786<p>Find an appropriate place in Clang to do whatever your attribute needs to do.
1787Check for the attribute's presence using <tt>Decl::getAttr&lt;YourAttr>()</tt>.</p>
1788
1789<p>Update the <a href="LanguageExtensions.html">Clang Language Extensions</a>
1790document to describe your new attribute.</p>
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1791
Douglas Gregor1f634c62011-09-30 21:32:37 +0000[diff] [blame]1792<!-- ======================================================================= -->
1793<h3 id="AddingExprStmt">How to add an expression or statement</h3>
1794<!-- ======================================================================= -->
1795
1796<p>Expressions and statements are one of the most fundamental constructs within a
1797compiler, because they interact with many different parts of the AST,
1798semantic analysis, and IR generation. Therefore, adding a new
1799expression or statement kind into Clang requires some care. The following list
1800details the various places in Clang where an expression or statement needs to be
1801introduced, along with patterns to follow to ensure that the new
1802expression or statement works well across all of the C languages. We
1803focus on expressions, but statements are similar.</p>
1804
1805<ol>
1806 <li>Introduce parsing actions into the parser. Recursive-descent
1807 parsing is mostly self-explanatory, but there are a few things that
1808 are worth keeping in mind:
1809 <ul>
1810 <li>Keep as much source location information as possible! You'll
1811 want it later to produce great diagnostics and support Clang's
1812 various features that map between source code and the AST.</li>
1813 <li>Write tests for all of the "bad" parsing cases, to make sure
1814 your recovery is good. If you have matched delimiters (e.g.,
1815 parentheses, square brackets, etc.), use
Douglas Gregor4a8dfb52011-10-12 16:37:45 +0000[diff] [blame]1816 <tt>Parser::BalancedDelimiterTracker</tt> to give nice diagnostics when
Douglas Gregor1f634c62011-09-30 21:32:37 +0000[diff] [blame]1817 things go wrong.</li>
1818 </ul>
1819 </li>
1820
1821 <li>Introduce semantic analysis actions into <tt>Sema</tt>. Semantic
1822 analysis should always involve two functions: an <tt>ActOnXXX</tt>
1823 function that will be called directly from the parser, and a
1824 <tt>BuildXXX</tt> function that performs the actual semantic
1825 analysis and will (eventually!) build the AST node. It's fairly
1826 common for the <tt>ActOnCXX</tt> function to do very little (often
1827 just some minor translation from the parser's representation to
1828 <tt>Sema</tt>'s representation of the same thing), but the separation
1829 is still important: C++ template instantiation, for example,
1830 should always call the <tt>BuildXXX</tt> variant. Several notes on
1831 semantic analysis before we get into construction of the AST:
1832 <ul>
1833 <li>Your expression probably involves some types and some
1834 subexpressions. Make sure to fully check that those types, and the
1835 types of those subexpressions, meet your expectations. Add
1836 implicit conversions where necessary to make sure that all of the
1837 types line up exactly the way you want them. Write extensive tests
1838 to check that you're getting good diagnostics for mistakes and
1839 that you can use various forms of subexpressions with your
1840 expression.</li>
1841 <li>When type-checking a type or subexpression, make sure to first
1842 check whether the type is "dependent"
1843 (<tt>Type::isDependentType()</tt>) or whether a subexpression is
1844 type-dependent (<tt>Expr::isTypeDependent()</tt>). If any of these
1845 return true, then you're inside a template and you can't do much
1846 type-checking now. That's normal, and your AST node (when you get
1847 there) will have to deal with this case. At this point, you can
1848 write tests that use your expression within templates, but don't
1849 try to instantiate the templates.</li>
1850 <li>For each subexpression, be sure to call
1851 <tt>Sema::CheckPlaceholderExpr()</tt> to deal with "weird"
1852 expressions that don't behave well as subexpressions. Then,
1853 determine whether you need to perform
1854 lvalue-to-rvalue conversions
1855 (<tt>Sema::DefaultLvalueConversion</tt>e) or
1856 the usual unary conversions
1857 (<tt>Sema::UsualUnaryConversions</tt>), for places where the
1858 subexpression is producing a value you intend to use.</li>
1859 <li>Your <tt>BuildXXX</tt> function will probably just return
1860 <tt>ExprError()</tt> at this point, since you don't have an AST.
1861 That's perfectly fine, and shouldn't impact your testing.</li>
1862 </ul>
1863 </li>
1864
1865 <li>Introduce an AST node for your new expression. This starts with
1866 declaring the node in <tt>include/Basic/StmtNodes.td</tt> and
1867 creating a new class for your expression in the appropriate
1868 <tt>include/AST/Expr*.h</tt> header. It's best to look at the class
1869 for a similar expression to get ideas, and there are some specific
1870 things to watch for:
1871 <ul>
1872 <li>If you need to allocate memory, use the <tt>ASTContext</tt>
1873 allocator to allocate memory. Never use raw <tt>malloc</tt> or
1874 <tt>new</tt>, and never hold any resources in an AST node, because
1875 the destructor of an AST node is never called.</li>
1876
1877 <li>Make sure that <tt>getSourceRange()</tt> covers the exact
1878 source range of your expression. This is needed for diagnostics
1879 and for IDE support.</li>
1880
1881 <li>Make sure that <tt>children()</tt> visits all of the
1882 subexpressions. This is important for a number of features (e.g., IDE
1883 support, C++ variadic templates). If you have sub-types, you'll
1884 also need to visit those sub-types in the
1885 <tt>RecursiveASTVisitor</tt>.</li>
1886
1887 <li>Add printing support (<tt>StmtPrinter.cpp</tt>) and dumping
1888 support (<tt>StmtDumper.cpp</tt>) for your expression.</li>
1889
1890 <li>Add profiling support (<tt>StmtProfile.cpp</tt>) for your AST
1891 node, noting the distinguishing (non-source location)
1892 characteristics of an instance of your expression. Omitting this
1893 step will lead to hard-to-diagnose failures regarding matching of
1894 template declarations.</li>
1895 </ul>
1896 </li>
1897
1898 <li>Teach semantic analysis to build your AST node! At this point,
1899 you can wire up your <tt>Sema::BuildXXX</tt> function to actually
1900 create your AST. A few things to check at this point:
1901 <ul>
1902 <li>If your expression can construct a new C++ class or return a
1903 new Objective-C object, be sure to update and then call
1904 <tt>Sema::MaybeBindToTemporary</tt> for your just-created AST node
1905 to be sure that the object gets properly destructed. An easy way
1906 to test this is to return a C++ class with a private destructor:
1907 semantic analysis should flag an error here with the attempt to
1908 call the destructor.</li>
1909 <li>Inspect the generated AST by printing it using <tt>clang -cc1
1910 -ast-print</tt>, to make sure you're capturing all of the
1911 important information about how the AST was written.</li>
1912 <li>Inspect the generated AST under <tt>clang -cc1 -ast-dump</tt>
1913 to verify that all of the types in the generated AST line up the
1914 way you want them. Remember that clients of the AST should never
1915 have to "think" to understand what's going on. For example, all
1916 implicit conversions should show up explicitly in the AST.</li>
1917 <li>Write tests that use your expression as a subexpression of
1918 other, well-known expressions. Can you call a function using your
1919 expression as an argument? Can you use the ternary operator?</li>
1920 </ul>
1921 </li>
1922
1923 <li>Teach code generation to create IR to your AST node. This step
1924 is the first (and only) that requires knowledge of LLVM IR. There
1925 are several things to keep in mind:
1926 <ul>
1927 <li>Code generation is separated into scalar/aggregate/complex and
1928 lvalue/rvalue paths, depending on what kind of result your
1929 expression produces. On occasion, this requires some careful
1930 factoring of code to avoid duplication.</li>
1931
1932 <li><tt>CodeGenFunction</tt> contains functions
1933 <tt>ConvertType</tt> and <tt>ConvertTypeForMem</tt> that convert
1934 Clang's types (<tt>clang::Type*</tt> or <tt>clang::QualType</tt>)
1935 to LLVM types.
1936 Use the former for values, and the later for memory locations:
1937 test with the C++ "bool" type to check this. If you find
1938 that you are having to use LLVM bitcasts to make
1939 the subexpressions of your expression have the type that your
1940 expression expects, STOP! Go fix semantic analysis and the AST so
1941 that you don't need these bitcasts.</li>
1942
1943 <li>The <tt>CodeGenFunction</tt> class has a number of helper
1944 functions to make certain operations easy, such as generating code
1945 to produce an lvalue or an rvalue, or to initialize a memory
1946 location with a given value. Prefer to use these functions rather
1947 than directly writing loads and stores, because these functions
1948 take care of some of the tricky details for you (e.g., for
1949 exceptions).</li>
1950
1951 <li>If your expression requires some special behavior in the event
1952 of an exception, look at the <tt>push*Cleanup</tt> functions in
1953 <tt>CodeGenFunction</tt> to introduce a cleanup. You shouldn't
1954 have to deal with exception-handling directly.</li>
1955
1956 <li>Testing is extremely important in IR generation. Use <tt>clang
1957 -cc1 -emit-llvm</tt> and <a
1958 href="http://llvm.org/cmds/FileCheck.html">FileCheck</a> to verify
1959 that you're generating the right IR.</li>
1960 </ul>
1961 </li>
1962
1963 <li>Teach template instantiation how to cope with your AST
1964 node, which requires some fairly simple code:
1965 <ul>
1966 <li>Make sure that your expression's constructor properly
1967 computes the flags for type dependence (i.e., the type your
1968 expression produces can change from one instantiation to the
1969 next), value dependence (i.e., the constant value your expression
1970 produces can change from one instantiation to the next),
Douglas Gregord1cb2dc2011-10-14 00:54:15 +0000[diff] [blame]1971 instantiation dependence (i.e., a template parameter occurs
Douglas Gregor1f634c62011-09-30 21:32:37 +0000[diff] [blame]1972 anywhere in your expression), and whether your expression contains
1973 a parameter pack (for variadic templates). Often, computing these
1974 flags just means combining the results from the various types and
1975 subexpressions.</li>
1976
1977 <li>Add <tt>TransformXXX</tt> and <tt>RebuildXXX</tt> functions to
1978 the
1979 <tt>TreeTransform</tt> class template in <tt>Sema</tt>.
1980 <tt>TransformXXX</tt> should (recursively) transform all of the
1981 subexpressions and types
1982 within your expression, using <tt>getDerived().TransformYYY</tt>.
1983 If all of the subexpressions and types transform without error, it
1984 will then call the <tt>RebuildXXX</tt> function, which will in
1985 turn call <tt>getSema().BuildXXX</tt> to perform semantic analysis
1986 and build your expression.</li>
1987
1988 <li>To test template instantiation, take those tests you wrote to
1989 make sure that you were type checking with type-dependent
1990 expressions and dependent types (from step #2) and instantiate
1991 those templates with various types, some of which type-check and
1992 some that don't, and test the error messages in each case.</li>
1993 </ul>
1994 </li>
1995
1996 <li>There are some "extras" that make other features work better.
1997 It's worth handling these extras to give your expression complete
1998 integration into Clang:
1999 <ul>
2000 <li>Add code completion support for your expression in
2001 <tt>SemaCodeComplete.cpp</tt>.</li>
2002
2003 <li>If your expression has types in it, or has any "interesting"
2004 features other than subexpressions, extend libclang's
2005 <tt>CursorVisitor</tt> to provide proper visitation for your
2006 expression, enabling various IDE features such as syntax
2007 highlighting, cross-referencing, and so on. The
2008 <tt>c-index-test</tt> helper program can be used to test these
2009 features.</li>
2010 </ul>
2011 </li>
2012</ol>
2013
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]2014</div>
2015</body>
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]2016</html>