blob: 1ae1968299e780a769374f87b12c03be22ff62e7 [file] [log] [blame]
Benjamin Kramer665a8dc2012-01-15 15:26:07 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
Ted Kremenek17a295d2008-06-11 06:19:49 +00003<html>
4<head>
Chris Lattner552de0a2008-11-23 08:16:56 +00005<title>"Clang" CFE Internals Manual</title>
Benjamin Kramer665a8dc2012-01-15 15:26:07 +00006<link type="text/css" rel="stylesheet" href="../menu.css">
7<link type="text/css" rel="stylesheet" href="../content.css">
Sebastian Redl68168562008-11-22 22:16:45 +00008<style type="text/css">
9td {
10 vertical-align: top;
11}
12</style>
Ted Kremenek17a295d2008-06-11 06:19:49 +000013</head>
14<body>
15
16<!--#include virtual="../menu.html.incl"-->
17
18<div id="content">
Chris Lattner86920d32007-07-31 05:42:17 +000019
Chris Lattner552de0a2008-11-23 08:16:56 +000020<h1>"Clang" CFE Internals Manual</h1>
Chris Lattner86920d32007-07-31 05:42:17 +000021
22<ul>
23<li><a href="#intro">Introduction</a></li>
Peter Collingbourne967c1182011-10-15 16:59:24 +000024<li><a href="#libsupport">LLVM Support Library</a></li>
Chris Lattner552de0a2008-11-23 08:16:56 +000025<li><a href="#libbasic">The Clang 'Basic' Library</a>
Chris Lattner86920d32007-07-31 05:42:17 +000026 <ul>
Chris Lattner62fd2782008-11-22 21:41:31 +000027 <li><a href="#Diagnostics">The Diagnostics Subsystem</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +000028 <li><a href="#SourceLocation">The SourceLocation and SourceManager
29 classes</a></li>
Douglas Gregor715c92a2010-10-27 16:02:28 +000030 <li><a href="#SourceRange">SourceRange and CharSourceRange</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +000031 </ul>
32</li>
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +000033<li><a href="#libdriver">The Driver Library</a>
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +000034</li>
Douglas Gregor32110df2009-05-20 00:16:32 +000035<li><a href="#pch">Precompiled Headers</a>
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +000036<li><a href="#libfrontend">The Frontend Library</a>
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +000037</li>
Chris Lattner86920d32007-07-31 05:42:17 +000038<li><a href="#liblex">The Lexer and Preprocessor Library</a>
39 <ul>
40 <li><a href="#Token">The Token class</a></li>
41 <li><a href="#Lexer">The Lexer class</a></li>
Chris Lattner3932fe02009-01-06 06:02:08 +000042 <li><a href="#AnnotationToken">Annotation Tokens</a></li>
Chris Lattner79281252008-03-09 02:27:26 +000043 <li><a href="#TokenLexer">The TokenLexer class</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +000044 <li><a href="#MultipleIncludeOpt">The MultipleIncludeOpt class</a></li>
45 </ul>
46</li>
47<li><a href="#libparse">The Parser Library</a>
Chris Lattner86920d32007-07-31 05:42:17 +000048</li>
49<li><a href="#libast">The AST Library</a>
50 <ul>
51 <li><a href="#Type">The Type class and its subclasses</a></li>
52 <li><a href="#QualType">The QualType class</a></li>
Douglas Gregor2e1cd422008-11-17 14:58:09 +000053 <li><a href="#DeclarationName">Declaration names</a></li>
Douglas Gregor074149e2009-01-05 19:45:36 +000054 <li><a href="#DeclContext">Declaration contexts</a>
55 <ul>
56 <li><a href="#Redeclarations">Redeclarations and Overloads</a></li>
57 <li><a href="#LexicalAndSemanticContexts">Lexical and Semantic
58 Contexts</a></li>
59 <li><a href="#TransparentContexts">Transparent Declaration Contexts</a></li>
60 <li><a href="#MultiDeclContext">Multiply-Defined Declaration Contexts</a></li>
61 </ul>
62 </li>
Ted Kremenek8bc05712007-10-10 23:01:43 +000063 <li><a href="#CFG">The CFG class</a></li>
Chris Lattner7bad1992008-11-16 21:48:07 +000064 <li><a href="#Constants">Constant Folding in the Clang AST</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +000065 </ul>
66</li>
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +000067<li><a href="#Howtos">Howto guides</a>
68 <ul>
69 <li><a href="#AddingAttributes">How to add an attribute</a></li>
Douglas Gregor1f634c62011-09-30 21:32:37 +000070 <li><a href="#AddingExprStmt">How to add a new expression or statement</a></li>
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +000071 </ul>
72</li>
Chris Lattner86920d32007-07-31 05:42:17 +000073</ul>
74
75
76<!-- ======================================================================= -->
77<h2 id="intro">Introduction</h2>
78<!-- ======================================================================= -->
79
80<p>This document describes some of the more important APIs and internal design
Chris Lattner552de0a2008-11-23 08:16:56 +000081decisions made in the Clang C front-end. The purpose of this document is to
Chris Lattner86920d32007-07-31 05:42:17 +000082both capture some of this high level information and also describe some of the
83design decisions behind it. This is meant for people interested in hacking on
Chris Lattner552de0a2008-11-23 08:16:56 +000084Clang, not for end-users. The description below is categorized by
Chris Lattner86920d32007-07-31 05:42:17 +000085libraries, and does not describe any of the clients of the libraries.</p>
86
87<!-- ======================================================================= -->
Peter Collingbourne967c1182011-10-15 16:59:24 +000088<h2 id="libsupport">LLVM Support Library</h2>
Chris Lattner86920d32007-07-31 05:42:17 +000089<!-- ======================================================================= -->
90
Peter Collingbourne967c1182011-10-15 16:59:24 +000091<p>The LLVM libsupport library provides many underlying libraries and
92<a href="http://llvm.org/docs/ProgrammersManual.html">data-structures</a>,
93including command line option processing, various containers and a system
94abstraction layer, which is used for file system access.</p>
Chris Lattner86920d32007-07-31 05:42:17 +000095
96<!-- ======================================================================= -->
Chris Lattner552de0a2008-11-23 08:16:56 +000097<h2 id="libbasic">The Clang 'Basic' Library</h2>
Chris Lattner86920d32007-07-31 05:42:17 +000098<!-- ======================================================================= -->
99
100<p>This library certainly needs a better name. The 'basic' library contains a
101number of low-level utilities for tracking and manipulating source buffers,
102locations within the source buffers, diagnostics, tokens, target abstraction,
103and information about the subset of the language being compiled for.</p>
104
105<p>Part of this infrastructure is specific to C (such as the TargetInfo class),
106other parts could be reused for other non-C-based languages (SourceLocation,
107SourceManager, Diagnostics, FileManager). When and if there is future demand
108we can figure out if it makes sense to introduce a new library, move the general
109classes somewhere else, or introduce some other solution.</p>
110
111<p>We describe the roles of these classes in order of their dependencies.</p>
112
Chris Lattner62fd2782008-11-22 21:41:31 +0000113
114<!-- ======================================================================= -->
115<h3 id="Diagnostics">The Diagnostics Subsystem</h3>
116<!-- ======================================================================= -->
117
118<p>The Clang Diagnostics subsystem is an important part of how the compiler
119communicates with the human. Diagnostics are the warnings and errors produced
120when the code is incorrect or dubious. In Clang, each diagnostic produced has
Sebastian Redl9bc2a992010-07-07 23:42:27 +0000121(at the minimum) a unique ID, an English translation associated with it, a <a
122href="#SourceLocation">SourceLocation</a> to "put the caret", and a severity (e.g.
Chris Lattner62fd2782008-11-22 21:41:31 +0000123<tt>WARNING</tt> or <tt>ERROR</tt>). They can also optionally include a number
124of arguments to the dianostic (which fill in "%0"'s in the string) as well as a
125number of source ranges that related to the diagnostic.</p>
126
Chris Lattner552de0a2008-11-23 08:16:56 +0000127<p>In this section, we'll be giving examples produced by the Clang command line
Chris Lattner62fd2782008-11-22 21:41:31 +0000128driver, but diagnostics can be <a href="#DiagnosticClient">rendered in many
129different ways</a> depending on how the DiagnosticClient interface is
Sebastian Redl9bc2a992010-07-07 23:42:27 +0000130implemented. A representative example of a diagnostic is:</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000131
132<pre>
133t.c:38:15: error: invalid operands to binary expression ('int *' and '_Complex float')
Benjamin Kramer665a8dc2012-01-15 15:26:07 +0000134 <span style="color:darkgreen">P = (P-42) + Gamma*4;</span>
135 <span style="color:blue">~~~~~~ ^ ~~~~~~~</span>
Chris Lattner62fd2782008-11-22 21:41:31 +0000136</pre>
137
138<p>In this example, you can see the English translation, the severity (error),
139you can see the source location (the caret ("^") and file/line/column info),
140the source ranges "~~~~", arguments to the diagnostic ("int*" and "_Complex
141float"). You'll have to believe me that there is a unique ID backing the
142diagnostic :).</p>
143
144<p>Getting all of this to happen has several steps and involves many moving
145pieces, this section describes them and talks about best practices when adding
146a new diagnostic.</p>
147
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000148<!-- ============================= -->
149<h4>The Diagnostic*Kinds.td files</h4>
150<!-- ============================= -->
Chris Lattner62fd2782008-11-22 21:41:31 +0000151
Chris Lattner4c50b692010-05-01 17:35:19 +0000152<p>Diagnostics are created by adding an entry to one of the <tt>
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000153clang/Basic/Diagnostic*Kinds.td</tt> files, depending on what library will
154be using it. From this file, tblgen generates the unique ID of the diagnostic,
155the severity of the diagnostic and the English translation + format string.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000156
157<p>There is little sanity with the naming of the unique ID's right now. Some
158start with err_, warn_, ext_ to encode the severity into the name. Since the
159enum is referenced in the C++ code that produces the diagnostic, it is somewhat
160useful for it to be reasonably short.</p>
161
162<p>The severity of the diagnostic comes from the set {<tt>NOTE</tt>,
163<tt>WARNING</tt>, <tt>EXTENSION</tt>, <tt>EXTWARN</tt>, <tt>ERROR</tt>}. The
164<tt>ERROR</tt> severity is used for diagnostics indicating the program is never
165acceptable under any circumstances. When an error is emitted, the AST for the
166input code may not be fully built. The <tt>EXTENSION</tt> and <tt>EXTWARN</tt>
167severities are used for extensions to the language that Clang accepts. This
168means that Clang fully understands and can represent them in the AST, but we
169produce diagnostics to tell the user their code is non-portable. The difference
170is that the former are ignored by default, and the later warn by default. The
171<tt>WARNING</tt> severity is used for constructs that are valid in the currently
172selected source language but that are dubious in some way. The <tt>NOTE</tt>
Daniel Dunbar426b8632009-02-17 15:49:03 +0000173level is used to staple more information onto previous diagnostics.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000174
175<p>These <em>severities</em> are mapped into a smaller set (the
176Diagnostic::Level enum, {<tt>Ignored</tt>, <tt>Note</tt>, <tt>Warning</tt>,
Chris Lattner0aad2972009-02-05 22:49:08 +0000177<tt>Error</tt>, <tt>Fatal</tt> }) of output <em>levels</em> by the diagnostics
Chris Lattnera180fdd2009-02-17 07:07:29 +0000178subsystem based on various configuration options. Clang internally supports a
179fully fine grained mapping mechanism that allows you to map almost any
180diagnostic to the output level that you want. The only diagnostics that cannot
181be mapped are <tt>NOTE</tt>s, which always follow the severity of the previously
182emitted diagnostic and <tt>ERROR</tt>s, which can only be mapped to
183<tt>Fatal</tt> (it is not possible to turn an error into a warning,
184for example).</p>
185
186<p>Diagnostic mappings are used in many ways. For example, if the user
187specifies <tt>-pedantic</tt>, <tt>EXTENSION</tt> maps to <tt>Warning</tt>, if
188they specify <tt>-pedantic-errors</tt>, it turns into <tt>Error</tt>. This is
189used to implement options like <tt>-Wunused_macros</tt>, <tt>-Wundef</tt> etc.
190</p>
191
192<p>
193Mapping to <tt>Fatal</tt> should only be used for diagnostics that are
194considered so severe that error recovery won't be able to recover sensibly from
195them (thus spewing a ton of bogus errors). One example of this class of error
196are failure to #include a file.
197</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000198
199<!-- ================= -->
200<h4>The Format String</h4>
201<!-- ================= -->
202
203<p>The format string for the diagnostic is very simple, but it has some power.
204It takes the form of a string in English with markers that indicate where and
205how arguments to the diagnostic are inserted and formatted. For example, here
206are some simple format strings:</p>
207
208<pre>
209 "binary integer literals are an extension"
210 "format string contains '\\0' within the string body"
211 "more '<b>%%</b>' conversions than data arguments"
Chris Lattner545b3682008-11-23 20:27:13 +0000212 "invalid operands to binary expression (<b>%0</b> and <b>%1</b>)"
Chris Lattner62fd2782008-11-22 21:41:31 +0000213 "overloaded '<b>%0</b>' must be a <b>%select{unary|binary|unary or binary}2</b> operator"
214 " (has <b>%1</b> parameter<b>%s1</b>)"
215</pre>
216
217<p>These examples show some important points of format strings. You can use any
218 plain ASCII character in the diagnostic string except "%" without a problem,
219 but these are C strings, so you have to use and be aware of all the C escape
220 sequences (as in the second example). If you want to produce a "%" in the
221 output, use the "%%" escape sequence, like the third diagnostic. Finally,
Chris Lattner552de0a2008-11-23 08:16:56 +0000222 Clang uses the "%...[digit]" sequences to specify where and how arguments to
Chris Lattner62fd2782008-11-22 21:41:31 +0000223 the diagnostic are formatted.</p>
224
225<p>Arguments to the diagnostic are numbered according to how they are specified
226 by the C++ code that <a href="#producingdiag">produces them</a>, and are
227 referenced by <tt>%0</tt> .. <tt>%9</tt>. If you have more than 10 arguments
Chris Lattner552de0a2008-11-23 08:16:56 +0000228 to your diagnostic, you are doing something wrong :). Unlike printf, there
Chris Lattner62fd2782008-11-22 21:41:31 +0000229 is no requirement that arguments to the diagnostic end up in the output in
230 the same order as they are specified, you could have a format string with
231 <tt>"%1 %0"</tt> that swaps them, for example. The text in between the
232 percent and digit are formatting instructions. If there are no instructions,
233 the argument is just turned into a string and substituted in.</p>
234
235<p>Here are some "best practices" for writing the English format string:</p>
236
237<ul>
238<li>Keep the string short. It should ideally fit in the 80 column limit of the
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000239 <tt>DiagnosticKinds.td</tt> file. This avoids the diagnostic wrapping when
Chris Lattner62fd2782008-11-22 21:41:31 +0000240 printed, and forces you to think about the important point you are conveying
241 with the diagnostic.</li>
242<li>Take advantage of location information. The user will be able to see the
243 line and location of the caret, so you don't need to tell them that the
244 problem is with the 4th argument to the function: just point to it.</li>
245<li>Do not capitalize the diagnostic string, and do not end it with a
246 period.</li>
247<li>If you need to quote something in the diagnostic string, use single
248 quotes.</li>
249</ul>
250
251<p>Diagnostics should never take random English strings as arguments: you
252shouldn't use <tt>"you have a problem with %0"</tt> and pass in things like
253<tt>"your argument"</tt> or <tt>"your return value"</tt> as arguments. Doing
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000254this prevents <a href="#translation">translating</a> the Clang diagnostics to
Chris Lattner62fd2782008-11-22 21:41:31 +0000255other languages (because they'll get random English words in their otherwise
256localized diagnostic). The exceptions to this are C/C++ language keywords
257(e.g. auto, const, mutable, etc) and C/C++ operators (<tt>/=</tt>). Note
258that things like "pointer" and "reference" are not keywords. On the other
259hand, you <em>can</em> include anything that comes from the user's source code,
Chris Lattner552de0a2008-11-23 08:16:56 +0000260including variable names, types, labels, etc. The 'select' format can be
261used to achieve this sort of thing in a localizable way, see below.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000262
263<!-- ==================================== -->
Benjamin Kramer665a8dc2012-01-15 15:26:07 +0000264<h4>Formatting a Diagnostic Argument</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000265<!-- ==================================== -->
266
267<p>Arguments to diagnostics are fully typed internally, and come from a couple
268different classes: integers, types, names, and random strings. Depending on
269the class of the argument, it can be optionally formatted in different ways.
270This gives the DiagnosticClient information about what the argument means
271without requiring it to use a specific presentation (consider this MVC for
272Clang :).</p>
273
274<p>Here are the different diagnostic argument formats currently supported by
275Clang:</p>
276
277<table>
278<tr><td colspan="2"><b>"s" format</b></td></tr>
279<tr><td>Example:</td><td><tt>"requires %1 parameter%s1"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000280<tr><td>Class:</td><td>Integers</td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000281<tr><td>Description:</td><td>This is a simple formatter for integers that is
282 useful when producing English diagnostics. When the integer is 1, it prints
283 as nothing. When the integer is not 1, it prints as "s". This allows some
Chris Lattner627b7052008-11-23 00:28:33 +0000284 simple grammatical forms to be to be handled correctly, and eliminates the
285 need to use gross things like <tt>"requires %1 parameter(s)"</tt>.</td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000286
287<tr><td colspan="2"><b>"select" format</b></td></tr>
288<tr><td>Example:</td><td><tt>"must be a %select{unary|binary|unary or binary}2
289 operator"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000290<tr><td>Class:</td><td>Integers</td></tr>
John McCall3a47e232010-01-14 19:12:17 +0000291<tr><td>Description:</td><td><p>This format specifier is used to merge multiple
Chris Lattnercc543342008-11-22 23:50:47 +0000292 related diagnostics together into one common one, without requiring the
Chris Lattner552de0a2008-11-23 08:16:56 +0000293 difference to be specified as an English string argument. Instead of
Chris Lattnercc543342008-11-22 23:50:47 +0000294 specifying the string, the diagnostic gets an integer argument and the
295 format string selects the numbered option. In this case, the "%2" value
296 must be an integer in the range [0..2]. If it is 0, it prints 'unary', if
297 it is 1 it prints 'binary' if it is 2, it prints 'unary or binary'. This
298 allows other language translations to substitute reasonable words (or entire
299 phrases) based on the semantics of the diagnostic instead of having to do
John McCall3a47e232010-01-14 19:12:17 +0000300 things textually.</p>
301 <p>The selected string does undergo formatting.</p></td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000302
303<tr><td colspan="2"><b>"plural" format</b></td></tr>
Sebastian Redl68168562008-11-22 22:16:45 +0000304<tr><td>Example:</td><td><tt>"you have %1 %plural{1:mouse|:mice}1 connected to
305 your computer"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000306<tr><td>Class:</td><td>Integers</td></tr>
Sebastian Redl68168562008-11-22 22:16:45 +0000307<tr><td>Description:</td><td><p>This is a formatter for complex plural forms.
308 It is designed to handle even the requirements of languages with very
309 complex plural forms, as many Baltic languages have. The argument consists
310 of a series of expression/form pairs, separated by ':', where the first form
311 whose expression evaluates to true is the result of the modifier.</p>
312 <p>An expression can be empty, in which case it is always true. See the
313 example at the top. Otherwise, it is a series of one or more numeric
314 conditions, separated by ','. If any condition matches, the expression
315 matches. Each numeric condition can take one of three forms.</p>
316 <ul>
317 <li>number: A simple decimal number matches if the argument is the same
Chris Lattner627b7052008-11-23 00:28:33 +0000318 as the number. Example: <tt>"%plural{1:mouse|:mice}4"</tt></li>
Sebastian Redl68168562008-11-22 22:16:45 +0000319 <li>range: A range in square brackets matches if the argument is within
Chris Lattner552de0a2008-11-23 08:16:56 +0000320 the range. Then range is inclusive on both ends. Example:
Chris Lattner627b7052008-11-23 00:28:33 +0000321 <tt>"%plural{0:none|1:one|[2,5]:some|:many}2"</tt></li>
322 <li>modulo: A modulo operator is followed by a number, and
323 equals sign and either a number or a range. The tests are the
324 same as for plain
Sebastian Redl68168562008-11-22 22:16:45 +0000325 numbers and ranges, but the argument is taken modulo the number first.
Chris Lattner627b7052008-11-23 00:28:33 +0000326 Example: <tt>"%plural{%100=0:even hundred|%100=[1,50]:lower half|:everything
327 else}1"</tt></li>
Sebastian Redl68168562008-11-22 22:16:45 +0000328 </ul>
329 <p>The parser is very unforgiving. A syntax error, even whitespace, will
330 abort, as will a failure to match the argument against any
331 expression.</p></td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000332
John McCall3a47e232010-01-14 19:12:17 +0000333<tr><td colspan="2"><b>"ordinal" format</b></td></tr>
334<tr><td>Example:</td><td><tt>"ambiguity in %ordinal0 argument"</tt></td></tr>
335<tr><td>Class:</td><td>Integers</td></tr>
336<tr><td>Description:</td><td><p>This is a formatter which represents the
337 argument number as an ordinal: the value <tt>1</tt> becomes <tt>1st</tt>,
338 <tt>3</tt> becomes <tt>3rd</tt>, and so on. Values less than <tt>1</tt>
339 are not supported.</p>
340 <p>This formatter is currently hard-coded to use English ordinals.</p></td></tr>
341
Chris Lattner077bf5e2008-11-24 03:33:13 +0000342<tr><td colspan="2"><b>"objcclass" format</b></td></tr>
343<tr><td>Example:</td><td><tt>"method %objcclass0 not found"</tt></td></tr>
344<tr><td>Class:</td><td>DeclarationName</td></tr>
345<tr><td>Description:</td><td><p>This is a simple formatter that indicates the
346 DeclarationName corresponds to an Objective-C class method selector. As
347 such, it prints the selector with a leading '+'.</p></td></tr>
348
349<tr><td colspan="2"><b>"objcinstance" format</b></td></tr>
350<tr><td>Example:</td><td><tt>"method %objcinstance0 not found"</tt></td></tr>
351<tr><td>Class:</td><td>DeclarationName</td></tr>
352<tr><td>Description:</td><td><p>This is a simple formatter that indicates the
353 DeclarationName corresponds to an Objective-C instance method selector. As
354 such, it prints the selector with a leading '-'.</p></td></tr>
355
Douglas Gregor47b9a1c2009-02-04 17:27:36 +0000356<tr><td colspan="2"><b>"q" format</b></td></tr>
357<tr><td>Example:</td><td><tt>"candidate found by name lookup is %q0"</tt></td></tr>
358<tr><td>Class:</td><td>NamedDecl*</td></tr>
359<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>
360
Chris Lattner62fd2782008-11-22 21:41:31 +0000361</table>
362
Chris Lattnercc543342008-11-22 23:50:47 +0000363<p>It is really easy to add format specifiers to the Clang diagnostics system,
Chris Lattner552de0a2008-11-23 08:16:56 +0000364but they should be discussed before they are added. If you are creating a lot
365of repetitive diagnostics and/or have an idea for a useful formatter, please
366bring it up on the cfe-dev mailing list.</p>
Chris Lattnercc543342008-11-22 23:50:47 +0000367
Chris Lattner62fd2782008-11-22 21:41:31 +0000368<!-- ===================================================== -->
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000369<h4 id="producingdiag">Producing the Diagnostic</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000370<!-- ===================================================== -->
371
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000372<p>Now that you've created the diagnostic in the DiagnosticKinds.td file, you
Chris Lattner552de0a2008-11-23 08:16:56 +0000373need to write the code that detects the condition in question and emits the
374new diagnostic. Various components of Clang (e.g. the preprocessor, Sema,
Chris Lattner627b7052008-11-23 00:28:33 +0000375etc) provide a helper function named "Diag". It creates a diagnostic and
376accepts the arguments, ranges, and other information that goes along with
377it.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000378
Chris Lattner552de0a2008-11-23 08:16:56 +0000379<p>For example, the binary expression error comes from code like this:</p>
Chris Lattner627b7052008-11-23 00:28:33 +0000380
381<pre>
382 if (various things that are bad)
383 Diag(Loc, diag::err_typecheck_invalid_operands)
384 &lt;&lt; lex-&gt;getType() &lt;&lt; rex-&gt;getType()
385 &lt;&lt; lex-&gt;getSourceRange() &lt;&lt; rex-&gt;getSourceRange();
386</pre>
387
388<p>This shows that use of the Diag method: they take a location (a <a
389href="#SourceLocation">SourceLocation</a> object) and a diagnostic enum value
Chris Lattnercc2ac1e2011-02-14 06:42:50 +0000390(which matches the name from DiagnosticKinds.td). If the diagnostic takes
Chris Lattner627b7052008-11-23 00:28:33 +0000391arguments, they are specified with the &lt;&lt; operator: the first argument
392becomes %0, the second becomes %1, etc. The diagnostic interface allows you to
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000393specify arguments of many different types, including <tt>int</tt> and
394<tt>unsigned</tt> for integer arguments, <tt>const char*</tt> and
395<tt>std::string</tt> for string arguments, <tt>DeclarationName</tt> and
396<tt>const IdentifierInfo*</tt> for names, <tt>QualType</tt> for types, etc.
397SourceRanges are also specified with the &lt;&lt; operator, but do not have a
398specific ordering requirement.</p>
Chris Lattner627b7052008-11-23 00:28:33 +0000399
400<p>As you can see, adding and producing a diagnostic is pretty straightforward.
401The hard part is deciding exactly what you need to say to help the user, picking
402a suitable wording, and providing the information needed to format it correctly.
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000403The good news is that the call site that issues a diagnostic should be
404completely independent of how the diagnostic is formatted and in what language
405it is rendered.
Chris Lattner627b7052008-11-23 00:28:33 +0000406</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000407
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000408<!-- ==================================================== -->
Peter Collingbourne38448d32011-03-21 01:45:18 +0000409<h4 id="fix-it-hints">Fix-It Hints</h4>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000410<!-- ==================================================== -->
411
412<p>In some cases, the front end emits diagnostics when it is clear
413that some small change to the source code would fix the problem. For
414example, a missing semicolon at the end of a statement or a use of
Chris Lattner34c05332009-02-27 19:31:12 +0000415deprecated syntax that is easily rewritten into a more modern form.
416Clang tries very hard to emit the diagnostic and recover gracefully
417in these and other cases.</p>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000418
Peter Collingbourne38448d32011-03-21 01:45:18 +0000419<p>However, for these cases where the fix is obvious, the diagnostic
420can be annotated with a hint (referred to as a "fix-it hint") that
421describes how to change the code referenced by the diagnostic to fix
422the problem. For example, it might add the missing semicolon at the
423end of the statement or rewrite the use of a deprecated construct
424into something more palatable. Here is one such example from the C++
425front end, where we warn about the right-shift operator changing
David Blaikie5090e9f2011-10-18 05:49:30 +0000426meaning from C++98 to C++11:</p>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000427
428<pre>
David Blaikie5090e9f2011-10-18 05:49:30 +0000429test.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 +0000430A&lt;100 &gt;&gt; 2&gt; *a;
431 ^
432 ( )
433</pre>
434
Peter Collingbourne38448d32011-03-21 01:45:18 +0000435<p>Here, the fix-it hint is suggesting that parentheses be added,
436and showing exactly where those parentheses would be inserted into the
437source code. The fix-it hints themselves describe what changes to make
438to the source code in an abstract manner, which the text diagnostic
439printer renders as a line of "insertions" below the caret line. <a
440href="#DiagnosticClient">Other diagnostic clients</a> might choose
441to render the code differently (e.g., as markup inline) or even give
442the user the ability to automatically fix the problem.</p>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000443
Nico Weber80039632012-06-04 21:56:14 +0000444<p>Fix-it hints on errors and warnings need to obey these rules:</p>
445
446<ul>
447<li>Since they are automatically applied if <code>-Xclang -fixit</code>
448is passed to the driver, they should only be used when it's very likely they
449match the user's intent.</li>
450<li>Clang must recover from the error or warning as if the fix-it had been
451applied.</li>
452</ul>
453
454<p>If a fix-it can't obey these rules, put the fix-it on a note. Fix-its on
455notes are not applied automatically.</p>
456
Peter Collingbourne38448d32011-03-21 01:45:18 +0000457<p>All fix-it hints are described by the <code>FixItHint</code> class,
458instances of which should be attached to the diagnostic using the
459&lt;&lt; operator in the same way that highlighted source ranges and
460arguments are passed to the diagnostic. Fix-it hints can be created
461with one of three constructors:</p>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000462
463<dl>
Peter Collingbourne38448d32011-03-21 01:45:18 +0000464 <dt><code>FixItHint::CreateInsertion(Loc, Code)</code></dt>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000465 <dd>Specifies that the given <code>Code</code> (a string) should be inserted
466 before the source location <code>Loc</code>.</dd>
467
Peter Collingbourne38448d32011-03-21 01:45:18 +0000468 <dt><code>FixItHint::CreateRemoval(Range)</code></dt>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000469 <dd>Specifies that the code in the given source <code>Range</code>
470 should be removed.</dd>
471
Peter Collingbourne38448d32011-03-21 01:45:18 +0000472 <dt><code>FixItHint::CreateReplacement(Range, Code)</code></dt>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000473 <dd>Specifies that the code in the given source <code>Range</code>
474 should be removed, and replaced with the given <code>Code</code> string.</dd>
475</dl>
476
Chris Lattner62fd2782008-11-22 21:41:31 +0000477<!-- ============================================================= -->
478<h4><a name="DiagnosticClient">The DiagnosticClient Interface</a></h4>
479<!-- ============================================================= -->
480
Chris Lattner627b7052008-11-23 00:28:33 +0000481<p>Once code generates a diagnostic with all of the arguments and the rest of
482the relevant information, Clang needs to know what to do with it. As previously
483mentioned, the diagnostic machinery goes through some filtering to map a
484severity onto a diagnostic level, then (assuming the diagnostic is not mapped to
485"<tt>Ignore</tt>") it invokes an object that implements the DiagnosticClient
486interface with the information.</p>
487
488<p>It is possible to implement this interface in many different ways. For
489example, the normal Clang DiagnosticClient (named 'TextDiagnosticPrinter') turns
490the arguments into strings (according to the various formatting rules), prints
491out the file/line/column information and the string, then prints out the line of
492code, the source ranges, and the caret. However, this behavior isn't required.
493</p>
494
495<p>Another implementation of the DiagnosticClient interface is the
Chris Lattner552de0a2008-11-23 08:16:56 +0000496'TextDiagnosticBuffer' class, which is used when Clang is in -verify mode.
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000497Instead of formatting and printing out the diagnostics, this implementation just
498captures and remembers the diagnostics as they fly by. Then -verify compares
Chris Lattner552de0a2008-11-23 08:16:56 +0000499the list of produced diagnostics to the list of expected ones. If they disagree,
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000500it prints out its own output.
Chris Lattner627b7052008-11-23 00:28:33 +0000501</p>
502
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000503<p>There are many other possible implementations of this interface, and this is
504why we prefer diagnostics to pass down rich structured information in arguments.
505For example, an HTML output might want declaration names be linkified to where
506they come from in the source. Another example is that a GUI might let you click
507on typedefs to expand them. This application would want to pass significantly
508more information about types through to the GUI than a simple flat string. The
509interface allows this to happen.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000510
511<!-- ====================================================== -->
512<h4><a name="translation">Adding Translations to Clang</a></h4>
513<!-- ====================================================== -->
514
Chris Lattner627b7052008-11-23 00:28:33 +0000515<p>Not possible yet! Diagnostic strings should be written in UTF-8, the client
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000516can translate to the relevant code page if needed. Each translation completely
517replaces the format string for the diagnostic.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000518
519
Chris Lattner86920d32007-07-31 05:42:17 +0000520<!-- ======================================================================= -->
521<h3 id="SourceLocation">The SourceLocation and SourceManager classes</h3>
522<!-- ======================================================================= -->
523
524<p>Strangely enough, the SourceLocation class represents a location within the
525source code of the program. Important design points include:</p>
526
527<ol>
528<li>sizeof(SourceLocation) must be extremely small, as these are embedded into
529 many AST nodes and are passed around often. Currently it is 32 bits.</li>
530<li>SourceLocation must be a simple value object that can be efficiently
531 copied.</li>
532<li>We should be able to represent a source location for any byte of any input
533 file. This includes in the middle of tokens, in whitespace, in trigraphs,
534 etc.</li>
535<li>A SourceLocation must encode the current #include stack that was active when
536 the location was processed. For example, if the location corresponds to a
537 token, it should contain the set of #includes active when the token was
538 lexed. This allows us to print the #include stack for a diagnostic.</li>
539<li>SourceLocation must be able to describe macro expansions, capturing both
540 the ultimate instantiation point and the source of the original character
541 data.</li>
542</ol>
543
544<p>In practice, the SourceLocation works together with the SourceManager class
Nick Lewycky77561e52010-05-26 21:48:10 +0000545to encode two pieces of information about a location: its spelling location
546and its instantiation location. For most tokens, these will be the same.
547However, for a macro expansion (or tokens that came from a _Pragma directive)
548these will describe the location of the characters corresponding to the token
549and the location where the token was used (i.e. the macro instantiation point
550or the location of the _Pragma itself).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000551
Chris Lattner552de0a2008-11-23 08:16:56 +0000552<p>The Clang front-end inherently depends on the location of a token being
Chris Lattner86920d32007-07-31 05:42:17 +0000553tracked correctly. If it is ever incorrect, the front-end may get confused and
554die. The reason for this is that the notion of the 'spelling' of a Token in
Chris Lattner552de0a2008-11-23 08:16:56 +0000555Clang depends on being able to find the original input characters for the token.
Chris Lattner18376dd2009-01-16 07:00:50 +0000556This concept maps directly to the "spelling location" for the token.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000557
Douglas Gregor715c92a2010-10-27 16:02:28 +0000558
559<!-- ======================================================================= -->
560<h3 id="SourceRange">SourceRange and CharSourceRange</h3>
561<!-- ======================================================================= -->
562<!-- mostly taken from
563 http://lists.cs.uiuc.edu/pipermail/cfe-dev/2010-August/010595.html -->
564
565<p>Clang represents most source ranges by [first, last], where first and last
566each point to the beginning of their respective tokens. For example
567consider the SourceRange of the following statement:</p>
568<pre>
569x = foo + bar;
570^first ^last
571</pre>
572
573<p>To map from this representation to a character-based
574representation, the 'last' location needs to be adjusted to point to
575(or past) the end of that token with either
576<code>Lexer::MeasureTokenLength()</code> or
Chris Lattner7ef5c272010-11-17 07:05:50 +0000577<code>Lexer::getLocForEndOfToken()</code>. For the rare cases
Douglas Gregor715c92a2010-10-27 16:02:28 +0000578where character-level source ranges information is needed we use
579the <code>CharSourceRange</code> class.</p>
580
581
Chris Lattner86920d32007-07-31 05:42:17 +0000582<!-- ======================================================================= -->
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000583<h2 id="libdriver">The Driver Library</h2>
584<!-- ======================================================================= -->
585
Ted Kremenekcfa8d572009-04-09 18:08:18 +0000586<p>The clang Driver and library are documented <a
Benjamin Kramer665a8dc2012-01-15 15:26:07 +0000587href="DriverInternals.html">here</a>.<p>
Ted Kremenekcfa8d572009-04-09 18:08:18 +0000588
589<!-- ======================================================================= -->
Douglas Gregor32110df2009-05-20 00:16:32 +0000590<h2 id="pch">Precompiled Headers</h2>
Ted Kremenekcfa8d572009-04-09 18:08:18 +0000591<!-- ======================================================================= -->
592
Douglas Gregor32110df2009-05-20 00:16:32 +0000593<p>Clang supports two implementations of precompiled headers. The
594 default implementation, precompiled headers (<a
595 href="PCHInternals.html">PCH</a>) uses a serialized representation
596 of Clang's internal data structures, encoded with the <a
597 href="http://llvm.org/docs/BitCodeFormat.html">LLVM bitstream
598 format</a>. Pretokenized headers (<a
599 href="PTHInternals.html">PTH</a>), on the other hand, contain a
600 serialized representation of the tokens encountered when
601 preprocessing a header (and anything that header includes).</p>
602
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000603
604<!-- ======================================================================= -->
605<h2 id="libfrontend">The Frontend Library</h2>
606<!-- ======================================================================= -->
607
608<p>The Frontend library contains functionality useful for building
609tools on top of the clang libraries, for example several methods for
610outputting diagnostics.</p>
611
612<!-- ======================================================================= -->
Chris Lattner86920d32007-07-31 05:42:17 +0000613<h2 id="liblex">The Lexer and Preprocessor Library</h2>
614<!-- ======================================================================= -->
615
616<p>The Lexer library contains several tightly-connected classes that are involved
617with the nasty process of lexing and preprocessing C source code. The main
618interface to this library for outside clients is the large <a
619href="#Preprocessor">Preprocessor</a> class.
620It contains the various pieces of state that are required to coherently read
621tokens out of a translation unit.</p>
622
623<p>The core interface to the Preprocessor object (once it is set up) is the
624Preprocessor::Lex method, which returns the next <a href="#Token">Token</a> from
625the preprocessor stream. There are two types of token providers that the
626preprocessor is capable of reading from: a buffer lexer (provided by the <a
627href="#Lexer">Lexer</a> class) and a buffered token stream (provided by the <a
Chris Lattner79281252008-03-09 02:27:26 +0000628href="#TokenLexer">TokenLexer</a> class).
Chris Lattner86920d32007-07-31 05:42:17 +0000629
630
631<!-- ======================================================================= -->
632<h3 id="Token">The Token class</h3>
633<!-- ======================================================================= -->
634
635<p>The Token class is used to represent a single lexed token. Tokens are
636intended to be used by the lexer/preprocess and parser libraries, but are not
637intended to live beyond them (for example, they should not live in the ASTs).<p>
638
639<p>Tokens most often live on the stack (or some other location that is efficient
640to access) as the parser is running, but occasionally do get buffered up. For
641example, macro definitions are stored as a series of tokens, and the C++
Chris Lattner3fcbb892008-11-23 08:32:53 +0000642front-end periodically needs to buffer tokens up for tentative parsing and
Chris Lattner86920d32007-07-31 05:42:17 +0000643various pieces of look-ahead. As such, the size of a Token matter. On a 32-bit
644system, sizeof(Token) is currently 16 bytes.</p>
645
Chris Lattner3932fe02009-01-06 06:02:08 +0000646<p>Tokens occur in two forms: "<a href="#AnnotationToken">Annotation
647Tokens</a>" and normal tokens. Normal tokens are those returned by the lexer,
648annotation tokens represent semantic information and are produced by the parser,
649replacing normal tokens in the token stream. Normal tokens contain the
650following information:</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000651
652<ul>
653<li><b>A SourceLocation</b> - This indicates the location of the start of the
654token.</li>
655
656<li><b>A length</b> - This stores the length of the token as stored in the
657SourceBuffer. For tokens that include them, this length includes trigraphs and
658escaped newlines which are ignored by later phases of the compiler. By pointing
659into the original source buffer, it is always possible to get the original
660spelling of a token completely accurately.</li>
661
662<li><b>IdentifierInfo</b> - If a token takes the form of an identifier, and if
663identifier lookup was enabled when the token was lexed (e.g. the lexer was not
664reading in 'raw' mode) this contains a pointer to the unique hash value for the
665identifier. Because the lookup happens before keyword identification, this
666field is set even for language keywords like 'for'.</li>
667
668<li><b>TokenKind</b> - This indicates the kind of token as classified by the
669lexer. This includes things like <tt>tok::starequal</tt> (for the "*="
670operator), <tt>tok::ampamp</tt> for the "&amp;&amp;" token, and keyword values
671(e.g. <tt>tok::kw_for</tt>) for identifiers that correspond to keywords. Note
672that some tokens can be spelled multiple ways. For example, C++ supports
673"operator keywords", where things like "and" are treated exactly like the
674"&amp;&amp;" operator. In these cases, the kind value is set to
675<tt>tok::ampamp</tt>, which is good for the parser, which doesn't have to
676consider both forms. For something that cares about which form is used (e.g.
677the preprocessor 'stringize' operator) the spelling indicates the original
678form.</li>
679
680<li><b>Flags</b> - There are currently four flags tracked by the
681lexer/preprocessor system on a per-token basis:
682
683 <ol>
684 <li><b>StartOfLine</b> - This was the first token that occurred on its input
685 source line.</li>
686 <li><b>LeadingSpace</b> - There was a space character either immediately
687 before the token or transitively before the token as it was expanded
688 through a macro. The definition of this flag is very closely defined by
689 the stringizing requirements of the preprocessor.</li>
690 <li><b>DisableExpand</b> - This flag is used internally to the preprocessor to
691 represent identifier tokens which have macro expansion disabled. This
692 prevents them from being considered as candidates for macro expansion ever
693 in the future.</li>
694 <li><b>NeedsCleaning</b> - This flag is set if the original spelling for the
695 token includes a trigraph or escaped newline. Since this is uncommon,
696 many pieces of code can fast-path on tokens that did not need cleaning.
Chris Lattner86920d32007-07-31 05:42:17 +0000697 </ol>
698</li>
699</ul>
700
Chris Lattner3932fe02009-01-06 06:02:08 +0000701<p>One interesting (and somewhat unusual) aspect of normal tokens is that they
702don't contain any semantic information about the lexed value. For example, if
703the token was a pp-number token, we do not represent the value of the number
704that was lexed (this is left for later pieces of code to decide). Additionally,
705the lexer library has no notion of typedef names vs variable names: both are
Chris Lattner86920d32007-07-31 05:42:17 +0000706returned as identifiers, and the parser is left to decide whether a specific
707identifier is a typedef or a variable (tracking this requires scope information
Chris Lattner3932fe02009-01-06 06:02:08 +0000708among other things). The parser can do this translation by replacing tokens
709returned by the preprocessor with "Annotation Tokens".</p>
710
711<!-- ======================================================================= -->
712<h3 id="AnnotationToken">Annotation Tokens</h3>
713<!-- ======================================================================= -->
714
715<p>Annotation Tokens are tokens that are synthesized by the parser and injected
716into the preprocessor's token stream (replacing existing tokens) to record
717semantic information found by the parser. For example, if "foo" is found to be
718a typedef, the "foo" <tt>tok::identifier</tt> token is replaced with an
719<tt>tok::annot_typename</tt>. This is useful for a couple of reasons: 1) this
720makes it easy to handle qualified type names (e.g. "foo::bar::baz&lt;42&gt;::t")
721in C++ as a single "token" in the parser. 2) if the parser backtracks, the
722reparse does not need to redo semantic analysis to determine whether a token
723sequence is a variable, type, template, etc.</p>
724
725<p>Annotation Tokens are created by the parser and reinjected into the parser's
726token stream (when backtracking is enabled). Because they can only exist in
727tokens that the preprocessor-proper is done with, it doesn't need to keep around
728flags like "start of line" that the preprocessor uses to do its job.
729Additionally, an annotation token may "cover" a sequence of preprocessor tokens
730(e.g. <tt>a::b::c</tt> is five preprocessor tokens). As such, the valid fields
731of an annotation token are different than the fields for a normal token (but
732they are multiplexed into the normal Token fields):</p>
733
734<ul>
735<li><b>SourceLocation "Location"</b> - The SourceLocation for the annotation
736token indicates the first token replaced by the annotation token. In the example
737above, it would be the location of the "a" identifier.</li>
738
739<li><b>SourceLocation "AnnotationEndLoc"</b> - This holds the location of the
740last token replaced with the annotation token. In the example above, it would
741be the location of the "c" identifier.</li>
742
John McCall027ac442010-09-03 05:07:55 +0000743<li><b>void* "AnnotationValue"</b> - This contains an opaque object
744that the parser gets from Sema. The parser merely preserves the
745information for Sema to later interpret based on the annotation token
746kind.</li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000747
748<li><b>TokenKind "Kind"</b> - This indicates the kind of Annotation token this
749is. See below for the different valid kinds.</li>
750</ul>
751
752<p>Annotation tokens currently come in three kinds:</p>
753
754<ol>
755<li><b>tok::annot_typename</b>: This annotation token represents a
John McCall027ac442010-09-03 05:07:55 +0000756resolved typename token that is potentially qualified. The
757AnnotationValue field contains the <tt>QualType</tt> returned by
758Sema::getTypeName(), possibly with source location information
759attached.</li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000760
John McCall027ac442010-09-03 05:07:55 +0000761<li><b>tok::annot_cxxscope</b>: This annotation token represents a C++
762scope specifier, such as "A::B::". This corresponds to the grammar
763productions "::" and ":: [opt] nested-name-specifier". The
764AnnotationValue pointer is a <tt>NestedNameSpecifier*</tt> returned by
765the Sema::ActOnCXXGlobalScopeSpecifier and
766Sema::ActOnCXXNestedNameSpecifier callbacks.</li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000767
Douglas Gregor39a8de12009-02-25 19:37:18 +0000768<li><b>tok::annot_template_id</b>: This annotation token represents a
769C++ template-id such as "foo&lt;int, 4&gt;", where "foo" is the name
770of a template. The AnnotationValue pointer is a pointer to a malloc'd
John McCall027ac442010-09-03 05:07:55 +0000771TemplateIdAnnotation object. Depending on the context, a parsed
772template-id that names a type might become a typename annotation token
773(if all we care about is the named type, e.g., because it occurs in a
774type specifier) or might remain a template-id token (if we want to
775retain more source location information or produce a new type, e.g.,
776in a declaration of a class template specialization). template-id
777annotation tokens that refer to a type can be "upgraded" to typename
778annotation tokens by the parser.</li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000779
780</ol>
781
Cedric Venetda76b282009-01-06 16:22:54 +0000782<p>As mentioned above, annotation tokens are not returned by the preprocessor,
Chris Lattner3932fe02009-01-06 06:02:08 +0000783they are formed on demand by the parser. This means that the parser has to be
784aware of cases where an annotation could occur and form it where appropriate.
785This is somewhat similar to how the parser handles Translation Phase 6 of C99:
786String Concatenation (see C99 5.1.1.2). In the case of string concatenation,
787the preprocessor just returns distinct tok::string_literal and
788tok::wide_string_literal tokens and the parser eats a sequence of them wherever
789the grammar indicates that a string literal can occur.</p>
790
791<p>In order to do this, whenever the parser expects a tok::identifier or
792tok::coloncolon, it should call the TryAnnotateTypeOrScopeToken or
793TryAnnotateCXXScopeToken methods to form the annotation token. These methods
794will maximally form the specified annotation tokens and replace the current
795token with them, if applicable. If the current tokens is not valid for an
796annotation token, it will remain an identifier or :: token.</p>
797
798
Chris Lattner86920d32007-07-31 05:42:17 +0000799
800<!-- ======================================================================= -->
801<h3 id="Lexer">The Lexer class</h3>
802<!-- ======================================================================= -->
803
804<p>The Lexer class provides the mechanics of lexing tokens out of a source
805buffer and deciding what they mean. The Lexer is complicated by the fact that
806it operates on raw buffers that have not had spelling eliminated (this is a
807necessity to get decent performance), but this is countered with careful coding
808as well as standard performance techniques (for example, the comment handling
809code is vectorized on X86 and PowerPC hosts).</p>
810
811<p>The lexer has a couple of interesting modal features:</p>
812
813<ul>
814<li>The lexer can operate in 'raw' mode. This mode has several features that
815 make it possible to quickly lex the file (e.g. it stops identifier lookup,
816 doesn't specially handle preprocessor tokens, handles EOF differently, etc).
817 This mode is used for lexing within an "<tt>#if 0</tt>" block, for
818 example.</li>
819<li>The lexer can capture and return comments as tokens. This is required to
820 support the -C preprocessor mode, which passes comments through, and is
821 used by the diagnostic checker to identifier expect-error annotations.</li>
822<li>The lexer can be in ParsingFilename mode, which happens when preprocessing
Chris Lattner84386242007-09-16 19:25:23 +0000823 after reading a #include directive. This mode changes the parsing of '&lt;'
Chris Lattner86920d32007-07-31 05:42:17 +0000824 to return an "angled string" instead of a bunch of tokens for each thing
825 within the filename.</li>
826<li>When parsing a preprocessor directive (after "<tt>#</tt>") the
827 ParsingPreprocessorDirective mode is entered. This changes the parser to
Peter Collingbourne84021552011-02-28 02:37:51 +0000828 return EOD at a newline.</li>
Chris Lattner86920d32007-07-31 05:42:17 +0000829<li>The Lexer uses a LangOptions object to know whether trigraphs are enabled,
830 whether C++ or ObjC keywords are recognized, etc.</li>
831</ul>
832
833<p>In addition to these modes, the lexer keeps track of a couple of other
834 features that are local to a lexed buffer, which change as the buffer is
835 lexed:</p>
836
837<ul>
838<li>The Lexer uses BufferPtr to keep track of the current character being
839 lexed.</li>
840<li>The Lexer uses IsAtStartOfLine to keep track of whether the next lexed token
841 will start with its "start of line" bit set.</li>
842<li>The Lexer keeps track of the current #if directives that are active (which
843 can be nested).</li>
844<li>The Lexer keeps track of an <a href="#MultipleIncludeOpt">
845 MultipleIncludeOpt</a> object, which is used to
846 detect whether the buffer uses the standard "<tt>#ifndef XX</tt> /
847 <tt>#define XX</tt>" idiom to prevent multiple inclusion. If a buffer does,
848 subsequent includes can be ignored if the XX macro is defined.</li>
849</ul>
850
851<!-- ======================================================================= -->
Chris Lattner79281252008-03-09 02:27:26 +0000852<h3 id="TokenLexer">The TokenLexer class</h3>
Chris Lattner86920d32007-07-31 05:42:17 +0000853<!-- ======================================================================= -->
854
Chris Lattner79281252008-03-09 02:27:26 +0000855<p>The TokenLexer class is a token provider that returns tokens from a list
Chris Lattner86920d32007-07-31 05:42:17 +0000856of tokens that came from somewhere else. It typically used for two things: 1)
857returning tokens from a macro definition as it is being expanded 2) returning
858tokens from an arbitrary buffer of tokens. The later use is used by _Pragma and
859will most likely be used to handle unbounded look-ahead for the C++ parser.</p>
860
861<!-- ======================================================================= -->
862<h3 id="MultipleIncludeOpt">The MultipleIncludeOpt class</h3>
863<!-- ======================================================================= -->
864
865<p>The MultipleIncludeOpt class implements a really simple little state machine
866that is used to detect the standard "<tt>#ifndef XX</tt> / <tt>#define XX</tt>"
867idiom that people typically use to prevent multiple inclusion of headers. If a
868buffer uses this idiom and is subsequently #include'd, the preprocessor can
869simply check to see whether the guarding condition is defined or not. If so,
870the preprocessor can completely ignore the include of the header.</p>
871
872
873
874<!-- ======================================================================= -->
875<h2 id="libparse">The Parser Library</h2>
876<!-- ======================================================================= -->
877
878<!-- ======================================================================= -->
879<h2 id="libast">The AST Library</h2>
880<!-- ======================================================================= -->
881
882<!-- ======================================================================= -->
883<h3 id="Type">The Type class and its subclasses</h3>
884<!-- ======================================================================= -->
885
886<p>The Type class (and its subclasses) are an important part of the AST. Types
887are accessed through the ASTContext class, which implicitly creates and uniques
888them as they are needed. Types have a couple of non-obvious features: 1) they
889do not capture type qualifiers like const or volatile (See
890<a href="#QualType">QualType</a>), and 2) they implicitly capture typedef
Chris Lattner8a2bc622007-07-31 06:37:39 +0000891information. Once created, types are immutable (unlike decls).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000892
893<p>Typedefs in C make semantic analysis a bit more complex than it would
894be without them. The issue is that we want to capture typedef information
895and represent it in the AST perfectly, but the semantics of operations need to
896"see through" typedefs. For example, consider this code:</p>
897
898<code>
899void func() {<br>
Bill Wendling30d17752007-10-06 01:56:01 +0000900&nbsp;&nbsp;typedef int foo;<br>
901&nbsp;&nbsp;foo X, *Y;<br>
902&nbsp;&nbsp;typedef foo* bar;<br>
903&nbsp;&nbsp;bar Z;<br>
904&nbsp;&nbsp;*X; <i>// error</i><br>
905&nbsp;&nbsp;**Y; <i>// error</i><br>
906&nbsp;&nbsp;**Z; <i>// error</i><br>
Chris Lattner86920d32007-07-31 05:42:17 +0000907}<br>
908</code>
909
910<p>The code above is illegal, and thus we expect there to be diagnostics emitted
911on the annotated lines. In this example, we expect to get:</p>
912
913<pre>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000914<b>test.c:6:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner86920d32007-07-31 05:42:17 +0000915*X; // error
Benjamin Kramer665a8dc2012-01-15 15:26:07 +0000916<span style="color:blue">^~</span>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000917<b>test.c:7:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner86920d32007-07-31 05:42:17 +0000918**Y; // error
Benjamin Kramer665a8dc2012-01-15 15:26:07 +0000919<span style="color:blue">^~~</span>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000920<b>test.c:8:1: error: indirection requires pointer operand ('foo' invalid)</b>
921**Z; // error
Benjamin Kramer665a8dc2012-01-15 15:26:07 +0000922<span style="color:blue">^~~</span>
Chris Lattner86920d32007-07-31 05:42:17 +0000923</pre>
924
925<p>While this example is somewhat silly, it illustrates the point: we want to
926retain typedef information where possible, so that we can emit errors about
927"<tt>std::string</tt>" instead of "<tt>std::basic_string&lt;char, std:...</tt>".
928Doing this requires properly keeping typedef information (for example, the type
929of "X" is "foo", not "int"), and requires properly propagating it through the
Chris Lattner8a2bc622007-07-31 06:37:39 +0000930various operators (for example, the type of *Y is "foo", not "int"). In order
931to retain this information, the type of these expressions is an instance of the
932TypedefType class, which indicates that the type of these expressions is a
933typedef for foo.
934</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000935
Chris Lattner8a2bc622007-07-31 06:37:39 +0000936<p>Representing types like this is great for diagnostics, because the
937user-specified type is always immediately available. There are two problems
938with this: first, various semantic checks need to make judgements about the
Douglas Gregor2d1e21a2011-12-19 19:50:23 +0000939<em>actual structure</em> of a type, ignoring typedefs. Second, we need an
Chris Lattner33fc68a2007-07-31 18:54:50 +0000940efficient way to query whether two types are structurally identical to each
941other, ignoring typedefs. The solution to both of these problems is the idea of
Chris Lattner8a2bc622007-07-31 06:37:39 +0000942canonical types.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000943
Chris Lattner62fd2782008-11-22 21:41:31 +0000944<!-- =============== -->
Chris Lattner8a2bc622007-07-31 06:37:39 +0000945<h4>Canonical Types</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000946<!-- =============== -->
Chris Lattner86920d32007-07-31 05:42:17 +0000947
Chris Lattner8a2bc622007-07-31 06:37:39 +0000948<p>Every instance of the Type class contains a canonical type pointer. For
949simple types with no typedefs involved (e.g. "<tt>int</tt>", "<tt>int*</tt>",
950"<tt>int**</tt>"), the type just points to itself. For types that have a
951typedef somewhere in their structure (e.g. "<tt>foo</tt>", "<tt>foo*</tt>",
952"<tt>foo**</tt>", "<tt>bar</tt>"), the canonical type pointer points to their
953structurally equivalent type without any typedefs (e.g. "<tt>int</tt>",
954"<tt>int*</tt>", "<tt>int**</tt>", and "<tt>int*</tt>" respectively).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000955
Chris Lattner8a2bc622007-07-31 06:37:39 +0000956<p>This design provides a constant time operation (dereferencing the canonical
957type pointer) that gives us access to the structure of types. For example,
958we can trivially tell that "bar" and "foo*" are the same type by dereferencing
959their canonical type pointers and doing a pointer comparison (they both point
960to the single "<tt>int*</tt>" type).</p>
961
962<p>Canonical types and typedef types bring up some complexities that must be
963carefully managed. Specifically, the "isa/cast/dyncast" operators generally
964shouldn't be used in code that is inspecting the AST. For example, when type
965checking the indirection operator (unary '*' on a pointer), the type checker
966must verify that the operand has a pointer type. It would not be correct to
967check that with "<tt>isa&lt;PointerType&gt;(SubExpr-&gt;getType())</tt>",
968because this predicate would fail if the subexpression had a typedef type.</p>
969
970<p>The solution to this problem are a set of helper methods on Type, used to
971check their properties. In this case, it would be correct to use
972"<tt>SubExpr-&gt;getType()-&gt;isPointerType()</tt>" to do the check. This
973predicate will return true if the <em>canonical type is a pointer</em>, which is
974true any time the type is structurally a pointer type. The only hard part here
975is remembering not to use the <tt>isa/cast/dyncast</tt> operations.</p>
976
977<p>The second problem we face is how to get access to the pointer type once we
978know it exists. To continue the example, the result type of the indirection
979operator is the pointee type of the subexpression. In order to determine the
980type, we need to get the instance of PointerType that best captures the typedef
981information in the program. If the type of the expression is literally a
982PointerType, we can return that, otherwise we have to dig through the
983typedefs to find the pointer type. For example, if the subexpression had type
984"<tt>foo*</tt>", we could return that type as the result. If the subexpression
985had type "<tt>bar</tt>", we want to return "<tt>foo*</tt>" (note that we do
986<em>not</em> want "<tt>int*</tt>"). In order to provide all of this, Type has
Chris Lattner11406c12007-07-31 16:50:51 +0000987a getAsPointerType() method that checks whether the type is structurally a
Chris Lattner8a2bc622007-07-31 06:37:39 +0000988PointerType and, if so, returns the best one. If not, it returns a null
989pointer.</p>
990
991<p>This structure is somewhat mystical, but after meditating on it, it will
992make sense to you :).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000993
994<!-- ======================================================================= -->
995<h3 id="QualType">The QualType class</h3>
996<!-- ======================================================================= -->
997
John McCall027ac442010-09-03 05:07:55 +0000998<p>The QualType class is designed as a trivial value class that is
999small, passed by-value and is efficient to query. The idea of
1000QualType is that it stores the type qualifiers (const, volatile,
1001restrict, plus some extended qualifiers required by language
1002extensions) separately from the types themselves. QualType is
1003conceptually a pair of "Type*" and the bits for these type qualifiers.</p>
Chris Lattner86920d32007-07-31 05:42:17 +00001004
1005<p>By storing the type qualifiers as bits in the conceptual pair, it is
1006extremely efficient to get the set of qualifiers on a QualType (just return the
1007field of the pair), add a type qualifier (which is a trivial constant-time
1008operation that sets a bit), and remove one or more type qualifiers (just return
1009a QualType with the bitfield set to empty).</p>
1010
1011<p>Further, because the bits are stored outside of the type itself, we do not
1012need to create duplicates of types with different sets of qualifiers (i.e. there
1013is only a single heap allocated "int" type: "const int" and "volatile const int"
1014both point to the same heap allocated "int" type). This reduces the heap size
1015used to represent bits and also means we do not have to consider qualifiers when
1016uniquing types (<a href="#Type">Type</a> does not even contain qualifiers).</p>
1017
John McCall027ac442010-09-03 05:07:55 +00001018<p>In practice, the two most common type qualifiers (const and
1019restrict) are stored in the low bits of the pointer to the Type
1020object, together with a flag indicating whether extended qualifiers
1021are present (which must be heap-allocated). This means that QualType
1022is exactly the same size as a pointer.</p>
Ted Kremenek8bc05712007-10-10 23:01:43 +00001023
1024<!-- ======================================================================= -->
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001025<h3 id="DeclarationName">Declaration names</h3>
1026<!-- ======================================================================= -->
1027
1028<p>The <tt>DeclarationName</tt> class represents the name of a
1029 declaration in Clang. Declarations in the C family of languages can
Chris Lattner3fcbb892008-11-23 08:32:53 +00001030 take several different forms. Most declarations are named by
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001031 simple identifiers, e.g., "<code>f</code>" and "<code>x</code>" in
1032 the function declaration <code>f(int x)</code>. In C++, declaration
1033 names can also name class constructors ("<code>Class</code>"
1034 in <code>struct Class { Class(); }</code>), class destructors
1035 ("<code>~Class</code>"), overloaded operator names ("operator+"),
1036 and conversion functions ("<code>operator void const *</code>"). In
1037 Objective-C, declaration names can refer to the names of Objective-C
1038 methods, which involve the method name and the parameters,
Chris Lattner3fcbb892008-11-23 08:32:53 +00001039 collectively called a <i>selector</i>, e.g.,
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001040 "<code>setWidth:height:</code>". Since all of these kinds of
Chris Lattner3fcbb892008-11-23 08:32:53 +00001041 entities - variables, functions, Objective-C methods, C++
1042 constructors, destructors, and operators - are represented as
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001043 subclasses of Clang's common <code>NamedDecl</code>
1044 class, <code>DeclarationName</code> is designed to efficiently
1045 represent any kind of name.</p>
1046
1047<p>Given
1048 a <code>DeclarationName</code> <code>N</code>, <code>N.getNameKind()</code>
Douglas Gregor2def4832008-11-17 20:34:05 +00001049 will produce a value that describes what kind of name <code>N</code>
Douglas Gregore94ca9e42008-11-18 14:39:36 +00001050 stores. There are 8 options (all of the names are inside
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001051 the <code>DeclarationName</code> class)</p>
1052<dl>
1053 <dt>Identifier</dt>
1054 <dd>The name is a simple
1055 identifier. Use <code>N.getAsIdentifierInfo()</code> to retrieve the
1056 corresponding <code>IdentifierInfo*</code> pointing to the actual
1057 identifier. Note that C++ overloaded operators (e.g.,
1058 "<code>operator+</code>") are represented as special kinds of
1059 identifiers. Use <code>IdentifierInfo</code>'s <code>getOverloadedOperatorID</code>
1060 function to determine whether an identifier is an overloaded
1061 operator name.</dd>
1062
1063 <dt>ObjCZeroArgSelector, ObjCOneArgSelector,
1064 ObjCMultiArgSelector</dt>
1065 <dd>The name is an Objective-C selector, which can be retrieved as a
1066 <code>Selector</code> instance
1067 via <code>N.getObjCSelector()</code>. The three possible name
1068 kinds for Objective-C reflect an optimization within
1069 the <code>DeclarationName</code> class: both zero- and
1070 one-argument selectors are stored as a
1071 masked <code>IdentifierInfo</code> pointer, and therefore require
1072 very little space, since zero- and one-argument selectors are far
1073 more common than multi-argument selectors (which use a different
1074 structure).</dd>
1075
1076 <dt>CXXConstructorName</dt>
1077 <dd>The name is a C++ constructor
1078 name. Use <code>N.getCXXNameType()</code> to retrieve
1079 the <a href="#QualType">type</a> that this constructor is meant to
1080 construct. The type is always the canonical type, since all
1081 constructors for a given type have the same name.</dd>
1082
1083 <dt>CXXDestructorName</dt>
1084 <dd>The name is a C++ destructor
1085 name. Use <code>N.getCXXNameType()</code> to retrieve
1086 the <a href="#QualType">type</a> whose destructor is being
1087 named. This type is always a canonical type.</dd>
1088
1089 <dt>CXXConversionFunctionName</dt>
1090 <dd>The name is a C++ conversion function. Conversion functions are
1091 named according to the type they convert to, e.g., "<code>operator void
1092 const *</code>". Use <code>N.getCXXNameType()</code> to retrieve
1093 the type that this conversion function converts to. This type is
1094 always a canonical type.</dd>
Douglas Gregore94ca9e42008-11-18 14:39:36 +00001095
1096 <dt>CXXOperatorName</dt>
1097 <dd>The name is a C++ overloaded operator name. Overloaded operators
1098 are named according to their spelling, e.g.,
1099 "<code>operator+</code>" or "<code>operator new
1100 []</code>". Use <code>N.getCXXOverloadedOperator()</code> to
1101 retrieve the overloaded operator (a value of
1102 type <code>OverloadedOperatorKind</code>).</dd>
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001103</dl>
1104
1105<p><code>DeclarationName</code>s are cheap to create, copy, and
1106 compare. They require only a single pointer's worth of storage in
Douglas Gregore94ca9e42008-11-18 14:39:36 +00001107 the common cases (identifiers, zero-
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001108 and one-argument Objective-C selectors) and use dense, uniqued
1109 storage for the other kinds of
1110 names. Two <code>DeclarationName</code>s can be compared for
1111 equality (<code>==</code>, <code>!=</code>) using a simple bitwise
1112 comparison, can be ordered
1113 with <code>&lt;</code>, <code>&gt;</code>, <code>&lt;=</code>,
1114 and <code>&gt;=</code> (which provide a lexicographical ordering for
1115 normal identifiers but an unspecified ordering for other kinds of
1116 names), and can be placed into LLVM <code>DenseMap</code>s
1117 and <code>DenseSet</code>s.</p>
1118
1119<p><code>DeclarationName</code> instances can be created in different
1120 ways depending on what kind of name the instance will store. Normal
Douglas Gregore94ca9e42008-11-18 14:39:36 +00001121 identifiers (<code>IdentifierInfo</code> pointers) and Objective-C selectors
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001122 (<code>Selector</code>) can be implicitly converted
1123 to <code>DeclarationName</code>s. Names for C++ constructors,
Douglas Gregore94ca9e42008-11-18 14:39:36 +00001124 destructors, conversion functions, and overloaded operators can be retrieved from
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001125 the <code>DeclarationNameTable</code>, an instance of which is
1126 available as <code>ASTContext::DeclarationNames</code>. The member
1127 functions <code>getCXXConstructorName</code>, <code>getCXXDestructorName</code>,
Douglas Gregore94ca9e42008-11-18 14:39:36 +00001128 <code>getCXXConversionFunctionName</code>, and <code>getCXXOperatorName</code>, respectively,
1129 return <code>DeclarationName</code> instances for the four kinds of
Douglas Gregor2e1cd422008-11-17 14:58:09 +00001130 C++ special function names.</p>
1131
1132<!-- ======================================================================= -->
Douglas Gregor074149e2009-01-05 19:45:36 +00001133<h3 id="DeclContext">Declaration contexts</h3>
1134<!-- ======================================================================= -->
1135<p>Every declaration in a program exists within some <i>declaration
1136 context</i>, such as a translation unit, namespace, class, or
1137 function. Declaration contexts in Clang are represented by
1138 the <code>DeclContext</code> class, from which the various
1139 declaration-context AST nodes
1140 (<code>TranslationUnitDecl</code>, <code>NamespaceDecl</code>, <code>RecordDecl</code>, <code>FunctionDecl</code>,
1141 etc.) will derive. The <code>DeclContext</code> class provides
1142 several facilities common to each declaration context:</p>
1143<dl>
1144 <dt>Source-centric vs. Semantics-centric View of Declarations</dt>
1145 <dd><code>DeclContext</code> provides two views of the declarations
1146 stored within a declaration context. The source-centric view
1147 accurately represents the program source code as written, including
1148 multiple declarations of entities where present (see the
1149 section <a href="#Redeclarations">Redeclarations and
1150 Overloads</a>), while the semantics-centric view represents the
1151 program semantics. The two views are kept synchronized by semantic
1152 analysis while the ASTs are being constructed.</dd>
1153
1154 <dt>Storage of declarations within that context</dt>
1155 <dd>Every declaration context can contain some number of
1156 declarations. For example, a C++ class (represented
1157 by <code>RecordDecl</code>) contains various member functions,
1158 fields, nested types, and so on. All of these declarations will be
1159 stored within the <code>DeclContext</code>, and one can iterate
1160 over the declarations via
1161 [<code>DeclContext::decls_begin()</code>,
1162 <code>DeclContext::decls_end()</code>). This mechanism provides
1163 the source-centric view of declarations in the context.</dd>
1164
1165 <dt>Lookup of declarations within that context</dt>
1166 <dd>The <code>DeclContext</code> structure provides efficient name
1167 lookup for names within that declaration context. For example,
1168 if <code>N</code> is a namespace we can look for the
1169 name <code>N::f</code>
1170 using <code>DeclContext::lookup</code>. The lookup itself is
1171 based on a lazily-constructed array (for declaration contexts
1172 with a small number of declarations) or hash table (for
1173 declaration contexts with more declarations). The lookup
1174 operation provides the semantics-centric view of the declarations
1175 in the context.</dd>
1176
1177 <dt>Ownership of declarations</dt>
1178 <dd>The <code>DeclContext</code> owns all of the declarations that
1179 were declared within its declaration context, and is responsible
1180 for the management of their memory as well as their
1181 (de-)serialization.</dd>
1182</dl>
1183
Douglas Gregor4afa39d2009-01-20 01:17:11 +00001184<p>All declarations are stored within a declaration context, and one
1185 can query
1186 information about the context in which each declaration lives. One
Douglas Gregor074149e2009-01-05 19:45:36 +00001187 can retrieve the <code>DeclContext</code> that contains a
Douglas Gregor4afa39d2009-01-20 01:17:11 +00001188 particular <code>Decl</code>
1189 using <code>Decl::getDeclContext</code>. However, see the
Douglas Gregor074149e2009-01-05 19:45:36 +00001190 section <a href="#LexicalAndSemanticContexts">Lexical and Semantic
1191 Contexts</a> for more information about how to interpret this
1192 context information.</p>
1193
1194<h4 id="Redeclarations">Redeclarations and Overloads</h4>
1195<p>Within a translation unit, it is common for an entity to be
1196declared several times. For example, we might declare a function "f"
1197 and then later re-declare it as part of an inlined definition:</p>
1198
1199<pre>
1200void f(int x, int y, int z = 1);
1201
1202inline void f(int x, int y, int z) { /* ... */ }
1203</pre>
1204
1205<p>The representation of "f" differs in the source-centric and
1206 semantics-centric views of a declaration context. In the
1207 source-centric view, all redeclarations will be present, in the
1208 order they occurred in the source code, making
1209 this view suitable for clients that wish to see the structure of
1210 the source code. In the semantics-centric view, only the most recent "f"
1211 will be found by the lookup, since it effectively replaces the first
1212 declaration of "f".</p>
1213
1214<p>In the semantics-centric view, overloading of functions is
1215 represented explicitly. For example, given two declarations of a
1216 function "g" that are overloaded, e.g.,</p>
1217<pre>
1218void g();
1219void g(int);
1220</pre>
1221<p>the <code>DeclContext::lookup</code> operation will return
Jonathan D. Turnerd3224292011-07-06 18:12:36 +00001222 a <code>DeclContext::lookup_result</code> that contains a range of iterators
1223 over declarations of "g". Clients that perform semantic analysis on a
Douglas Gregor074149e2009-01-05 19:45:36 +00001224 program that is not concerned with the actual source code will
1225 primarily use this semantics-centric view.</p>
1226
1227<h4 id="LexicalAndSemanticContexts">Lexical and Semantic Contexts</h4>
Douglas Gregor4afa39d2009-01-20 01:17:11 +00001228<p>Each declaration has two potentially different
Douglas Gregor074149e2009-01-05 19:45:36 +00001229 declaration contexts: a <i>lexical</i> context, which corresponds to
1230 the source-centric view of the declaration context, and
1231 a <i>semantic</i> context, which corresponds to the
1232 semantics-centric view. The lexical context is accessible
Douglas Gregor4afa39d2009-01-20 01:17:11 +00001233 via <code>Decl::getLexicalDeclContext</code> while the
Douglas Gregor074149e2009-01-05 19:45:36 +00001234 semantic context is accessible
Douglas Gregor4afa39d2009-01-20 01:17:11 +00001235 via <code>Decl::getDeclContext</code>, both of which return
Douglas Gregor074149e2009-01-05 19:45:36 +00001236 <code>DeclContext</code> pointers. For most declarations, the two
1237 contexts are identical. For example:</p>
1238
1239<pre>
1240class X {
1241public:
1242 void f(int x);
1243};
1244</pre>
1245
1246<p>Here, the semantic and lexical contexts of <code>X::f</code> are
1247 the <code>DeclContext</code> associated with the
1248 class <code>X</code> (itself stored as a <code>RecordDecl</code> AST
1249 node). However, we can now define <code>X::f</code> out-of-line:</p>
1250
1251<pre>
1252void X::f(int x = 17) { /* ... */ }
1253</pre>
1254
1255<p>This definition of has different lexical and semantic
1256 contexts. The lexical context corresponds to the declaration
1257 context in which the actual declaration occurred in the source
1258 code, e.g., the translation unit containing <code>X</code>. Thus,
1259 this declaration of <code>X::f</code> can be found by traversing
1260 the declarations provided by
1261 [<code>decls_begin()</code>, <code>decls_end()</code>) in the
1262 translation unit.</p>
1263
1264<p>The semantic context of <code>X::f</code> corresponds to the
1265 class <code>X</code>, since this member function is (semantically) a
1266 member of <code>X</code>. Lookup of the name <code>f</code> into
1267 the <code>DeclContext</code> associated with <code>X</code> will
1268 then return the definition of <code>X::f</code> (including
1269 information about the default argument).</p>
1270
1271<h4 id="TransparentContexts">Transparent Declaration Contexts</h4>
1272<p>In C and C++, there are several contexts in which names that are
1273 logically declared inside another declaration will actually "leak"
1274 out into the enclosing scope from the perspective of name
1275 lookup. The most obvious instance of this behavior is in
1276 enumeration types, e.g.,</p>
1277<pre>
1278enum Color {
1279 Red,
1280 Green,
1281 Blue
1282};
1283</pre>
1284
1285<p>Here, <code>Color</code> is an enumeration, which is a declaration
1286 context that contains the
1287 enumerators <code>Red</code>, <code>Green</code>,
1288 and <code>Blue</code>. Thus, traversing the list of declarations
1289 contained in the enumeration <code>Color</code> will
1290 yield <code>Red</code>, <code>Green</code>,
1291 and <code>Blue</code>. However, outside of the scope
1292 of <code>Color</code> one can name the enumerator <code>Red</code>
1293 without qualifying the name, e.g.,</p>
1294
1295<pre>
1296Color c = Red;
1297</pre>
1298
1299<p>There are other entities in C++ that provide similar behavior. For
1300 example, linkage specifications that use curly braces:</p>
1301
1302<pre>
1303extern "C" {
1304 void f(int);
1305 void g(int);
1306}
1307// f and g are visible here
1308</pre>
1309
1310<p>For source-level accuracy, we treat the linkage specification and
1311 enumeration type as a
1312 declaration context in which its enclosed declarations ("Red",
1313 "Green", and "Blue"; "f" and "g")
1314 are declared. However, these declarations are visible outside of the
1315 scope of the declaration context.</p>
1316
1317<p>These language features (and several others, described below) have
1318 roughly the same set of
1319 requirements: declarations are declared within a particular lexical
1320 context, but the declarations are also found via name lookup in
1321 scopes enclosing the declaration itself. This feature is implemented
1322 via <i>transparent</i> declaration contexts
1323 (see <code>DeclContext::isTransparentContext()</code>), whose
1324 declarations are visible in the nearest enclosing non-transparent
1325 declaration context. This means that the lexical context of the
1326 declaration (e.g., an enumerator) will be the
1327 transparent <code>DeclContext</code> itself, as will the semantic
1328 context, but the declaration will be visible in every outer context
1329 up to and including the first non-transparent declaration context (since
1330 transparent declaration contexts can be nested).</p>
1331
1332<p>The transparent <code>DeclContexts</code> are:</p>
1333<ul>
David Blaikie5090e9f2011-10-18 05:49:30 +00001334 <li>Enumerations (but not C++11 "scoped enumerations"):
Douglas Gregor074149e2009-01-05 19:45:36 +00001335 <pre>
1336enum Color {
1337 Red,
1338 Green,
1339 Blue
1340};
1341// Red, Green, and Blue are in scope
1342 </pre></li>
1343 <li>C++ linkage specifications:
1344 <pre>
1345extern "C" {
1346 void f(int);
1347 void g(int);
1348}
1349// f and g are in scope
1350 </pre></li>
1351 <li>Anonymous unions and structs:
1352 <pre>
1353struct LookupTable {
1354 bool IsVector;
1355 union {
1356 std::vector&lt;Item&gt; *Vector;
1357 std::set&lt;Item&gt; *Set;
1358 };
1359};
1360
1361LookupTable LT;
1362LT.Vector = 0; // Okay: finds Vector inside the unnamed union
1363 </pre>
1364 </li>
David Blaikie5090e9f2011-10-18 05:49:30 +00001365 <li>C++11 inline namespaces:
Douglas Gregor074149e2009-01-05 19:45:36 +00001366<pre>
1367namespace mylib {
1368 inline namespace debug {
1369 class X;
1370 }
1371}
1372mylib::X *xp; // okay: mylib::X refers to mylib::debug::X
1373</pre>
1374</li>
1375</ul>
1376
1377
1378<h4 id="MultiDeclContext">Multiply-Defined Declaration Contexts</h4>
1379<p>C++ namespaces have the interesting--and, so far, unique--property that
1380the namespace can be defined multiple times, and the declarations
1381provided by each namespace definition are effectively merged (from
1382the semantic point of view). For example, the following two code
1383snippets are semantically indistinguishable:</p>
1384<pre>
1385// Snippet #1:
1386namespace N {
1387 void f();
1388}
1389namespace N {
1390 void f(int);
1391}
1392
1393// Snippet #2:
1394namespace N {
1395 void f();
1396 void f(int);
1397}
1398</pre>
1399
1400<p>In Clang's representation, the source-centric view of declaration
1401 contexts will actually have two separate <code>NamespaceDecl</code>
1402 nodes in Snippet #1, each of which is a declaration context that
1403 contains a single declaration of "f". However, the semantics-centric
1404 view provided by name lookup into the namespace <code>N</code> for
Jonathan D. Turnerd3224292011-07-06 18:12:36 +00001405 "f" will return a <code>DeclContext::lookup_result</code> that contains
1406 a range of iterators over declarations of "f".</p>
Douglas Gregor074149e2009-01-05 19:45:36 +00001407
1408<p><code>DeclContext</code> manages multiply-defined declaration
1409 contexts internally. The
1410 function <code>DeclContext::getPrimaryContext</code> retrieves the
1411 "primary" context for a given <code>DeclContext</code> instance,
1412 which is the <code>DeclContext</code> responsible for maintaining
1413 the lookup table used for the semantics-centric view. Given the
1414 primary context, one can follow the chain
1415 of <code>DeclContext</code> nodes that define additional
1416 declarations via <code>DeclContext::getNextContext</code>. Note that
1417 these functions are used internally within the lookup and insertion
1418 methods of the <code>DeclContext</code>, so the vast majority of
1419 clients can ignore them.</p>
1420
1421<!-- ======================================================================= -->
Ted Kremenek8bc05712007-10-10 23:01:43 +00001422<h3 id="CFG">The <tt>CFG</tt> class</h3>
1423<!-- ======================================================================= -->
1424
1425<p>The <tt>CFG</tt> class is designed to represent a source-level
1426control-flow graph for a single statement (<tt>Stmt*</tt>). Typically
1427instances of <tt>CFG</tt> are constructed for function bodies (usually
1428an instance of <tt>CompoundStmt</tt>), but can also be instantiated to
1429represent the control-flow of any class that subclasses <tt>Stmt</tt>,
1430which includes simple expressions. Control-flow graphs are especially
1431useful for performing
1432<a href="http://en.wikipedia.org/wiki/Data_flow_analysis#Sensitivities">flow-
1433or path-sensitive</a> program analyses on a given function.</p>
1434
Chris Lattner62fd2782008-11-22 21:41:31 +00001435<!-- ============ -->
Ted Kremenek8bc05712007-10-10 23:01:43 +00001436<h4>Basic Blocks</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +00001437<!-- ============ -->
Ted Kremenek8bc05712007-10-10 23:01:43 +00001438
1439<p>Concretely, an instance of <tt>CFG</tt> is a collection of basic
1440blocks. Each basic block is an instance of <tt>CFGBlock</tt>, which
1441simply contains an ordered sequence of <tt>Stmt*</tt> (each referring
1442to statements in the AST). The ordering of statements within a block
1443indicates unconditional flow of control from one statement to the
1444next. <a href="#ConditionalControlFlow">Conditional control-flow</a>
1445is represented using edges between basic blocks. The statements
1446within a given <tt>CFGBlock</tt> can be traversed using
1447the <tt>CFGBlock::*iterator</tt> interface.</p>
1448
1449<p>
Ted Kremenek18e17e72007-10-18 22:50:52 +00001450A <tt>CFG</tt> object owns the instances of <tt>CFGBlock</tt> within
Ted Kremenek8bc05712007-10-10 23:01:43 +00001451the control-flow graph it represents. Each <tt>CFGBlock</tt> within a
1452CFG is also uniquely numbered (accessible
1453via <tt>CFGBlock::getBlockID()</tt>). Currently the number is
1454based on the ordering the blocks were created, but no assumptions
1455should be made on how <tt>CFGBlock</tt>s are numbered other than their
1456numbers are unique and that they are numbered from 0..N-1 (where N is
1457the number of basic blocks in the CFG).</p>
1458
Chris Lattner62fd2782008-11-22 21:41:31 +00001459<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +00001460<h4>Entry and Exit Blocks</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +00001461<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +00001462
1463Each instance of <tt>CFG</tt> contains two special blocks:
1464an <i>entry</i> block (accessible via <tt>CFG::getEntry()</tt>), which
1465has no incoming edges, and an <i>exit</i> block (accessible
1466via <tt>CFG::getExit()</tt>), which has no outgoing edges. Neither
1467block contains any statements, and they serve the role of providing a
1468clear entrance and exit for a body of code such as a function body.
1469The presence of these empty blocks greatly simplifies the
1470implementation of many analyses built on top of CFGs.
1471
Chris Lattner62fd2782008-11-22 21:41:31 +00001472<!-- ===================================================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +00001473<h4 id ="ConditionalControlFlow">Conditional Control-Flow</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +00001474<!-- ===================================================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +00001475
1476<p>Conditional control-flow (such as those induced by if-statements
1477and loops) is represented as edges between <tt>CFGBlock</tt>s.
1478Because different C language constructs can induce control-flow,
1479each <tt>CFGBlock</tt> also records an extra <tt>Stmt*</tt> that
1480represents the <i>terminator</i> of the block. A terminator is simply
1481the statement that caused the control-flow, and is used to identify
1482the nature of the conditional control-flow between blocks. For
1483example, in the case of an if-statement, the terminator refers to
1484the <tt>IfStmt</tt> object in the AST that represented the given
1485branch.</p>
1486
1487<p>To illustrate, consider the following code example:</p>
1488
1489<code>
1490int foo(int x) {<br>
1491&nbsp;&nbsp;x = x + 1;<br>
1492<br>
1493&nbsp;&nbsp;if (x > 2) x++;<br>
1494&nbsp;&nbsp;else {<br>
1495&nbsp;&nbsp;&nbsp;&nbsp;x += 2;<br>
1496&nbsp;&nbsp;&nbsp;&nbsp;x *= 2;<br>
1497&nbsp;&nbsp;}<br>
1498<br>
1499&nbsp;&nbsp;return x;<br>
1500}
1501</code>
1502
1503<p>After invoking the parser+semantic analyzer on this code fragment,
1504the AST of the body of <tt>foo</tt> is referenced by a
1505single <tt>Stmt*</tt>. We can then construct an instance
1506of <tt>CFG</tt> representing the control-flow graph of this function
1507body by single call to a static class method:</p>
1508
1509<code>
1510&nbsp;&nbsp;Stmt* FooBody = ...<br>
1511&nbsp;&nbsp;CFG* FooCFG = <b>CFG::buildCFG</b>(FooBody);
1512</code>
1513
1514<p>It is the responsibility of the caller of <tt>CFG::buildCFG</tt>
1515to <tt>delete</tt> the returned <tt>CFG*</tt> when the CFG is no
1516longer needed.</p>
1517
1518<p>Along with providing an interface to iterate over
1519its <tt>CFGBlock</tt>s, the <tt>CFG</tt> class also provides methods
1520that are useful for debugging and visualizing CFGs. For example, the
1521method
1522<tt>CFG::dump()</tt> dumps a pretty-printed version of the CFG to
1523standard error. This is especially useful when one is using a
1524debugger such as gdb. For example, here is the output
1525of <tt>FooCFG->dump()</tt>:</p>
1526
1527<code>
1528&nbsp;[ B5 (ENTRY) ]<br>
1529&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (0):<br>
1530&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B4<br>
1531<br>
1532&nbsp;[ B4 ]<br>
1533&nbsp;&nbsp;&nbsp;&nbsp;1: x = x + 1<br>
1534&nbsp;&nbsp;&nbsp;&nbsp;2: (x > 2)<br>
1535&nbsp;&nbsp;&nbsp;&nbsp;<b>T: if [B4.2]</b><br>
1536&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B5<br>
1537&nbsp;&nbsp;&nbsp;&nbsp;Successors (2): B3 B2<br>
1538<br>
1539&nbsp;[ B3 ]<br>
1540&nbsp;&nbsp;&nbsp;&nbsp;1: x++<br>
1541&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B4<br>
1542&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B1<br>
1543<br>
1544&nbsp;[ B2 ]<br>
1545&nbsp;&nbsp;&nbsp;&nbsp;1: x += 2<br>
1546&nbsp;&nbsp;&nbsp;&nbsp;2: x *= 2<br>
1547&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B4<br>
1548&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B1<br>
1549<br>
1550&nbsp;[ B1 ]<br>
1551&nbsp;&nbsp;&nbsp;&nbsp;1: return x;<br>
1552&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (2): B2 B3<br>
1553&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B0<br>
1554<br>
1555&nbsp;[ B0 (EXIT) ]<br>
1556&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B1<br>
1557&nbsp;&nbsp;&nbsp;&nbsp;Successors (0):
1558</code>
1559
1560<p>For each block, the pretty-printed output displays for each block
1561the number of <i>predecessor</i> blocks (blocks that have outgoing
1562control-flow to the given block) and <i>successor</i> blocks (blocks
1563that have control-flow that have incoming control-flow from the given
1564block). We can also clearly see the special entry and exit blocks at
1565the beginning and end of the pretty-printed output. For the entry
1566block (block B5), the number of predecessor blocks is 0, while for the
1567exit block (block B0) the number of successor blocks is 0.</p>
1568
1569<p>The most interesting block here is B4, whose outgoing control-flow
1570represents the branching caused by the sole if-statement
1571in <tt>foo</tt>. Of particular interest is the second statement in
1572the block, <b><tt>(x > 2)</tt></b>, and the terminator, printed
1573as <b><tt>if [B4.2]</tt></b>. The second statement represents the
1574evaluation of the condition of the if-statement, which occurs before
1575the actual branching of control-flow. Within the <tt>CFGBlock</tt>
1576for B4, the <tt>Stmt*</tt> for the second statement refers to the
1577actual expression in the AST for <b><tt>(x > 2)</tt></b>. Thus
1578pointers to subclasses of <tt>Expr</tt> can appear in the list of
1579statements in a block, and not just subclasses of <tt>Stmt</tt> that
1580refer to proper C statements.</p>
1581
1582<p>The terminator of block B4 is a pointer to the <tt>IfStmt</tt>
1583object in the AST. The pretty-printer outputs <b><tt>if
1584[B4.2]</tt></b> because the condition expression of the if-statement
1585has an actual place in the basic block, and thus the terminator is
1586essentially
1587<i>referring</i> to the expression that is the second statement of
1588block B4 (i.e., B4.2). In this manner, conditions for control-flow
1589(which also includes conditions for loops and switch statements) are
1590hoisted into the actual basic block.</p>
1591
Chris Lattner62fd2782008-11-22 21:41:31 +00001592<!-- ===================== -->
1593<!-- <h4>Implicit Control-Flow</h4> -->
1594<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +00001595
1596<!--
1597<p>A key design principle of the <tt>CFG</tt> class was to not require
1598any transformations to the AST in order to represent control-flow.
1599Thus the <tt>CFG</tt> does not perform any "lowering" of the
1600statements in an AST: loops are not transformed into guarded gotos,
1601short-circuit operations are not converted to a set of if-statements,
1602and so on.</p>
1603-->
Ted Kremenek17a295d2008-06-11 06:19:49 +00001604
Chris Lattner7bad1992008-11-16 21:48:07 +00001605
1606<!-- ======================================================================= -->
1607<h3 id="Constants">Constant Folding in the Clang AST</h3>
1608<!-- ======================================================================= -->
1609
1610<p>There are several places where constants and constant folding matter a lot to
1611the Clang front-end. First, in general, we prefer the AST to retain the source
1612code as close to how the user wrote it as possible. This means that if they
1613wrote "5+4", we want to keep the addition and two constants in the AST, we don't
1614want to fold to "9". This means that constant folding in various ways turns
1615into a tree walk that needs to handle the various cases.</p>
1616
1617<p>However, there are places in both C and C++ that require constants to be
1618folded. For example, the C standard defines what an "integer constant
1619expression" (i-c-e) is with very precise and specific requirements. The
1620language then requires i-c-e's in a lot of places (for example, the size of a
1621bitfield, the value for a case statement, etc). For these, we have to be able
1622to constant fold the constants, to do semantic checks (e.g. verify bitfield size
1623is non-negative and that case statements aren't duplicated). We aim for Clang
1624to be very pedantic about this, diagnosing cases when the code does not use an
1625i-c-e where one is required, but accepting the code unless running with
1626<tt>-pedantic-errors</tt>.</p>
1627
1628<p>Things get a little bit more tricky when it comes to compatibility with
1629real-world source code. Specifically, GCC has historically accepted a huge
1630superset of expressions as i-c-e's, and a lot of real world code depends on this
1631unfortuate accident of history (including, e.g., the glibc system headers). GCC
1632accepts anything its "fold" optimizer is capable of reducing to an integer
1633constant, which means that the definition of what it accepts changes as its
1634optimizer does. One example is that GCC accepts things like "case X-X:" even
1635when X is a variable, because it can fold this to 0.</p>
1636
1637<p>Another issue are how constants interact with the extensions we support, such
1638as __builtin_constant_p, __builtin_inf, __extension__ and many others. C99
1639obviously does not specify the semantics of any of these extensions, and the
1640definition of i-c-e does not include them. However, these extensions are often
1641used in real code, and we have to have a way to reason about them.</p>
1642
1643<p>Finally, this is not just a problem for semantic analysis. The code
1644generator and other clients have to be able to fold constants (e.g. to
1645initialize global variables) and has to handle a superset of what C99 allows.
1646Further, these clients can benefit from extended information. For example, we
1647know that "foo()||1" always evaluates to true, but we can't replace the
1648expression with true because it has side effects.</p>
1649
1650<!-- ======================= -->
1651<h4>Implementation Approach</h4>
1652<!-- ======================= -->
1653
1654<p>After trying several different approaches, we've finally converged on a
1655design (Note, at the time of this writing, not all of this has been implemented,
1656consider this a design goal!). Our basic approach is to define a single
1657recursive method evaluation method (<tt>Expr::Evaluate</tt>), which is
1658implemented in <tt>AST/ExprConstant.cpp</tt>. Given an expression with 'scalar'
1659type (integer, fp, complex, or pointer) this method returns the following
1660information:</p>
1661
1662<ul>
1663<li>Whether the expression is an integer constant expression, a general
1664 constant that was folded but has no side effects, a general constant that
1665 was folded but that does have side effects, or an uncomputable/unfoldable
1666 value.
1667</li>
1668<li>If the expression was computable in any way, this method returns the APValue
1669 for the result of the expression.</li>
1670<li>If the expression is not evaluatable at all, this method returns
1671 information on one of the problems with the expression. This includes a
1672 SourceLocation for where the problem is, and a diagnostic ID that explains
1673 the problem. The diagnostic should be have ERROR type.</li>
1674<li>If the expression is not an integer constant expression, this method returns
1675 information on one of the problems with the expression. This includes a
1676 SourceLocation for where the problem is, and a diagnostic ID that explains
1677 the problem. The diagnostic should be have EXTENSION type.</li>
1678</ul>
1679
1680<p>This information gives various clients the flexibility that they want, and we
1681will eventually have some helper methods for various extensions. For example,
1682Sema should have a <tt>Sema::VerifyIntegerConstantExpression</tt> method, which
1683calls Evaluate on the expression. If the expression is not foldable, the error
1684is emitted, and it would return true. If the expression is not an i-c-e, the
1685EXTENSION diagnostic is emitted. Finally it would return false to indicate that
1686the AST is ok.</p>
1687
1688<p>Other clients can use the information in other ways, for example, codegen can
1689just use expressions that are foldable in any way.</p>
1690
1691<!-- ========== -->
1692<h4>Extensions</h4>
1693<!-- ========== -->
1694
Chris Lattner552de0a2008-11-23 08:16:56 +00001695<p>This section describes how some of the various extensions Clang supports
Chris Lattner7bad1992008-11-16 21:48:07 +00001696interacts with constant evaluation:</p>
1697
1698<ul>
1699<li><b><tt>__extension__</tt></b>: The expression form of this extension causes
1700 any evaluatable subexpression to be accepted as an integer constant
1701 expression.</li>
1702<li><b><tt>__builtin_constant_p</tt></b>: This returns true (as a integer
Richard Smith8a0f1552011-12-09 03:40:28 +00001703 constant expression) if the operand evaluates to either a numeric value
1704 (that is, not a pointer cast to integral type) of integral, enumeration,
1705 floating or complex type, or if it evaluates to the address of the first
1706 character of a string literal (possibly cast to some other type). As a
Chris Lattner28daa532008-12-12 06:55:44 +00001707 special case, if <tt>__builtin_constant_p</tt> is the (potentially
1708 parenthesized) condition of a conditional operator expression ("?:"), only
Chris Lattner42b83dd2008-12-12 18:00:51 +00001709 the true side of the conditional operator is considered, and it is evaluated
1710 with full constant folding.</li>
Chris Lattner7bad1992008-11-16 21:48:07 +00001711<li><b><tt>__builtin_choose_expr</tt></b>: The condition is required to be an
1712 integer constant expression, but we accept any constant as an "extension of
1713 an extension". This only evaluates one operand depending on which way the
1714 condition evaluates.</li>
1715<li><b><tt>__builtin_classify_type</tt></b>: This always returns an integer
1716 constant expression.</li>
1717<li><b><tt>__builtin_inf,nan,..</tt></b>: These are treated just like a
1718 floating-point literal.</li>
1719<li><b><tt>__builtin_abs,copysign,..</tt></b>: These are constant folded as
1720 general constant expressions.</li>
Richard Smith8a0f1552011-12-09 03:40:28 +00001721<li><b><tt>__builtin_strlen</tt></b> and <b><tt>strlen</tt></b>: These are
1722 constant folded as integer constant expressions if the argument is a string
1723 literal.</li>
Chris Lattner7bad1992008-11-16 21:48:07 +00001724</ul>
1725
1726
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +00001727<!-- ======================================================================= -->
1728<h2 id="Howtos">How to change Clang</h2>
1729<!-- ======================================================================= -->
Chris Lattner7bad1992008-11-16 21:48:07 +00001730
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +00001731<!-- ======================================================================= -->
1732<h3 id="AddingAttributes">How to add an attribute</h3>
1733<!-- ======================================================================= -->
1734
1735<p>To add an attribute, you'll have to add it to the list of attributes, add it
1736to the parsing phase, and look for it in the AST scan.
Benjamin Kramer665a8dc2012-01-15 15:26:07 +00001737<a href="http://llvm.org/viewvc/llvm-project?view=rev&amp;revision=124217">r124217</a>
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +00001738has a good example of adding a warning attribute.</p>
1739
1740<p>(Beware that this hasn't been reviewed/fixed by the people who designed the
1741attributes system yet.)</p>
1742
1743<h4><a
1744href="http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/Attr.td?view=markup">include/clang/Basic/Attr.td</a></h4>
1745
1746<p>Each attribute gets a <tt>def</tt> inheriting from <tt>Attr</tt> or one of
1747its subclasses. <tt>InheritableAttr</tt> means that the attribute also applies
1748to subsequent declarations of the same name.</p>
1749
1750<p><tt>Spellings</tt> lists the strings that can appear in
1751<tt>__attribute__((here))</tt> or <tt>[[here]]</tt>. All such strings
David Blaikie5090e9f2011-10-18 05:49:30 +00001752will be synonymous. If you want to allow the <tt>[[]]</tt> C++11
Jeffrey Yasskin28dadd62011-01-28 23:41:54 +00001753syntax, you have to define a list of <tt>Namespaces</tt>, which will
1754let users write <tt>[[namespace:spelling]]</tt>. Using the empty
1755string for a namespace will allow users to write just the spelling
1756with no "<tt>:</tt>".</p>
1757
1758<p><tt>Subjects</tt> restricts what kinds of AST node to which this attribute
1759can appertain (roughly, attach).</p>
1760
1761<p><tt>Args</tt> names the arguments the attribute takes, in order. If
1762<tt>Args</tt> is <tt>[StringArgument&lt;"Arg1">, IntArgument&lt;"Arg2">]</tt>
1763then <tt>__attribute__((myattribute("Hello", 3)))</tt> will be a valid use.</p>
1764
1765<h4>Boilerplate</h4>
1766
1767<p>Add an element to the <tt>AttributeList::Kind</tt> enum in <a
1768href="http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Sema/AttributeList.h?view=markup">include/clang/Sema/AttributeList.h</a>
1769named <tt>AT_lower_with_underscores</tt>. That is, a CamelCased
1770<tt>AttributeName</tt> in <tt>Attr.td</tt> name should become
1771<tt>AT_attribute_name</tt>.</p>
1772
1773<p>Add a case to the <tt>StringSwitch</tt> in <tt>AttributeList::getKind()</tt>
1774in <a
1775href="http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/AttributeList.cpp?view=markup">lib/Sema/AttributeList.cpp</a>
1776for each spelling of your attribute. Less common attributes should come toward
1777the end of that list.</p>
1778
1779<p>Write a new <tt>HandleYourAttr()</tt> function in <a
1780href="http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaDeclAttr.cpp?view=markup">lib/Sema/SemaDeclAttr.cpp</a>,
1781and add a case to the switch in <tt>ProcessNonInheritableDeclAttr()</tt> or
1782<tt>ProcessInheritableDeclAttr()</tt> forwarding to it.</p>
1783
1784<p>If your attribute causes extra warnings to fire, define a <tt>DiagGroup</tt>
1785in <a
1786href="http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticGroups.td?view=markup">include/clang/Basic/DiagnosticGroups.td</a>
1787named after the attribute's <tt>Spelling</tt> with "_"s replaced by "-"s. If
1788you're only defining one diagnostic, you can skip <tt>DiagnosticGroups.td</tt>
1789and use <tt>InGroup&lt;DiagGroup&lt;"your-attribute">></tt> directly in <a
1790href="http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticSemaKinds.td?view=markup">DiagnosticSemaKinds.td</a></p>
1791
1792<h4>The meat of your attribute</h4>
1793
1794<p>Find an appropriate place in Clang to do whatever your attribute needs to do.
1795Check for the attribute's presence using <tt>Decl::getAttr&lt;YourAttr>()</tt>.</p>
1796
1797<p>Update the <a href="LanguageExtensions.html">Clang Language Extensions</a>
1798document to describe your new attribute.</p>
Chris Lattner7bad1992008-11-16 21:48:07 +00001799
Douglas Gregor1f634c62011-09-30 21:32:37 +00001800<!-- ======================================================================= -->
1801<h3 id="AddingExprStmt">How to add an expression or statement</h3>
1802<!-- ======================================================================= -->
1803
1804<p>Expressions and statements are one of the most fundamental constructs within a
1805compiler, because they interact with many different parts of the AST,
1806semantic analysis, and IR generation. Therefore, adding a new
1807expression or statement kind into Clang requires some care. The following list
1808details the various places in Clang where an expression or statement needs to be
1809introduced, along with patterns to follow to ensure that the new
1810expression or statement works well across all of the C languages. We
1811focus on expressions, but statements are similar.</p>
1812
1813<ol>
1814 <li>Introduce parsing actions into the parser. Recursive-descent
1815 parsing is mostly self-explanatory, but there are a few things that
1816 are worth keeping in mind:
1817 <ul>
1818 <li>Keep as much source location information as possible! You'll
1819 want it later to produce great diagnostics and support Clang's
1820 various features that map between source code and the AST.</li>
1821 <li>Write tests for all of the "bad" parsing cases, to make sure
1822 your recovery is good. If you have matched delimiters (e.g.,
1823 parentheses, square brackets, etc.), use
Douglas Gregor4a8dfb52011-10-12 16:37:45 +00001824 <tt>Parser::BalancedDelimiterTracker</tt> to give nice diagnostics when
Douglas Gregor1f634c62011-09-30 21:32:37 +00001825 things go wrong.</li>
1826 </ul>
1827 </li>
1828
1829 <li>Introduce semantic analysis actions into <tt>Sema</tt>. Semantic
1830 analysis should always involve two functions: an <tt>ActOnXXX</tt>
1831 function that will be called directly from the parser, and a
1832 <tt>BuildXXX</tt> function that performs the actual semantic
1833 analysis and will (eventually!) build the AST node. It's fairly
1834 common for the <tt>ActOnCXX</tt> function to do very little (often
1835 just some minor translation from the parser's representation to
1836 <tt>Sema</tt>'s representation of the same thing), but the separation
1837 is still important: C++ template instantiation, for example,
1838 should always call the <tt>BuildXXX</tt> variant. Several notes on
1839 semantic analysis before we get into construction of the AST:
1840 <ul>
1841 <li>Your expression probably involves some types and some
1842 subexpressions. Make sure to fully check that those types, and the
1843 types of those subexpressions, meet your expectations. Add
1844 implicit conversions where necessary to make sure that all of the
1845 types line up exactly the way you want them. Write extensive tests
1846 to check that you're getting good diagnostics for mistakes and
1847 that you can use various forms of subexpressions with your
1848 expression.</li>
1849 <li>When type-checking a type or subexpression, make sure to first
1850 check whether the type is "dependent"
1851 (<tt>Type::isDependentType()</tt>) or whether a subexpression is
1852 type-dependent (<tt>Expr::isTypeDependent()</tt>). If any of these
1853 return true, then you're inside a template and you can't do much
1854 type-checking now. That's normal, and your AST node (when you get
1855 there) will have to deal with this case. At this point, you can
1856 write tests that use your expression within templates, but don't
1857 try to instantiate the templates.</li>
1858 <li>For each subexpression, be sure to call
1859 <tt>Sema::CheckPlaceholderExpr()</tt> to deal with "weird"
1860 expressions that don't behave well as subexpressions. Then,
1861 determine whether you need to perform
1862 lvalue-to-rvalue conversions
1863 (<tt>Sema::DefaultLvalueConversion</tt>e) or
1864 the usual unary conversions
1865 (<tt>Sema::UsualUnaryConversions</tt>), for places where the
1866 subexpression is producing a value you intend to use.</li>
1867 <li>Your <tt>BuildXXX</tt> function will probably just return
1868 <tt>ExprError()</tt> at this point, since you don't have an AST.
1869 That's perfectly fine, and shouldn't impact your testing.</li>
1870 </ul>
1871 </li>
1872
1873 <li>Introduce an AST node for your new expression. This starts with
1874 declaring the node in <tt>include/Basic/StmtNodes.td</tt> and
1875 creating a new class for your expression in the appropriate
1876 <tt>include/AST/Expr*.h</tt> header. It's best to look at the class
1877 for a similar expression to get ideas, and there are some specific
1878 things to watch for:
1879 <ul>
1880 <li>If you need to allocate memory, use the <tt>ASTContext</tt>
1881 allocator to allocate memory. Never use raw <tt>malloc</tt> or
1882 <tt>new</tt>, and never hold any resources in an AST node, because
1883 the destructor of an AST node is never called.</li>
1884
1885 <li>Make sure that <tt>getSourceRange()</tt> covers the exact
1886 source range of your expression. This is needed for diagnostics
1887 and for IDE support.</li>
1888
1889 <li>Make sure that <tt>children()</tt> visits all of the
1890 subexpressions. This is important for a number of features (e.g., IDE
1891 support, C++ variadic templates). If you have sub-types, you'll
1892 also need to visit those sub-types in the
1893 <tt>RecursiveASTVisitor</tt>.</li>
1894
1895 <li>Add printing support (<tt>StmtPrinter.cpp</tt>) and dumping
1896 support (<tt>StmtDumper.cpp</tt>) for your expression.</li>
1897
1898 <li>Add profiling support (<tt>StmtProfile.cpp</tt>) for your AST
1899 node, noting the distinguishing (non-source location)
1900 characteristics of an instance of your expression. Omitting this
1901 step will lead to hard-to-diagnose failures regarding matching of
1902 template declarations.</li>
1903 </ul>
1904 </li>
1905
1906 <li>Teach semantic analysis to build your AST node! At this point,
1907 you can wire up your <tt>Sema::BuildXXX</tt> function to actually
1908 create your AST. A few things to check at this point:
1909 <ul>
1910 <li>If your expression can construct a new C++ class or return a
1911 new Objective-C object, be sure to update and then call
1912 <tt>Sema::MaybeBindToTemporary</tt> for your just-created AST node
1913 to be sure that the object gets properly destructed. An easy way
1914 to test this is to return a C++ class with a private destructor:
1915 semantic analysis should flag an error here with the attempt to
1916 call the destructor.</li>
1917 <li>Inspect the generated AST by printing it using <tt>clang -cc1
1918 -ast-print</tt>, to make sure you're capturing all of the
1919 important information about how the AST was written.</li>
1920 <li>Inspect the generated AST under <tt>clang -cc1 -ast-dump</tt>
1921 to verify that all of the types in the generated AST line up the
1922 way you want them. Remember that clients of the AST should never
1923 have to "think" to understand what's going on. For example, all
1924 implicit conversions should show up explicitly in the AST.</li>
1925 <li>Write tests that use your expression as a subexpression of
1926 other, well-known expressions. Can you call a function using your
1927 expression as an argument? Can you use the ternary operator?</li>
1928 </ul>
1929 </li>
1930
1931 <li>Teach code generation to create IR to your AST node. This step
1932 is the first (and only) that requires knowledge of LLVM IR. There
1933 are several things to keep in mind:
1934 <ul>
1935 <li>Code generation is separated into scalar/aggregate/complex and
1936 lvalue/rvalue paths, depending on what kind of result your
1937 expression produces. On occasion, this requires some careful
1938 factoring of code to avoid duplication.</li>
1939
1940 <li><tt>CodeGenFunction</tt> contains functions
1941 <tt>ConvertType</tt> and <tt>ConvertTypeForMem</tt> that convert
1942 Clang's types (<tt>clang::Type*</tt> or <tt>clang::QualType</tt>)
1943 to LLVM types.
1944 Use the former for values, and the later for memory locations:
1945 test with the C++ "bool" type to check this. If you find
1946 that you are having to use LLVM bitcasts to make
1947 the subexpressions of your expression have the type that your
1948 expression expects, STOP! Go fix semantic analysis and the AST so
1949 that you don't need these bitcasts.</li>
1950
1951 <li>The <tt>CodeGenFunction</tt> class has a number of helper
1952 functions to make certain operations easy, such as generating code
1953 to produce an lvalue or an rvalue, or to initialize a memory
1954 location with a given value. Prefer to use these functions rather
1955 than directly writing loads and stores, because these functions
1956 take care of some of the tricky details for you (e.g., for
1957 exceptions).</li>
1958
1959 <li>If your expression requires some special behavior in the event
1960 of an exception, look at the <tt>push*Cleanup</tt> functions in
1961 <tt>CodeGenFunction</tt> to introduce a cleanup. You shouldn't
1962 have to deal with exception-handling directly.</li>
1963
1964 <li>Testing is extremely important in IR generation. Use <tt>clang
1965 -cc1 -emit-llvm</tt> and <a
1966 href="http://llvm.org/cmds/FileCheck.html">FileCheck</a> to verify
1967 that you're generating the right IR.</li>
1968 </ul>
1969 </li>
1970
1971 <li>Teach template instantiation how to cope with your AST
1972 node, which requires some fairly simple code:
1973 <ul>
1974 <li>Make sure that your expression's constructor properly
1975 computes the flags for type dependence (i.e., the type your
1976 expression produces can change from one instantiation to the
1977 next), value dependence (i.e., the constant value your expression
1978 produces can change from one instantiation to the next),
Douglas Gregord1cb2dc2011-10-14 00:54:15 +00001979 instantiation dependence (i.e., a template parameter occurs
Douglas Gregor1f634c62011-09-30 21:32:37 +00001980 anywhere in your expression), and whether your expression contains
1981 a parameter pack (for variadic templates). Often, computing these
1982 flags just means combining the results from the various types and
1983 subexpressions.</li>
1984
1985 <li>Add <tt>TransformXXX</tt> and <tt>RebuildXXX</tt> functions to
1986 the
1987 <tt>TreeTransform</tt> class template in <tt>Sema</tt>.
1988 <tt>TransformXXX</tt> should (recursively) transform all of the
1989 subexpressions and types
1990 within your expression, using <tt>getDerived().TransformYYY</tt>.
1991 If all of the subexpressions and types transform without error, it
1992 will then call the <tt>RebuildXXX</tt> function, which will in
1993 turn call <tt>getSema().BuildXXX</tt> to perform semantic analysis
1994 and build your expression.</li>
1995
1996 <li>To test template instantiation, take those tests you wrote to
1997 make sure that you were type checking with type-dependent
1998 expressions and dependent types (from step #2) and instantiate
1999 those templates with various types, some of which type-check and
2000 some that don't, and test the error messages in each case.</li>
2001 </ul>
2002 </li>
2003
2004 <li>There are some "extras" that make other features work better.
2005 It's worth handling these extras to give your expression complete
2006 integration into Clang:
2007 <ul>
2008 <li>Add code completion support for your expression in
2009 <tt>SemaCodeComplete.cpp</tt>.</li>
2010
2011 <li>If your expression has types in it, or has any "interesting"
2012 features other than subexpressions, extend libclang's
2013 <tt>CursorVisitor</tt> to provide proper visitation for your
2014 expression, enabling various IDE features such as syntax
2015 highlighting, cross-referencing, and so on. The
2016 <tt>c-index-test</tt> helper program can be used to test these
2017 features.</li>
2018 </ul>
2019 </li>
2020</ol>
2021
Ted Kremenek17a295d2008-06-11 06:19:49 +00002022</div>
2023</body>
Douglas Gregor2e1cd422008-11-17 14:58:09 +00002024</html>