Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / clang / cfa8d57f83697da82de4dbc452e822f759ba83e6 / . / docs / InternalsManual.html
blob: 05b45a6ab300254dcbd61c6b33bc09153590927f [file] [log] [blame]
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]1<html>
2<head>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]3<title>"Clang" CFE Internals Manual</title>
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]4<link type="text/css" rel="stylesheet" href="../menu.css" />
5<link type="text/css" rel="stylesheet" href="../content.css" />
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]6<style type="text/css">
7td {
8 vertical-align: top;
9}
10</style>
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]11</head>
12<body>
13
14<!--#include virtual="../menu.html.incl"-->
15
16<div id="content">
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]17
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]18<h1>"Clang" CFE Internals Manual</h1>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]19
20<ul>
21<li><a href="#intro">Introduction</a></li>
22<li><a href="#libsystem">LLVM System and Support Libraries</a></li>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]23<li><a href="#libbasic">The Clang 'Basic' Library</a>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]24 <ul>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]25 <li><a href="#Diagnostics">The Diagnostics Subsystem</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]26 <li><a href="#SourceLocation">The SourceLocation and SourceManager
27 classes</a></li>
28 </ul>
29</li>
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000[diff] [blame]30<li><a href="#libdriver">The Driver Library</a>
31 <ul>
32 </ul>
33</li>
Ted Kremenekcfa8d572009-04-09 18:08:18 +0000[diff] [blame^]34<li><a href="#pth">Pretokenized Headers (PTH)</a>
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000[diff] [blame]35<li><a href="#libfrontend">The Frontend Library</a>
36 <ul>
37 </ul>
38</li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]39<li><a href="#liblex">The Lexer and Preprocessor Library</a>
40 <ul>
41 <li><a href="#Token">The Token class</a></li>
42 <li><a href="#Lexer">The Lexer class</a></li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]43 <li><a href="#AnnotationToken">Annotation Tokens</a></li>
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]44 <li><a href="#TokenLexer">The TokenLexer class</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]45 <li><a href="#MultipleIncludeOpt">The MultipleIncludeOpt class</a></li>
46 </ul>
47</li>
48<li><a href="#libparse">The Parser Library</a>
49 <ul>
50 </ul>
51</li>
52<li><a href="#libast">The AST Library</a>
53 <ul>
54 <li><a href="#Type">The Type class and its subclasses</a></li>
55 <li><a href="#QualType">The QualType class</a></li>
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]56 <li><a href="#DeclarationName">Declaration names</a></li>
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]57 <li><a href="#DeclContext">Declaration contexts</a>
58 <ul>
59 <li><a href="#Redeclarations">Redeclarations and Overloads</a></li>
60 <li><a href="#LexicalAndSemanticContexts">Lexical and Semantic
61 Contexts</a></li>
62 <li><a href="#TransparentContexts">Transparent Declaration Contexts</a></li>
63 <li><a href="#MultiDeclContext">Multiply-Defined Declaration Contexts</a></li>
64 </ul>
65 </li>
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]66 <li><a href="#CFG">The CFG class</a></li>
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]67 <li><a href="#Constants">Constant Folding in the Clang AST</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]68 </ul>
69</li>
70</ul>
71
72
73<!-- ======================================================================= -->
74<h2 id="intro">Introduction</h2>
75<!-- ======================================================================= -->
76
77<p>This document describes some of the more important APIs and internal design
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]78decisions made in the Clang C front-end. The purpose of this document is to
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]79both capture some of this high level information and also describe some of the
80design decisions behind it. This is meant for people interested in hacking on
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]81Clang, not for end-users. The description below is categorized by
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]82libraries, and does not describe any of the clients of the libraries.</p>
83
84<!-- ======================================================================= -->
85<h2 id="libsystem">LLVM System and Support Libraries</h2>
86<!-- ======================================================================= -->
87
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]88<p>The LLVM libsystem library provides the basic Clang system abstraction layer,
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]89which is used for file system access. The LLVM libsupport library provides many
90underlying libraries and <a
91href="http://llvm.org/docs/ProgrammersManual.html">data-structures</a>,
92 including command line option
93processing and various containers.</p>
94
95<!-- ======================================================================= -->
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]96<h2 id="libbasic">The Clang 'Basic' Library</h2>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]97<!-- ======================================================================= -->
98
99<p>This library certainly needs a better name. The 'basic' library contains a
100number of low-level utilities for tracking and manipulating source buffers,
101locations within the source buffers, diagnostics, tokens, target abstraction,
102and information about the subset of the language being compiled for.</p>
103
104<p>Part of this infrastructure is specific to C (such as the TargetInfo class),
105other parts could be reused for other non-C-based languages (SourceLocation,
106SourceManager, Diagnostics, FileManager). When and if there is future demand
107we can figure out if it makes sense to introduce a new library, move the general
108classes somewhere else, or introduce some other solution.</p>
109
110<p>We describe the roles of these classes in order of their dependencies.</p>
111
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]112
113<!-- ======================================================================= -->
114<h3 id="Diagnostics">The Diagnostics Subsystem</h3>
115<!-- ======================================================================= -->
116
117<p>The Clang Diagnostics subsystem is an important part of how the compiler
118communicates with the human. Diagnostics are the warnings and errors produced
119when the code is incorrect or dubious. In Clang, each diagnostic produced has
120(at the minimum) a unique ID, a <a href="#SourceLocation">SourceLocation</a> to
121"put the caret", an English translation associated with it, and a severity (e.g.
122<tt>WARNING</tt> or <tt>ERROR</tt>). They can also optionally include a number
123of arguments to the dianostic (which fill in "%0"'s in the string) as well as a
124number of source ranges that related to the diagnostic.</p>
125
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]126<p>In this section, we'll be giving examples produced by the Clang command line
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]127driver, but diagnostics can be <a href="#DiagnosticClient">rendered in many
128different ways</a> depending on how the DiagnosticClient interface is
129implemented. A representative example of a diagonstic is:</p>
130
131<pre>
132t.c:38:15: error: invalid operands to binary expression ('int *' and '_Complex float')
133 <font color="darkgreen">P = (P-42) + Gamma*4;</font>
134 <font color="blue">~~~~~~ ^ ~~~~~~~</font>
135</pre>
136
137<p>In this example, you can see the English translation, the severity (error),
138you can see the source location (the caret ("^") and file/line/column info),
139the source ranges "~~~~", arguments to the diagnostic ("int*" and "_Complex
140float"). You'll have to believe me that there is a unique ID backing the
141diagnostic :).</p>
142
143<p>Getting all of this to happen has several steps and involves many moving
144pieces, this section describes them and talks about best practices when adding
145a new diagnostic.</p>
146
147<!-- ============================ -->
148<h4>The DiagnosticKinds.def file</h4>
149<!-- ============================ -->
150
151<p>Diagnostics are created by adding an entry to the <tt><a
152href="http://llvm.org/svn/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticKinds.def"
153>DiagnosticKinds.def</a></tt> file. This file encodes the unique ID of the
154diagnostic (as an enum, the first argument), the severity of the diagnostic
155(second argument) and the English translation + format string.</p>
156
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 +0000[diff] [blame]173level is used to staple more information onto previous diagnostics.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]174
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 +0000[diff] [blame]177<tt>Error</tt>, <tt>Fatal</tt> }) of output <em>levels</em> by the diagnostics
Chris Lattnera180fdd2009-02-17 07:07:29 +0000[diff] [blame]178subsystem 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 +0000[diff] [blame]198
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 +0000[diff] [blame]212 "invalid operands to binary expression (<b>%0</b> and <b>%1</b>)"
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]213 "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 +0000[diff] [blame]222 Clang uses the "%...[digit]" sequences to specify where and how arguments to
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]223 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 +0000[diff] [blame]228 to your diagnostic, you are doing something wrong :). Unlike printf, there
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]229 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
239 <tt>DiagnosticKinds.def</tt> file. This avoids the diagnostic wrapping when
240 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
254this prevents <a href="translation">translating</a> the Clang diagnostics to
255other 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 +0000[diff] [blame]260including 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 +0000[diff] [blame]262
263<!-- ==================================== -->
264<h4>Formatting a Diagnostic Argument</a></h4>
265<!-- ==================================== -->
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 +0000[diff] [blame]280<tr><td>Class:</td><td>Integers</td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]281<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 +0000[diff] [blame]284 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 +0000[diff] [blame]286
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 +0000[diff] [blame]290<tr><td>Class:</td><td>Integers</td></tr>
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]291<tr><td>Description:</td><td>This format specifier is used to merge multiple
292 related diagnostics together into one common one, without requiring the
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]293 difference to be specified as an English string argument. Instead of
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]294 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
300 things textually.</td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]301
302<tr><td colspan="2"><b>"plural" format</b></td></tr>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]303<tr><td>Example:</td><td><tt>"you have %1 %plural{1:mouse|:mice}1 connected to
304 your computer"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]305<tr><td>Class:</td><td>Integers</td></tr>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]306<tr><td>Description:</td><td><p>This is a formatter for complex plural forms.
307 It is designed to handle even the requirements of languages with very
308 complex plural forms, as many Baltic languages have. The argument consists
309 of a series of expression/form pairs, separated by ':', where the first form
310 whose expression evaluates to true is the result of the modifier.</p>
311 <p>An expression can be empty, in which case it is always true. See the
312 example at the top. Otherwise, it is a series of one or more numeric
313 conditions, separated by ','. If any condition matches, the expression
314 matches. Each numeric condition can take one of three forms.</p>
315 <ul>
316 <li>number: A simple decimal number matches if the argument is the same
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]317 as the number. Example: <tt>"%plural{1:mouse|:mice}4"</tt></li>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]318 <li>range: A range in square brackets matches if the argument is within
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]319 the range. Then range is inclusive on both ends. Example:
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]320 <tt>"%plural{0:none|1:one|[2,5]:some|:many}2"</tt></li>
321 <li>modulo: A modulo operator is followed by a number, and
322 equals sign and either a number or a range. The tests are the
323 same as for plain
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]324 numbers and ranges, but the argument is taken modulo the number first.
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]325 Example: <tt>"%plural{%100=0:even hundred|%100=[1,50]:lower half|:everything
326 else}1"</tt></li>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]327 </ul>
328 <p>The parser is very unforgiving. A syntax error, even whitespace, will
329 abort, as will a failure to match the argument against any
330 expression.</p></td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]331
Chris Lattner077bf5e2008-11-24 03:33:13 +0000[diff] [blame]332<tr><td colspan="2"><b>"objcclass" format</b></td></tr>
333<tr><td>Example:</td><td><tt>"method %objcclass0 not found"</tt></td></tr>
334<tr><td>Class:</td><td>DeclarationName</td></tr>
335<tr><td>Description:</td><td><p>This is a simple formatter that indicates the
336 DeclarationName corresponds to an Objective-C class method selector. As
337 such, it prints the selector with a leading '+'.</p></td></tr>
338
339<tr><td colspan="2"><b>"objcinstance" format</b></td></tr>
340<tr><td>Example:</td><td><tt>"method %objcinstance0 not found"</tt></td></tr>
341<tr><td>Class:</td><td>DeclarationName</td></tr>
342<tr><td>Description:</td><td><p>This is a simple formatter that indicates the
343 DeclarationName corresponds to an Objective-C instance method selector. As
344 such, it prints the selector with a leading '-'.</p></td></tr>
345
Douglas Gregor47b9a1c2009-02-04 17:27:36 +0000[diff] [blame]346<tr><td colspan="2"><b>"q" format</b></td></tr>
347<tr><td>Example:</td><td><tt>"candidate found by name lookup is %q0"</tt></td></tr>
348<tr><td>Class:</td><td>NamedDecl*</td></tr>
349<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>
350
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]351</table>
352
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]353<p>It is really easy to add format specifiers to the Clang diagnostics system,
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]354but they should be discussed before they are added. If you are creating a lot
355of repetitive diagnostics and/or have an idea for a useful formatter, please
356bring it up on the cfe-dev mailing list.</p>
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]357
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]358<!-- ===================================================== -->
359<h4><a name="#producingdiag">Producing the Diagnostic</a></h4>
360<!-- ===================================================== -->
361
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]362<p>Now that you've created the diagnostic in the DiagnosticKinds.def file, you
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]363need to write the code that detects the condition in question and emits the
364new diagnostic. Various components of Clang (e.g. the preprocessor, Sema,
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]365etc) provide a helper function named "Diag". It creates a diagnostic and
366accepts the arguments, ranges, and other information that goes along with
367it.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]368
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]369<p>For example, the binary expression error comes from code like this:</p>
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]370
371<pre>
372 if (various things that are bad)
373 Diag(Loc, diag::err_typecheck_invalid_operands)
374 &lt;&lt; lex-&gt;getType() &lt;&lt; rex-&gt;getType()
375 &lt;&lt; lex-&gt;getSourceRange() &lt;&lt; rex-&gt;getSourceRange();
376</pre>
377
378<p>This shows that use of the Diag method: they take a location (a <a
379href="#SourceLocation">SourceLocation</a> object) and a diagnostic enum value
380(which matches the name from DiagnosticKinds.def). If the diagnostic takes
381arguments, they are specified with the &lt;&lt; operator: the first argument
382becomes %0, the second becomes %1, etc. The diagnostic interface allows you to
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]383specify arguments of many different types, including <tt>int</tt> and
384<tt>unsigned</tt> for integer arguments, <tt>const char*</tt> and
385<tt>std::string</tt> for string arguments, <tt>DeclarationName</tt> and
386<tt>const IdentifierInfo*</tt> for names, <tt>QualType</tt> for types, etc.
387SourceRanges are also specified with the &lt;&lt; operator, but do not have a
388specific ordering requirement.</p>
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]389
390<p>As you can see, adding and producing a diagnostic is pretty straightforward.
391The hard part is deciding exactly what you need to say to help the user, picking
392a suitable wording, and providing the information needed to format it correctly.
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]393The good news is that the call site that issues a diagnostic should be
394completely independent of how the diagnostic is formatted and in what language
395it is rendered.
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]396</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]397
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]398<!-- ==================================================== -->
399<h4 id="code-modification-hints">Code Modification Hints</h4>
400<!-- ==================================================== -->
401
402<p>In some cases, the front end emits diagnostics when it is clear
403that some small change to the source code would fix the problem. For
404example, a missing semicolon at the end of a statement or a use of
Chris Lattner34c05332009-02-27 19:31:12 +0000[diff] [blame]405deprecated syntax that is easily rewritten into a more modern form.
406Clang tries very hard to emit the diagnostic and recover gracefully
407in these and other cases.</p>
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]408
Chris Lattner34c05332009-02-27 19:31:12 +0000[diff] [blame]409<p>However, for these cases where the fix is obvious, the diagnostic
410can be annotated with a code
411modification "hint" that describes how to change the code referenced
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000[diff] [blame]412by the diagnostic to fix the problem. For example, it might add the
413missing semicolon at the end of the statement or rewrite the use of a
414deprecated construct into something more palatable. Here is one such
415example C++ front end, where we warn about the right-shift operator
416changing meaning from C++98 to C++0x:</p>
417
418<pre>
419test.cpp:3:7: warning: use of right-shift operator ('&gt;&gt;') in template argument will require parentheses in C++0x
420A&lt;100 &gt;&gt; 2&gt; *a;
421 ^
422 ( )
423</pre>
424
425<p>Here, the code modification hint is suggesting that parentheses be
426added, and showing exactly where those parentheses would be inserted
427into the source code. The code modification hints themselves describe
428what changes to make to the source code in an abstract manner, which
429the text diagnostic printer renders as a line of "insertions" below
430the caret line. <a href="#DiagnosticClient">Other diagnostic
431clients</a> might choose to render the code differently (e.g., as
432markup inline) or even give the user the ability to automatically fix
433the problem.</p>
434
435<p>All code modification hints are described by the
436<code>CodeModificationHint</code> class, instances of which should be
437attached to the diagnostic using the &lt;&lt; operator in the same way
438that highlighted source ranges and arguments are passed to the
439diagnostic. Code modification hints can be created with one of three
440constructors:</p>
441
442<dl>
443 <dt><code>CodeModificationHint::CreateInsertion(Loc, Code)</code></dt>
444 <dd>Specifies that the given <code>Code</code> (a string) should be inserted
445 before the source location <code>Loc</code>.</dd>
446
447 <dt><code>CodeModificationHint::CreateRemoval(Range)</code></dt>
448 <dd>Specifies that the code in the given source <code>Range</code>
449 should be removed.</dd>
450
451 <dt><code>CodeModificationHint::CreateReplacement(Range, Code)</code></dt>
452 <dd>Specifies that the code in the given source <code>Range</code>
453 should be removed, and replaced with the given <code>Code</code> string.</dd>
454</dl>
455
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]456<!-- ============================================================= -->
457<h4><a name="DiagnosticClient">The DiagnosticClient Interface</a></h4>
458<!-- ============================================================= -->
459
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]460<p>Once code generates a diagnostic with all of the arguments and the rest of
461the relevant information, Clang needs to know what to do with it. As previously
462mentioned, the diagnostic machinery goes through some filtering to map a
463severity onto a diagnostic level, then (assuming the diagnostic is not mapped to
464"<tt>Ignore</tt>") it invokes an object that implements the DiagnosticClient
465interface with the information.</p>
466
467<p>It is possible to implement this interface in many different ways. For
468example, the normal Clang DiagnosticClient (named 'TextDiagnosticPrinter') turns
469the arguments into strings (according to the various formatting rules), prints
470out the file/line/column information and the string, then prints out the line of
471code, the source ranges, and the caret. However, this behavior isn't required.
472</p>
473
474<p>Another implementation of the DiagnosticClient interface is the
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]475'TextDiagnosticBuffer' class, which is used when Clang is in -verify mode.
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]476Instead of formatting and printing out the diagnostics, this implementation just
477captures and remembers the diagnostics as they fly by. Then -verify compares
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]478the list of produced diagnostics to the list of expected ones. If they disagree,
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]479it prints out its own output.
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]480</p>
481
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]482<p>There are many other possible implementations of this interface, and this is
483why we prefer diagnostics to pass down rich structured information in arguments.
484For example, an HTML output might want declaration names be linkified to where
485they come from in the source. Another example is that a GUI might let you click
486on typedefs to expand them. This application would want to pass significantly
487more information about types through to the GUI than a simple flat string. The
488interface allows this to happen.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]489
490<!-- ====================================================== -->
491<h4><a name="translation">Adding Translations to Clang</a></h4>
492<!-- ====================================================== -->
493
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]494<p>Not possible yet! Diagnostic strings should be written in UTF-8, the client
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]495can translate to the relevant code page if needed. Each translation completely
496replaces the format string for the diagnostic.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]497
498
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]499<!-- ======================================================================= -->
500<h3 id="SourceLocation">The SourceLocation and SourceManager classes</h3>
501<!-- ======================================================================= -->
502
503<p>Strangely enough, the SourceLocation class represents a location within the
504source code of the program. Important design points include:</p>
505
506<ol>
507<li>sizeof(SourceLocation) must be extremely small, as these are embedded into
508 many AST nodes and are passed around often. Currently it is 32 bits.</li>
509<li>SourceLocation must be a simple value object that can be efficiently
510 copied.</li>
511<li>We should be able to represent a source location for any byte of any input
512 file. This includes in the middle of tokens, in whitespace, in trigraphs,
513 etc.</li>
514<li>A SourceLocation must encode the current #include stack that was active when
515 the location was processed. For example, if the location corresponds to a
516 token, it should contain the set of #includes active when the token was
517 lexed. This allows us to print the #include stack for a diagnostic.</li>
518<li>SourceLocation must be able to describe macro expansions, capturing both
519 the ultimate instantiation point and the source of the original character
520 data.</li>
521</ol>
522
523<p>In practice, the SourceLocation works together with the SourceManager class
Chris Lattner18376dd2009-01-16 07:00:50 +0000[diff] [blame]524to encode two pieces of information about a location: it's spelling location
Chris Lattner88054de2009-01-16 07:15:35 +0000[diff] [blame]525and it's instantiation location. For most tokens, these will be the same. However,
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]526for a macro expansion (or tokens that came from a _Pragma directive) these will
527describe the location of the characters corresponding to the token and the
528location where the token was used (i.e. the macro instantiation point or the
529location of the _Pragma itself).</p>
530
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]531<p>For efficiency, we only track one level of macro instantiations: if a token was
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]532produced by multiple instantiations, we only track the source and ultimate
533destination. Though we could track the intermediate instantiation points, this
534would require extra bookkeeping and no known client would benefit substantially
535from this.</p>
536
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]537<p>The Clang front-end inherently depends on the location of a token being
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]538tracked correctly. If it is ever incorrect, the front-end may get confused and
539die. The reason for this is that the notion of the 'spelling' of a Token in
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]540Clang depends on being able to find the original input characters for the token.
Chris Lattner18376dd2009-01-16 07:00:50 +0000[diff] [blame]541This concept maps directly to the "spelling location" for the token.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]542
543<!-- ======================================================================= -->
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000[diff] [blame]544<h2 id="libdriver">The Driver Library</h2>
545<!-- ======================================================================= -->
546
Ted Kremenekcfa8d572009-04-09 18:08:18 +0000[diff] [blame^]547<p>The clang Driver and library are documented <a
548href="DriverInternals.html">here<a>.<p>
549
550<!-- ======================================================================= -->
551<h2 id="pth">Pretokenized Headers (PTH)</h2>
552<!-- ======================================================================= -->
553
554<p>Clang's current implementation of precompiled headers, known as PTH, is
555documented <a href="PTHInternals.html">here</a>.</p>
Daniel Dunbar27d9e9f2009-03-30 06:50:01 +0000[diff] [blame]556
557<!-- ======================================================================= -->
558<h2 id="libfrontend">The Frontend Library</h2>
559<!-- ======================================================================= -->
560
561<p>The Frontend library contains functionality useful for building
562tools on top of the clang libraries, for example several methods for
563outputting diagnostics.</p>
564
565<!-- ======================================================================= -->
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]566<h2 id="liblex">The Lexer and Preprocessor Library</h2>
567<!-- ======================================================================= -->
568
569<p>The Lexer library contains several tightly-connected classes that are involved
570with the nasty process of lexing and preprocessing C source code. The main
571interface to this library for outside clients is the large <a
572href="#Preprocessor">Preprocessor</a> class.
573It contains the various pieces of state that are required to coherently read
574tokens out of a translation unit.</p>
575
576<p>The core interface to the Preprocessor object (once it is set up) is the
577Preprocessor::Lex method, which returns the next <a href="#Token">Token</a> from
578the preprocessor stream. There are two types of token providers that the
579preprocessor is capable of reading from: a buffer lexer (provided by the <a
580href="#Lexer">Lexer</a> class) and a buffered token stream (provided by the <a
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]581href="#TokenLexer">TokenLexer</a> class).
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]582
583
584<!-- ======================================================================= -->
585<h3 id="Token">The Token class</h3>
586<!-- ======================================================================= -->
587
588<p>The Token class is used to represent a single lexed token. Tokens are
589intended to be used by the lexer/preprocess and parser libraries, but are not
590intended to live beyond them (for example, they should not live in the ASTs).<p>
591
592<p>Tokens most often live on the stack (or some other location that is efficient
593to access) as the parser is running, but occasionally do get buffered up. For
594example, macro definitions are stored as a series of tokens, and the C++
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]595front-end periodically needs to buffer tokens up for tentative parsing and
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]596various pieces of look-ahead. As such, the size of a Token matter. On a 32-bit
597system, sizeof(Token) is currently 16 bytes.</p>
598
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]599<p>Tokens occur in two forms: "<a href="#AnnotationToken">Annotation
600Tokens</a>" and normal tokens. Normal tokens are those returned by the lexer,
601annotation tokens represent semantic information and are produced by the parser,
602replacing normal tokens in the token stream. Normal tokens contain the
603following information:</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]604
605<ul>
606<li><b>A SourceLocation</b> - This indicates the location of the start of the
607token.</li>
608
609<li><b>A length</b> - This stores the length of the token as stored in the
610SourceBuffer. For tokens that include them, this length includes trigraphs and
611escaped newlines which are ignored by later phases of the compiler. By pointing
612into the original source buffer, it is always possible to get the original
613spelling of a token completely accurately.</li>
614
615<li><b>IdentifierInfo</b> - If a token takes the form of an identifier, and if
616identifier lookup was enabled when the token was lexed (e.g. the lexer was not
617reading in 'raw' mode) this contains a pointer to the unique hash value for the
618identifier. Because the lookup happens before keyword identification, this
619field is set even for language keywords like 'for'.</li>
620
621<li><b>TokenKind</b> - This indicates the kind of token as classified by the
622lexer. This includes things like <tt>tok::starequal</tt> (for the "*="
623operator), <tt>tok::ampamp</tt> for the "&amp;&amp;" token, and keyword values
624(e.g. <tt>tok::kw_for</tt>) for identifiers that correspond to keywords. Note
625that some tokens can be spelled multiple ways. For example, C++ supports
626"operator keywords", where things like "and" are treated exactly like the
627"&amp;&amp;" operator. In these cases, the kind value is set to
628<tt>tok::ampamp</tt>, which is good for the parser, which doesn't have to
629consider both forms. For something that cares about which form is used (e.g.
630the preprocessor 'stringize' operator) the spelling indicates the original
631form.</li>
632
633<li><b>Flags</b> - There are currently four flags tracked by the
634lexer/preprocessor system on a per-token basis:
635
636 <ol>
637 <li><b>StartOfLine</b> - This was the first token that occurred on its input
638 source line.</li>
639 <li><b>LeadingSpace</b> - There was a space character either immediately
640 before the token or transitively before the token as it was expanded
641 through a macro. The definition of this flag is very closely defined by
642 the stringizing requirements of the preprocessor.</li>
643 <li><b>DisableExpand</b> - This flag is used internally to the preprocessor to
644 represent identifier tokens which have macro expansion disabled. This
645 prevents them from being considered as candidates for macro expansion ever
646 in the future.</li>
647 <li><b>NeedsCleaning</b> - This flag is set if the original spelling for the
648 token includes a trigraph or escaped newline. Since this is uncommon,
649 many pieces of code can fast-path on tokens that did not need cleaning.
650 </p>
651 </ol>
652</li>
653</ul>
654
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]655<p>One interesting (and somewhat unusual) aspect of normal tokens is that they
656don't contain any semantic information about the lexed value. For example, if
657the token was a pp-number token, we do not represent the value of the number
658that was lexed (this is left for later pieces of code to decide). Additionally,
659the lexer library has no notion of typedef names vs variable names: both are
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]660returned as identifiers, and the parser is left to decide whether a specific
661identifier is a typedef or a variable (tracking this requires scope information
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]662among other things). The parser can do this translation by replacing tokens
663returned by the preprocessor with "Annotation Tokens".</p>
664
665<!-- ======================================================================= -->
666<h3 id="AnnotationToken">Annotation Tokens</h3>
667<!-- ======================================================================= -->
668
669<p>Annotation Tokens are tokens that are synthesized by the parser and injected
670into the preprocessor's token stream (replacing existing tokens) to record
671semantic information found by the parser. For example, if "foo" is found to be
672a typedef, the "foo" <tt>tok::identifier</tt> token is replaced with an
673<tt>tok::annot_typename</tt>. This is useful for a couple of reasons: 1) this
674makes it easy to handle qualified type names (e.g. "foo::bar::baz&lt;42&gt;::t")
675in C++ as a single "token" in the parser. 2) if the parser backtracks, the
676reparse does not need to redo semantic analysis to determine whether a token
677sequence is a variable, type, template, etc.</p>
678
679<p>Annotation Tokens are created by the parser and reinjected into the parser's
680token stream (when backtracking is enabled). Because they can only exist in
681tokens that the preprocessor-proper is done with, it doesn't need to keep around
682flags like "start of line" that the preprocessor uses to do its job.
683Additionally, an annotation token may "cover" a sequence of preprocessor tokens
684(e.g. <tt>a::b::c</tt> is five preprocessor tokens). As such, the valid fields
685of an annotation token are different than the fields for a normal token (but
686they are multiplexed into the normal Token fields):</p>
687
688<ul>
689<li><b>SourceLocation "Location"</b> - The SourceLocation for the annotation
690token indicates the first token replaced by the annotation token. In the example
691above, it would be the location of the "a" identifier.</li>
692
693<li><b>SourceLocation "AnnotationEndLoc"</b> - This holds the location of the
694last token replaced with the annotation token. In the example above, it would
695be the location of the "c" identifier.</li>
696
697<li><b>void* "AnnotationValue"</b> - This contains an opaque object that the
698parser gets from Sema through an Actions module, it is passed around and Sema
699intepretes it, based on the type of annotation token.</li>
700
701<li><b>TokenKind "Kind"</b> - This indicates the kind of Annotation token this
702is. See below for the different valid kinds.</li>
703</ul>
704
705<p>Annotation tokens currently come in three kinds:</p>
706
707<ol>
708<li><b>tok::annot_typename</b>: This annotation token represents a
709resolved typename token that is potentially qualified. The AnnotationValue
Steve Naroffb43a50f2009-01-28 19:39:02 +0000[diff] [blame]710field contains a pointer returned by Action::getTypeName(). In the case of the
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]711Sema actions module, this is a <tt>Decl*</tt> for the type.</li>
712
713<li><b>tok::annot_cxxscope</b>: This annotation token represents a C++ scope
714specifier, such as "A::B::". This corresponds to the grammar productions "::"
715and ":: [opt] nested-name-specifier". The AnnotationValue pointer is returned
716by the Action::ActOnCXXGlobalScopeSpecifier and
717Action::ActOnCXXNestedNameSpecifier callbacks. In the case of Sema, this is a
718<tt>DeclContext*</tt>.</li>
719
Douglas Gregor39a8de12009-02-25 19:37:18 +0000[diff] [blame]720<li><b>tok::annot_template_id</b>: This annotation token represents a
721C++ template-id such as "foo&lt;int, 4&gt;", where "foo" is the name
722of a template. The AnnotationValue pointer is a pointer to a malloc'd
723TemplateIdAnnotation object. Depending on the context, a parsed template-id that names a type might become a typename annotation token (if all we care about is the named type, e.g., because it occurs in a type specifier) or might remain a template-id token (if we want to retain more source location information or produce a new type, e.g., in a declaration of a class template specialization). template-id annotation tokens that refer to a type can be "upgraded" to typename annotation tokens by the parser.</li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]724
725</ol>
726
Cedric Venetda76b282009-01-06 16:22:54 +0000[diff] [blame]727<p>As mentioned above, annotation tokens are not returned by the preprocessor,
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]728they are formed on demand by the parser. This means that the parser has to be
729aware of cases where an annotation could occur and form it where appropriate.
730This is somewhat similar to how the parser handles Translation Phase 6 of C99:
731String Concatenation (see C99 5.1.1.2). In the case of string concatenation,
732the preprocessor just returns distinct tok::string_literal and
733tok::wide_string_literal tokens and the parser eats a sequence of them wherever
734the grammar indicates that a string literal can occur.</p>
735
736<p>In order to do this, whenever the parser expects a tok::identifier or
737tok::coloncolon, it should call the TryAnnotateTypeOrScopeToken or
738TryAnnotateCXXScopeToken methods to form the annotation token. These methods
739will maximally form the specified annotation tokens and replace the current
740token with them, if applicable. If the current tokens is not valid for an
741annotation token, it will remain an identifier or :: token.</p>
742
743
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]744
745<!-- ======================================================================= -->
746<h3 id="Lexer">The Lexer class</h3>
747<!-- ======================================================================= -->
748
749<p>The Lexer class provides the mechanics of lexing tokens out of a source
750buffer and deciding what they mean. The Lexer is complicated by the fact that
751it operates on raw buffers that have not had spelling eliminated (this is a
752necessity to get decent performance), but this is countered with careful coding
753as well as standard performance techniques (for example, the comment handling
754code is vectorized on X86 and PowerPC hosts).</p>
755
756<p>The lexer has a couple of interesting modal features:</p>
757
758<ul>
759<li>The lexer can operate in 'raw' mode. This mode has several features that
760 make it possible to quickly lex the file (e.g. it stops identifier lookup,
761 doesn't specially handle preprocessor tokens, handles EOF differently, etc).
762 This mode is used for lexing within an "<tt>#if 0</tt>" block, for
763 example.</li>
764<li>The lexer can capture and return comments as tokens. This is required to
765 support the -C preprocessor mode, which passes comments through, and is
766 used by the diagnostic checker to identifier expect-error annotations.</li>
767<li>The lexer can be in ParsingFilename mode, which happens when preprocessing
Chris Lattner84386242007-09-16 19:25:23 +0000[diff] [blame]768 after reading a #include directive. This mode changes the parsing of '&lt;'
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]769 to return an "angled string" instead of a bunch of tokens for each thing
770 within the filename.</li>
771<li>When parsing a preprocessor directive (after "<tt>#</tt>") the
772 ParsingPreprocessorDirective mode is entered. This changes the parser to
773 return EOM at a newline.</li>
774<li>The Lexer uses a LangOptions object to know whether trigraphs are enabled,
775 whether C++ or ObjC keywords are recognized, etc.</li>
776</ul>
777
778<p>In addition to these modes, the lexer keeps track of a couple of other
779 features that are local to a lexed buffer, which change as the buffer is
780 lexed:</p>
781
782<ul>
783<li>The Lexer uses BufferPtr to keep track of the current character being
784 lexed.</li>
785<li>The Lexer uses IsAtStartOfLine to keep track of whether the next lexed token
786 will start with its "start of line" bit set.</li>
787<li>The Lexer keeps track of the current #if directives that are active (which
788 can be nested).</li>
789<li>The Lexer keeps track of an <a href="#MultipleIncludeOpt">
790 MultipleIncludeOpt</a> object, which is used to
791 detect whether the buffer uses the standard "<tt>#ifndef XX</tt> /
792 <tt>#define XX</tt>" idiom to prevent multiple inclusion. If a buffer does,
793 subsequent includes can be ignored if the XX macro is defined.</li>
794</ul>
795
796<!-- ======================================================================= -->
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]797<h3 id="TokenLexer">The TokenLexer class</h3>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]798<!-- ======================================================================= -->
799
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]800<p>The TokenLexer class is a token provider that returns tokens from a list
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]801of tokens that came from somewhere else. It typically used for two things: 1)
802returning tokens from a macro definition as it is being expanded 2) returning
803tokens from an arbitrary buffer of tokens. The later use is used by _Pragma and
804will most likely be used to handle unbounded look-ahead for the C++ parser.</p>
805
806<!-- ======================================================================= -->
807<h3 id="MultipleIncludeOpt">The MultipleIncludeOpt class</h3>
808<!-- ======================================================================= -->
809
810<p>The MultipleIncludeOpt class implements a really simple little state machine
811that is used to detect the standard "<tt>#ifndef XX</tt> / <tt>#define XX</tt>"
812idiom that people typically use to prevent multiple inclusion of headers. If a
813buffer uses this idiom and is subsequently #include'd, the preprocessor can
814simply check to see whether the guarding condition is defined or not. If so,
815the preprocessor can completely ignore the include of the header.</p>
816
817
818
819<!-- ======================================================================= -->
820<h2 id="libparse">The Parser Library</h2>
821<!-- ======================================================================= -->
822
823<!-- ======================================================================= -->
824<h2 id="libast">The AST Library</h2>
825<!-- ======================================================================= -->
826
827<!-- ======================================================================= -->
828<h3 id="Type">The Type class and its subclasses</h3>
829<!-- ======================================================================= -->
830
831<p>The Type class (and its subclasses) are an important part of the AST. Types
832are accessed through the ASTContext class, which implicitly creates and uniques
833them as they are needed. Types have a couple of non-obvious features: 1) they
834do not capture type qualifiers like const or volatile (See
835<a href="#QualType">QualType</a>), and 2) they implicitly capture typedef
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]836information. Once created, types are immutable (unlike decls).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]837
838<p>Typedefs in C make semantic analysis a bit more complex than it would
839be without them. The issue is that we want to capture typedef information
840and represent it in the AST perfectly, but the semantics of operations need to
841"see through" typedefs. For example, consider this code:</p>
842
843<code>
844void func() {<br>
Bill Wendling30d17752007-10-06 01:56:01 +0000[diff] [blame]845&nbsp;&nbsp;typedef int foo;<br>
846&nbsp;&nbsp;foo X, *Y;<br>
847&nbsp;&nbsp;typedef foo* bar;<br>
848&nbsp;&nbsp;bar Z;<br>
849&nbsp;&nbsp;*X; <i>// error</i><br>
850&nbsp;&nbsp;**Y; <i>// error</i><br>
851&nbsp;&nbsp;**Z; <i>// error</i><br>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]852}<br>
853</code>
854
855<p>The code above is illegal, and thus we expect there to be diagnostics emitted
856on the annotated lines. In this example, we expect to get:</p>
857
858<pre>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]859<b>test.c:6:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]860*X; // error
861<font color="blue">^~</font>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]862<b>test.c:7:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]863**Y; // error
864<font color="blue">^~~</font>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]865<b>test.c:8:1: error: indirection requires pointer operand ('foo' invalid)</b>
866**Z; // error
867<font color="blue">^~~</font>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]868</pre>
869
870<p>While this example is somewhat silly, it illustrates the point: we want to
871retain typedef information where possible, so that we can emit errors about
872"<tt>std::string</tt>" instead of "<tt>std::basic_string&lt;char, std:...</tt>".
873Doing this requires properly keeping typedef information (for example, the type
874of "X" is "foo", not "int"), and requires properly propagating it through the
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]875various operators (for example, the type of *Y is "foo", not "int"). In order
876to retain this information, the type of these expressions is an instance of the
877TypedefType class, which indicates that the type of these expressions is a
878typedef for foo.
879</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]880
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]881<p>Representing types like this is great for diagnostics, because the
882user-specified type is always immediately available. There are two problems
883with this: first, various semantic checks need to make judgements about the
Chris Lattner33fc68a2007-07-31 18:54:50 +0000[diff] [blame]884<em>actual structure</em> of a type, ignoring typdefs. Second, we need an
885efficient way to query whether two types are structurally identical to each
886other, ignoring typedefs. The solution to both of these problems is the idea of
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]887canonical types.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]888
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]889<!-- =============== -->
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]890<h4>Canonical Types</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]891<!-- =============== -->
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]892
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]893<p>Every instance of the Type class contains a canonical type pointer. For
894simple types with no typedefs involved (e.g. "<tt>int</tt>", "<tt>int*</tt>",
895"<tt>int**</tt>"), the type just points to itself. For types that have a
896typedef somewhere in their structure (e.g. "<tt>foo</tt>", "<tt>foo*</tt>",
897"<tt>foo**</tt>", "<tt>bar</tt>"), the canonical type pointer points to their
898structurally equivalent type without any typedefs (e.g. "<tt>int</tt>",
899"<tt>int*</tt>", "<tt>int**</tt>", and "<tt>int*</tt>" respectively).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]900
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]901<p>This design provides a constant time operation (dereferencing the canonical
902type pointer) that gives us access to the structure of types. For example,
903we can trivially tell that "bar" and "foo*" are the same type by dereferencing
904their canonical type pointers and doing a pointer comparison (they both point
905to the single "<tt>int*</tt>" type).</p>
906
907<p>Canonical types and typedef types bring up some complexities that must be
908carefully managed. Specifically, the "isa/cast/dyncast" operators generally
909shouldn't be used in code that is inspecting the AST. For example, when type
910checking the indirection operator (unary '*' on a pointer), the type checker
911must verify that the operand has a pointer type. It would not be correct to
912check that with "<tt>isa&lt;PointerType&gt;(SubExpr-&gt;getType())</tt>",
913because this predicate would fail if the subexpression had a typedef type.</p>
914
915<p>The solution to this problem are a set of helper methods on Type, used to
916check their properties. In this case, it would be correct to use
917"<tt>SubExpr-&gt;getType()-&gt;isPointerType()</tt>" to do the check. This
918predicate will return true if the <em>canonical type is a pointer</em>, which is
919true any time the type is structurally a pointer type. The only hard part here
920is remembering not to use the <tt>isa/cast/dyncast</tt> operations.</p>
921
922<p>The second problem we face is how to get access to the pointer type once we
923know it exists. To continue the example, the result type of the indirection
924operator is the pointee type of the subexpression. In order to determine the
925type, we need to get the instance of PointerType that best captures the typedef
926information in the program. If the type of the expression is literally a
927PointerType, we can return that, otherwise we have to dig through the
928typedefs to find the pointer type. For example, if the subexpression had type
929"<tt>foo*</tt>", we could return that type as the result. If the subexpression
930had type "<tt>bar</tt>", we want to return "<tt>foo*</tt>" (note that we do
931<em>not</em> want "<tt>int*</tt>"). In order to provide all of this, Type has
Chris Lattner11406c12007-07-31 16:50:51 +0000[diff] [blame]932a getAsPointerType() method that checks whether the type is structurally a
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]933PointerType and, if so, returns the best one. If not, it returns a null
934pointer.</p>
935
936<p>This structure is somewhat mystical, but after meditating on it, it will
937make sense to you :).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]938
939<!-- ======================================================================= -->
940<h3 id="QualType">The QualType class</h3>
941<!-- ======================================================================= -->
942
943<p>The QualType class is designed as a trivial value class that is small,
944passed by-value and is efficient to query. The idea of QualType is that it
945stores the type qualifiers (const, volatile, restrict) separately from the types
946themselves: QualType is conceptually a pair of "Type*" and bits for the type
947qualifiers.</p>
948
949<p>By storing the type qualifiers as bits in the conceptual pair, it is
950extremely efficient to get the set of qualifiers on a QualType (just return the
951field of the pair), add a type qualifier (which is a trivial constant-time
952operation that sets a bit), and remove one or more type qualifiers (just return
953a QualType with the bitfield set to empty).</p>
954
955<p>Further, because the bits are stored outside of the type itself, we do not
956need to create duplicates of types with different sets of qualifiers (i.e. there
957is only a single heap allocated "int" type: "const int" and "volatile const int"
958both point to the same heap allocated "int" type). This reduces the heap size
959used to represent bits and also means we do not have to consider qualifiers when
960uniquing types (<a href="#Type">Type</a> does not even contain qualifiers).</p>
961
962<p>In practice, on hosts where it is safe, the 3 type qualifiers are stored in
963the low bit of the pointer to the Type object. This means that QualType is
964exactly the same size as a pointer, and this works fine on any system where
965malloc'd objects are at least 8 byte aligned.</p>
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]966
967<!-- ======================================================================= -->
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]968<h3 id="DeclarationName">Declaration names</h3>
969<!-- ======================================================================= -->
970
971<p>The <tt>DeclarationName</tt> class represents the name of a
972 declaration in Clang. Declarations in the C family of languages can
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]973 take several different forms. Most declarations are named by
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]974 simple identifiers, e.g., "<code>f</code>" and "<code>x</code>" in
975 the function declaration <code>f(int x)</code>. In C++, declaration
976 names can also name class constructors ("<code>Class</code>"
977 in <code>struct Class { Class(); }</code>), class destructors
978 ("<code>~Class</code>"), overloaded operator names ("operator+"),
979 and conversion functions ("<code>operator void const *</code>"). In
980 Objective-C, declaration names can refer to the names of Objective-C
981 methods, which involve the method name and the parameters,
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]982 collectively called a <i>selector</i>, e.g.,
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]983 "<code>setWidth:height:</code>". Since all of these kinds of
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]984 entities - variables, functions, Objective-C methods, C++
985 constructors, destructors, and operators - are represented as
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]986 subclasses of Clang's common <code>NamedDecl</code>
987 class, <code>DeclarationName</code> is designed to efficiently
988 represent any kind of name.</p>
989
990<p>Given
991 a <code>DeclarationName</code> <code>N</code>, <code>N.getNameKind()</code>
Douglas Gregor2def4832008-11-17 20:34:05 +0000[diff] [blame]992 will produce a value that describes what kind of name <code>N</code>
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]993 stores. There are 8 options (all of the names are inside
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]994 the <code>DeclarationName</code> class)</p>
995<dl>
996 <dt>Identifier</dt>
997 <dd>The name is a simple
998 identifier. Use <code>N.getAsIdentifierInfo()</code> to retrieve the
999 corresponding <code>IdentifierInfo*</code> pointing to the actual
1000 identifier. Note that C++ overloaded operators (e.g.,
1001 "<code>operator+</code>") are represented as special kinds of
1002 identifiers. Use <code>IdentifierInfo</code>'s <code>getOverloadedOperatorID</code>
1003 function to determine whether an identifier is an overloaded
1004 operator name.</dd>
1005
1006 <dt>ObjCZeroArgSelector, ObjCOneArgSelector,
1007 ObjCMultiArgSelector</dt>
1008 <dd>The name is an Objective-C selector, which can be retrieved as a
1009 <code>Selector</code> instance
1010 via <code>N.getObjCSelector()</code>. The three possible name
1011 kinds for Objective-C reflect an optimization within
1012 the <code>DeclarationName</code> class: both zero- and
1013 one-argument selectors are stored as a
1014 masked <code>IdentifierInfo</code> pointer, and therefore require
1015 very little space, since zero- and one-argument selectors are far
1016 more common than multi-argument selectors (which use a different
1017 structure).</dd>
1018
1019 <dt>CXXConstructorName</dt>
1020 <dd>The name is a C++ constructor
1021 name. Use <code>N.getCXXNameType()</code> to retrieve
1022 the <a href="#QualType">type</a> that this constructor is meant to
1023 construct. The type is always the canonical type, since all
1024 constructors for a given type have the same name.</dd>
1025
1026 <dt>CXXDestructorName</dt>
1027 <dd>The name is a C++ destructor
1028 name. Use <code>N.getCXXNameType()</code> to retrieve
1029 the <a href="#QualType">type</a> whose destructor is being
1030 named. This type is always a canonical type.</dd>
1031
1032 <dt>CXXConversionFunctionName</dt>
1033 <dd>The name is a C++ conversion function. Conversion functions are
1034 named according to the type they convert to, e.g., "<code>operator void
1035 const *</code>". Use <code>N.getCXXNameType()</code> to retrieve
1036 the type that this conversion function converts to. This type is
1037 always a canonical type.</dd>
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1038
1039 <dt>CXXOperatorName</dt>
1040 <dd>The name is a C++ overloaded operator name. Overloaded operators
1041 are named according to their spelling, e.g.,
1042 "<code>operator+</code>" or "<code>operator new
1043 []</code>". Use <code>N.getCXXOverloadedOperator()</code> to
1044 retrieve the overloaded operator (a value of
1045 type <code>OverloadedOperatorKind</code>).</dd>
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1046</dl>
1047
1048<p><code>DeclarationName</code>s are cheap to create, copy, and
1049 compare. They require only a single pointer's worth of storage in
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1050 the common cases (identifiers, zero-
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1051 and one-argument Objective-C selectors) and use dense, uniqued
1052 storage for the other kinds of
1053 names. Two <code>DeclarationName</code>s can be compared for
1054 equality (<code>==</code>, <code>!=</code>) using a simple bitwise
1055 comparison, can be ordered
1056 with <code>&lt;</code>, <code>&gt;</code>, <code>&lt;=</code>,
1057 and <code>&gt;=</code> (which provide a lexicographical ordering for
1058 normal identifiers but an unspecified ordering for other kinds of
1059 names), and can be placed into LLVM <code>DenseMap</code>s
1060 and <code>DenseSet</code>s.</p>
1061
1062<p><code>DeclarationName</code> instances can be created in different
1063 ways depending on what kind of name the instance will store. Normal
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1064 identifiers (<code>IdentifierInfo</code> pointers) and Objective-C selectors
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1065 (<code>Selector</code>) can be implicitly converted
1066 to <code>DeclarationName</code>s. Names for C++ constructors,
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1067 destructors, conversion functions, and overloaded operators can be retrieved from
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1068 the <code>DeclarationNameTable</code>, an instance of which is
1069 available as <code>ASTContext::DeclarationNames</code>. The member
1070 functions <code>getCXXConstructorName</code>, <code>getCXXDestructorName</code>,
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]1071 <code>getCXXConversionFunctionName</code>, and <code>getCXXOperatorName</code>, respectively,
1072 return <code>DeclarationName</code> instances for the four kinds of
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1073 C++ special function names.</p>
1074
1075<!-- ======================================================================= -->
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1076<h3 id="DeclContext">Declaration contexts</h3>
1077<!-- ======================================================================= -->
1078<p>Every declaration in a program exists within some <i>declaration
1079 context</i>, such as a translation unit, namespace, class, or
1080 function. Declaration contexts in Clang are represented by
1081 the <code>DeclContext</code> class, from which the various
1082 declaration-context AST nodes
1083 (<code>TranslationUnitDecl</code>, <code>NamespaceDecl</code>, <code>RecordDecl</code>, <code>FunctionDecl</code>,
1084 etc.) will derive. The <code>DeclContext</code> class provides
1085 several facilities common to each declaration context:</p>
1086<dl>
1087 <dt>Source-centric vs. Semantics-centric View of Declarations</dt>
1088 <dd><code>DeclContext</code> provides two views of the declarations
1089 stored within a declaration context. The source-centric view
1090 accurately represents the program source code as written, including
1091 multiple declarations of entities where present (see the
1092 section <a href="#Redeclarations">Redeclarations and
1093 Overloads</a>), while the semantics-centric view represents the
1094 program semantics. The two views are kept synchronized by semantic
1095 analysis while the ASTs are being constructed.</dd>
1096
1097 <dt>Storage of declarations within that context</dt>
1098 <dd>Every declaration context can contain some number of
1099 declarations. For example, a C++ class (represented
1100 by <code>RecordDecl</code>) contains various member functions,
1101 fields, nested types, and so on. All of these declarations will be
1102 stored within the <code>DeclContext</code>, and one can iterate
1103 over the declarations via
1104 [<code>DeclContext::decls_begin()</code>,
1105 <code>DeclContext::decls_end()</code>). This mechanism provides
1106 the source-centric view of declarations in the context.</dd>
1107
1108 <dt>Lookup of declarations within that context</dt>
1109 <dd>The <code>DeclContext</code> structure provides efficient name
1110 lookup for names within that declaration context. For example,
1111 if <code>N</code> is a namespace we can look for the
1112 name <code>N::f</code>
1113 using <code>DeclContext::lookup</code>. The lookup itself is
1114 based on a lazily-constructed array (for declaration contexts
1115 with a small number of declarations) or hash table (for
1116 declaration contexts with more declarations). The lookup
1117 operation provides the semantics-centric view of the declarations
1118 in the context.</dd>
1119
1120 <dt>Ownership of declarations</dt>
1121 <dd>The <code>DeclContext</code> owns all of the declarations that
1122 were declared within its declaration context, and is responsible
1123 for the management of their memory as well as their
1124 (de-)serialization.</dd>
1125</dl>
1126
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1127<p>All declarations are stored within a declaration context, and one
1128 can query
1129 information about the context in which each declaration lives. One
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1130 can retrieve the <code>DeclContext</code> that contains a
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1131 particular <code>Decl</code>
1132 using <code>Decl::getDeclContext</code>. However, see the
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1133 section <a href="#LexicalAndSemanticContexts">Lexical and Semantic
1134 Contexts</a> for more information about how to interpret this
1135 context information.</p>
1136
1137<h4 id="Redeclarations">Redeclarations and Overloads</h4>
1138<p>Within a translation unit, it is common for an entity to be
1139declared several times. For example, we might declare a function "f"
1140 and then later re-declare it as part of an inlined definition:</p>
1141
1142<pre>
1143void f(int x, int y, int z = 1);
1144
1145inline void f(int x, int y, int z) { /* ... */ }
1146</pre>
1147
1148<p>The representation of "f" differs in the source-centric and
1149 semantics-centric views of a declaration context. In the
1150 source-centric view, all redeclarations will be present, in the
1151 order they occurred in the source code, making
1152 this view suitable for clients that wish to see the structure of
1153 the source code. In the semantics-centric view, only the most recent "f"
1154 will be found by the lookup, since it effectively replaces the first
1155 declaration of "f".</p>
1156
1157<p>In the semantics-centric view, overloading of functions is
1158 represented explicitly. For example, given two declarations of a
1159 function "g" that are overloaded, e.g.,</p>
1160<pre>
1161void g();
1162void g(int);
1163</pre>
1164<p>the <code>DeclContext::lookup</code> operation will return
1165 an <code>OverloadedFunctionDecl</code> that contains both
1166 declarations of "g". Clients that perform semantic analysis on a
1167 program that is not concerned with the actual source code will
1168 primarily use this semantics-centric view.</p>
1169
1170<h4 id="LexicalAndSemanticContexts">Lexical and Semantic Contexts</h4>
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1171<p>Each declaration has two potentially different
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1172 declaration contexts: a <i>lexical</i> context, which corresponds to
1173 the source-centric view of the declaration context, and
1174 a <i>semantic</i> context, which corresponds to the
1175 semantics-centric view. The lexical context is accessible
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1176 via <code>Decl::getLexicalDeclContext</code> while the
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1177 semantic context is accessible
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1178 via <code>Decl::getDeclContext</code>, both of which return
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1179 <code>DeclContext</code> pointers. For most declarations, the two
1180 contexts are identical. For example:</p>
1181
1182<pre>
1183class X {
1184public:
1185 void f(int x);
1186};
1187</pre>
1188
1189<p>Here, the semantic and lexical contexts of <code>X::f</code> are
1190 the <code>DeclContext</code> associated with the
1191 class <code>X</code> (itself stored as a <code>RecordDecl</code> AST
1192 node). However, we can now define <code>X::f</code> out-of-line:</p>
1193
1194<pre>
1195void X::f(int x = 17) { /* ... */ }
1196</pre>
1197
1198<p>This definition of has different lexical and semantic
1199 contexts. The lexical context corresponds to the declaration
1200 context in which the actual declaration occurred in the source
1201 code, e.g., the translation unit containing <code>X</code>. Thus,
1202 this declaration of <code>X::f</code> can be found by traversing
1203 the declarations provided by
1204 [<code>decls_begin()</code>, <code>decls_end()</code>) in the
1205 translation unit.</p>
1206
1207<p>The semantic context of <code>X::f</code> corresponds to the
1208 class <code>X</code>, since this member function is (semantically) a
1209 member of <code>X</code>. Lookup of the name <code>f</code> into
1210 the <code>DeclContext</code> associated with <code>X</code> will
1211 then return the definition of <code>X::f</code> (including
1212 information about the default argument).</p>
1213
1214<h4 id="TransparentContexts">Transparent Declaration Contexts</h4>
1215<p>In C and C++, there are several contexts in which names that are
1216 logically declared inside another declaration will actually "leak"
1217 out into the enclosing scope from the perspective of name
1218 lookup. The most obvious instance of this behavior is in
1219 enumeration types, e.g.,</p>
1220<pre>
1221enum Color {
1222 Red,
1223 Green,
1224 Blue
1225};
1226</pre>
1227
1228<p>Here, <code>Color</code> is an enumeration, which is a declaration
1229 context that contains the
1230 enumerators <code>Red</code>, <code>Green</code>,
1231 and <code>Blue</code>. Thus, traversing the list of declarations
1232 contained in the enumeration <code>Color</code> will
1233 yield <code>Red</code>, <code>Green</code>,
1234 and <code>Blue</code>. However, outside of the scope
1235 of <code>Color</code> one can name the enumerator <code>Red</code>
1236 without qualifying the name, e.g.,</p>
1237
1238<pre>
1239Color c = Red;
1240</pre>
1241
1242<p>There are other entities in C++ that provide similar behavior. For
1243 example, linkage specifications that use curly braces:</p>
1244
1245<pre>
1246extern "C" {
1247 void f(int);
1248 void g(int);
1249}
1250// f and g are visible here
1251</pre>
1252
1253<p>For source-level accuracy, we treat the linkage specification and
1254 enumeration type as a
1255 declaration context in which its enclosed declarations ("Red",
1256 "Green", and "Blue"; "f" and "g")
1257 are declared. However, these declarations are visible outside of the
1258 scope of the declaration context.</p>
1259
1260<p>These language features (and several others, described below) have
1261 roughly the same set of
1262 requirements: declarations are declared within a particular lexical
1263 context, but the declarations are also found via name lookup in
1264 scopes enclosing the declaration itself. This feature is implemented
1265 via <i>transparent</i> declaration contexts
1266 (see <code>DeclContext::isTransparentContext()</code>), whose
1267 declarations are visible in the nearest enclosing non-transparent
1268 declaration context. This means that the lexical context of the
1269 declaration (e.g., an enumerator) will be the
1270 transparent <code>DeclContext</code> itself, as will the semantic
1271 context, but the declaration will be visible in every outer context
1272 up to and including the first non-transparent declaration context (since
1273 transparent declaration contexts can be nested).</p>
1274
1275<p>The transparent <code>DeclContexts</code> are:</p>
1276<ul>
1277 <li>Enumerations (but not C++0x "scoped enumerations"):
1278 <pre>
1279enum Color {
1280 Red,
1281 Green,
1282 Blue
1283};
1284// Red, Green, and Blue are in scope
1285 </pre></li>
1286 <li>C++ linkage specifications:
1287 <pre>
1288extern "C" {
1289 void f(int);
1290 void g(int);
1291}
1292// f and g are in scope
1293 </pre></li>
1294 <li>Anonymous unions and structs:
1295 <pre>
1296struct LookupTable {
1297 bool IsVector;
1298 union {
1299 std::vector&lt;Item&gt; *Vector;
1300 std::set&lt;Item&gt; *Set;
1301 };
1302};
1303
1304LookupTable LT;
1305LT.Vector = 0; // Okay: finds Vector inside the unnamed union
1306 </pre>
1307 </li>
1308 <li>C++0x inline namespaces:
1309<pre>
1310namespace mylib {
1311 inline namespace debug {
1312 class X;
1313 }
1314}
1315mylib::X *xp; // okay: mylib::X refers to mylib::debug::X
1316</pre>
1317</li>
1318</ul>
1319
1320
1321<h4 id="MultiDeclContext">Multiply-Defined Declaration Contexts</h4>
1322<p>C++ namespaces have the interesting--and, so far, unique--property that
1323the namespace can be defined multiple times, and the declarations
1324provided by each namespace definition are effectively merged (from
1325the semantic point of view). For example, the following two code
1326snippets are semantically indistinguishable:</p>
1327<pre>
1328// Snippet #1:
1329namespace N {
1330 void f();
1331}
1332namespace N {
1333 void f(int);
1334}
1335
1336// Snippet #2:
1337namespace N {
1338 void f();
1339 void f(int);
1340}
1341</pre>
1342
1343<p>In Clang's representation, the source-centric view of declaration
1344 contexts will actually have two separate <code>NamespaceDecl</code>
1345 nodes in Snippet #1, each of which is a declaration context that
1346 contains a single declaration of "f". However, the semantics-centric
1347 view provided by name lookup into the namespace <code>N</code> for
1348 "f" will return an <code>OverloadedFunctionDecl</code> that contains
1349 both declarations of "f".</p>
1350
1351<p><code>DeclContext</code> manages multiply-defined declaration
1352 contexts internally. The
1353 function <code>DeclContext::getPrimaryContext</code> retrieves the
1354 "primary" context for a given <code>DeclContext</code> instance,
1355 which is the <code>DeclContext</code> responsible for maintaining
1356 the lookup table used for the semantics-centric view. Given the
1357 primary context, one can follow the chain
1358 of <code>DeclContext</code> nodes that define additional
1359 declarations via <code>DeclContext::getNextContext</code>. Note that
1360 these functions are used internally within the lookup and insertion
1361 methods of the <code>DeclContext</code>, so the vast majority of
1362 clients can ignore them.</p>
1363
1364<!-- ======================================================================= -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1365<h3 id="CFG">The <tt>CFG</tt> class</h3>
1366<!-- ======================================================================= -->
1367
1368<p>The <tt>CFG</tt> class is designed to represent a source-level
1369control-flow graph for a single statement (<tt>Stmt*</tt>). Typically
1370instances of <tt>CFG</tt> are constructed for function bodies (usually
1371an instance of <tt>CompoundStmt</tt>), but can also be instantiated to
1372represent the control-flow of any class that subclasses <tt>Stmt</tt>,
1373which includes simple expressions. Control-flow graphs are especially
1374useful for performing
1375<a href="http://en.wikipedia.org/wiki/Data_flow_analysis#Sensitivities">flow-
1376or path-sensitive</a> program analyses on a given function.</p>
1377
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1378<!-- ============ -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1379<h4>Basic Blocks</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1380<!-- ============ -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1381
1382<p>Concretely, an instance of <tt>CFG</tt> is a collection of basic
1383blocks. Each basic block is an instance of <tt>CFGBlock</tt>, which
1384simply contains an ordered sequence of <tt>Stmt*</tt> (each referring
1385to statements in the AST). The ordering of statements within a block
1386indicates unconditional flow of control from one statement to the
1387next. <a href="#ConditionalControlFlow">Conditional control-flow</a>
1388is represented using edges between basic blocks. The statements
1389within a given <tt>CFGBlock</tt> can be traversed using
1390the <tt>CFGBlock::*iterator</tt> interface.</p>
1391
1392<p>
Ted Kremenek18e17e72007-10-18 22:50:52 +0000[diff] [blame]1393A <tt>CFG</tt> object owns the instances of <tt>CFGBlock</tt> within
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1394the control-flow graph it represents. Each <tt>CFGBlock</tt> within a
1395CFG is also uniquely numbered (accessible
1396via <tt>CFGBlock::getBlockID()</tt>). Currently the number is
1397based on the ordering the blocks were created, but no assumptions
1398should be made on how <tt>CFGBlock</tt>s are numbered other than their
1399numbers are unique and that they are numbered from 0..N-1 (where N is
1400the number of basic blocks in the CFG).</p>
1401
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1402<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1403<h4>Entry and Exit Blocks</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1404<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1405
1406Each instance of <tt>CFG</tt> contains two special blocks:
1407an <i>entry</i> block (accessible via <tt>CFG::getEntry()</tt>), which
1408has no incoming edges, and an <i>exit</i> block (accessible
1409via <tt>CFG::getExit()</tt>), which has no outgoing edges. Neither
1410block contains any statements, and they serve the role of providing a
1411clear entrance and exit for a body of code such as a function body.
1412The presence of these empty blocks greatly simplifies the
1413implementation of many analyses built on top of CFGs.
1414
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1415<!-- ===================================================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1416<h4 id ="ConditionalControlFlow">Conditional Control-Flow</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1417<!-- ===================================================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1418
1419<p>Conditional control-flow (such as those induced by if-statements
1420and loops) is represented as edges between <tt>CFGBlock</tt>s.
1421Because different C language constructs can induce control-flow,
1422each <tt>CFGBlock</tt> also records an extra <tt>Stmt*</tt> that
1423represents the <i>terminator</i> of the block. A terminator is simply
1424the statement that caused the control-flow, and is used to identify
1425the nature of the conditional control-flow between blocks. For
1426example, in the case of an if-statement, the terminator refers to
1427the <tt>IfStmt</tt> object in the AST that represented the given
1428branch.</p>
1429
1430<p>To illustrate, consider the following code example:</p>
1431
1432<code>
1433int foo(int x) {<br>
1434&nbsp;&nbsp;x = x + 1;<br>
1435<br>
1436&nbsp;&nbsp;if (x > 2) x++;<br>
1437&nbsp;&nbsp;else {<br>
1438&nbsp;&nbsp;&nbsp;&nbsp;x += 2;<br>
1439&nbsp;&nbsp;&nbsp;&nbsp;x *= 2;<br>
1440&nbsp;&nbsp;}<br>
1441<br>
1442&nbsp;&nbsp;return x;<br>
1443}
1444</code>
1445
1446<p>After invoking the parser+semantic analyzer on this code fragment,
1447the AST of the body of <tt>foo</tt> is referenced by a
1448single <tt>Stmt*</tt>. We can then construct an instance
1449of <tt>CFG</tt> representing the control-flow graph of this function
1450body by single call to a static class method:</p>
1451
1452<code>
1453&nbsp;&nbsp;Stmt* FooBody = ...<br>
1454&nbsp;&nbsp;CFG* FooCFG = <b>CFG::buildCFG</b>(FooBody);
1455</code>
1456
1457<p>It is the responsibility of the caller of <tt>CFG::buildCFG</tt>
1458to <tt>delete</tt> the returned <tt>CFG*</tt> when the CFG is no
1459longer needed.</p>
1460
1461<p>Along with providing an interface to iterate over
1462its <tt>CFGBlock</tt>s, the <tt>CFG</tt> class also provides methods
1463that are useful for debugging and visualizing CFGs. For example, the
1464method
1465<tt>CFG::dump()</tt> dumps a pretty-printed version of the CFG to
1466standard error. This is especially useful when one is using a
1467debugger such as gdb. For example, here is the output
1468of <tt>FooCFG->dump()</tt>:</p>
1469
1470<code>
1471&nbsp;[ B5 (ENTRY) ]<br>
1472&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (0):<br>
1473&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B4<br>
1474<br>
1475&nbsp;[ B4 ]<br>
1476&nbsp;&nbsp;&nbsp;&nbsp;1: x = x + 1<br>
1477&nbsp;&nbsp;&nbsp;&nbsp;2: (x > 2)<br>
1478&nbsp;&nbsp;&nbsp;&nbsp;<b>T: if [B4.2]</b><br>
1479&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B5<br>
1480&nbsp;&nbsp;&nbsp;&nbsp;Successors (2): B3 B2<br>
1481<br>
1482&nbsp;[ B3 ]<br>
1483&nbsp;&nbsp;&nbsp;&nbsp;1: x++<br>
1484&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B4<br>
1485&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B1<br>
1486<br>
1487&nbsp;[ B2 ]<br>
1488&nbsp;&nbsp;&nbsp;&nbsp;1: x += 2<br>
1489&nbsp;&nbsp;&nbsp;&nbsp;2: x *= 2<br>
1490&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B4<br>
1491&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B1<br>
1492<br>
1493&nbsp;[ B1 ]<br>
1494&nbsp;&nbsp;&nbsp;&nbsp;1: return x;<br>
1495&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (2): B2 B3<br>
1496&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B0<br>
1497<br>
1498&nbsp;[ B0 (EXIT) ]<br>
1499&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B1<br>
1500&nbsp;&nbsp;&nbsp;&nbsp;Successors (0):
1501</code>
1502
1503<p>For each block, the pretty-printed output displays for each block
1504the number of <i>predecessor</i> blocks (blocks that have outgoing
1505control-flow to the given block) and <i>successor</i> blocks (blocks
1506that have control-flow that have incoming control-flow from the given
1507block). We can also clearly see the special entry and exit blocks at
1508the beginning and end of the pretty-printed output. For the entry
1509block (block B5), the number of predecessor blocks is 0, while for the
1510exit block (block B0) the number of successor blocks is 0.</p>
1511
1512<p>The most interesting block here is B4, whose outgoing control-flow
1513represents the branching caused by the sole if-statement
1514in <tt>foo</tt>. Of particular interest is the second statement in
1515the block, <b><tt>(x > 2)</tt></b>, and the terminator, printed
1516as <b><tt>if [B4.2]</tt></b>. The second statement represents the
1517evaluation of the condition of the if-statement, which occurs before
1518the actual branching of control-flow. Within the <tt>CFGBlock</tt>
1519for B4, the <tt>Stmt*</tt> for the second statement refers to the
1520actual expression in the AST for <b><tt>(x > 2)</tt></b>. Thus
1521pointers to subclasses of <tt>Expr</tt> can appear in the list of
1522statements in a block, and not just subclasses of <tt>Stmt</tt> that
1523refer to proper C statements.</p>
1524
1525<p>The terminator of block B4 is a pointer to the <tt>IfStmt</tt>
1526object in the AST. The pretty-printer outputs <b><tt>if
1527[B4.2]</tt></b> because the condition expression of the if-statement
1528has an actual place in the basic block, and thus the terminator is
1529essentially
1530<i>referring</i> to the expression that is the second statement of
1531block B4 (i.e., B4.2). In this manner, conditions for control-flow
1532(which also includes conditions for loops and switch statements) are
1533hoisted into the actual basic block.</p>
1534
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1535<!-- ===================== -->
1536<!-- <h4>Implicit Control-Flow</h4> -->
1537<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1538
1539<!--
1540<p>A key design principle of the <tt>CFG</tt> class was to not require
1541any transformations to the AST in order to represent control-flow.
1542Thus the <tt>CFG</tt> does not perform any "lowering" of the
1543statements in an AST: loops are not transformed into guarded gotos,
1544short-circuit operations are not converted to a set of if-statements,
1545and so on.</p>
1546-->
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]1547
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1548
1549<!-- ======================================================================= -->
1550<h3 id="Constants">Constant Folding in the Clang AST</h3>
1551<!-- ======================================================================= -->
1552
1553<p>There are several places where constants and constant folding matter a lot to
1554the Clang front-end. First, in general, we prefer the AST to retain the source
1555code as close to how the user wrote it as possible. This means that if they
1556wrote "5+4", we want to keep the addition and two constants in the AST, we don't
1557want to fold to "9". This means that constant folding in various ways turns
1558into a tree walk that needs to handle the various cases.</p>
1559
1560<p>However, there are places in both C and C++ that require constants to be
1561folded. For example, the C standard defines what an "integer constant
1562expression" (i-c-e) is with very precise and specific requirements. The
1563language then requires i-c-e's in a lot of places (for example, the size of a
1564bitfield, the value for a case statement, etc). For these, we have to be able
1565to constant fold the constants, to do semantic checks (e.g. verify bitfield size
1566is non-negative and that case statements aren't duplicated). We aim for Clang
1567to be very pedantic about this, diagnosing cases when the code does not use an
1568i-c-e where one is required, but accepting the code unless running with
1569<tt>-pedantic-errors</tt>.</p>
1570
1571<p>Things get a little bit more tricky when it comes to compatibility with
1572real-world source code. Specifically, GCC has historically accepted a huge
1573superset of expressions as i-c-e's, and a lot of real world code depends on this
1574unfortuate accident of history (including, e.g., the glibc system headers). GCC
1575accepts anything its "fold" optimizer is capable of reducing to an integer
1576constant, which means that the definition of what it accepts changes as its
1577optimizer does. One example is that GCC accepts things like "case X-X:" even
1578when X is a variable, because it can fold this to 0.</p>
1579
1580<p>Another issue are how constants interact with the extensions we support, such
1581as __builtin_constant_p, __builtin_inf, __extension__ and many others. C99
1582obviously does not specify the semantics of any of these extensions, and the
1583definition of i-c-e does not include them. However, these extensions are often
1584used in real code, and we have to have a way to reason about them.</p>
1585
1586<p>Finally, this is not just a problem for semantic analysis. The code
1587generator and other clients have to be able to fold constants (e.g. to
1588initialize global variables) and has to handle a superset of what C99 allows.
1589Further, these clients can benefit from extended information. For example, we
1590know that "foo()||1" always evaluates to true, but we can't replace the
1591expression with true because it has side effects.</p>
1592
1593<!-- ======================= -->
1594<h4>Implementation Approach</h4>
1595<!-- ======================= -->
1596
1597<p>After trying several different approaches, we've finally converged on a
1598design (Note, at the time of this writing, not all of this has been implemented,
1599consider this a design goal!). Our basic approach is to define a single
1600recursive method evaluation method (<tt>Expr::Evaluate</tt>), which is
1601implemented in <tt>AST/ExprConstant.cpp</tt>. Given an expression with 'scalar'
1602type (integer, fp, complex, or pointer) this method returns the following
1603information:</p>
1604
1605<ul>
1606<li>Whether the expression is an integer constant expression, a general
1607 constant that was folded but has no side effects, a general constant that
1608 was folded but that does have side effects, or an uncomputable/unfoldable
1609 value.
1610</li>
1611<li>If the expression was computable in any way, this method returns the APValue
1612 for the result of the expression.</li>
1613<li>If the expression is not evaluatable at all, this method returns
1614 information on one of the problems with the expression. This includes a
1615 SourceLocation for where the problem is, and a diagnostic ID that explains
1616 the problem. The diagnostic should be have ERROR type.</li>
1617<li>If the expression is not an integer constant expression, this method returns
1618 information on one of the problems with the expression. This includes a
1619 SourceLocation for where the problem is, and a diagnostic ID that explains
1620 the problem. The diagnostic should be have EXTENSION type.</li>
1621</ul>
1622
1623<p>This information gives various clients the flexibility that they want, and we
1624will eventually have some helper methods for various extensions. For example,
1625Sema should have a <tt>Sema::VerifyIntegerConstantExpression</tt> method, which
1626calls Evaluate on the expression. If the expression is not foldable, the error
1627is emitted, and it would return true. If the expression is not an i-c-e, the
1628EXTENSION diagnostic is emitted. Finally it would return false to indicate that
1629the AST is ok.</p>
1630
1631<p>Other clients can use the information in other ways, for example, codegen can
1632just use expressions that are foldable in any way.</p>
1633
1634<!-- ========== -->
1635<h4>Extensions</h4>
1636<!-- ========== -->
1637
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]1638<p>This section describes how some of the various extensions Clang supports
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1639interacts with constant evaluation:</p>
1640
1641<ul>
1642<li><b><tt>__extension__</tt></b>: The expression form of this extension causes
1643 any evaluatable subexpression to be accepted as an integer constant
1644 expression.</li>
1645<li><b><tt>__builtin_constant_p</tt></b>: This returns true (as a integer
Chris Lattner28daa532008-12-12 06:55:44 +0000[diff] [blame]1646 constant expression) if the operand is any evaluatable constant. As a
1647 special case, if <tt>__builtin_constant_p</tt> is the (potentially
1648 parenthesized) condition of a conditional operator expression ("?:"), only
Chris Lattner42b83dd2008-12-12 18:00:51 +0000[diff] [blame]1649 the true side of the conditional operator is considered, and it is evaluated
1650 with full constant folding.</li>
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1651<li><b><tt>__builtin_choose_expr</tt></b>: The condition is required to be an
1652 integer constant expression, but we accept any constant as an "extension of
1653 an extension". This only evaluates one operand depending on which way the
1654 condition evaluates.</li>
1655<li><b><tt>__builtin_classify_type</tt></b>: This always returns an integer
1656 constant expression.</li>
1657<li><b><tt>__builtin_inf,nan,..</tt></b>: These are treated just like a
1658 floating-point literal.</li>
1659<li><b><tt>__builtin_abs,copysign,..</tt></b>: These are constant folded as
1660 general constant expressions.</li>
1661</ul>
1662
1663
1664
1665
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]1666</div>
1667</body>
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1668</html>