1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" 2 "http://www.w3.org/TR/html4/strict.dtd"> 3<html> 4<head> 5 <meta http-equiv="content-type" content="text/html; charset=iso-8859-1"> 6 <title>Clang - Expressive Diagnostics</title> 7 <link type="text/css" rel="stylesheet" href="menu.css"> 8 <link type="text/css" rel="stylesheet" href="content.css"> 9 <style type="text/css"> 10 .loc { font-weight: bold; } 11 .err { color:red; font-weight: bold; } 12 .warn { color:magenta; font-weight: bold; } 13 .note { color:gray; font-weight: bold; } 14 .msg { font-weight: bold; } 15 .cmd { font-style: italic; } 16 .snip { } 17 .point { color:green; font-weight: bold; } 18 </style> 19</head> 20<body> 21 22<!--#include virtual="menu.html.incl"--> 23 24<div id="content"> 25 26 27<!--=======================================================================--> 28<h1>Expressive Diagnostics</h1> 29<!--=======================================================================--> 30 31<p>In addition to being fast and functional, we aim to make Clang extremely user 32friendly. As far as a command-line compiler goes, this basically boils down to 33making the diagnostics (error and warning messages) generated by the compiler 34be as useful as possible. There are several ways that we do this. This section 35talks about the experience provided by the command line compiler, contrasting 36Clang output to GCC 4.9's output in some cases. 37</p> 38 39<h2>Column Numbers and Caret Diagnostics</h2> 40 41<p>First, all diagnostics produced by clang include full column number 42information. The clang command-line compiler driver uses this information 43to print "point diagnostics". 44(IDEs can use the information to display in-line error markup.) 45This is nice because it makes it very easy to understand exactly 46what is wrong in a particular piece of code.</p> 47 48<p>The point (the green "^" character) exactly shows where the problem is, even 49inside of a string. This makes it really easy to jump to the problem and 50helps when multiple instances of the same character occur on a line. (We'll 51revisit this more in following examples.)</p> 52 53<pre> 54 $ <span class="cmd">clang -fsyntax-only format-strings.c</span> 55 <span class="loc">format-strings.c:91:13:</span> <span class="warn">warning:</span> <span class="msg">'.*' specified field precision is missing a matching 'int' argument</span> 56 <span class="snip" > printf("%.*d");</span> 57 <span class="point"> ^</span> 58</pre> 59 60<p>Note that modern versions of GCC have followed Clang's lead, and are 61now able to give a column for a diagnostic, and include a snippet of source 62text in the result. However, Clang's column number is much more accurate, 63pointing at the problematic format specifier, rather than the <tt>)</tt> 64character the parser had reached when the problem was detected. 65Also, Clang's diagnostic is colored by default, making it easier to 66distinguish from nearby text.</p> 67 68<h2>Range Highlighting for Related Text</h2> 69 70<p>Clang captures and accurately tracks range information for expressions, 71statements, and other constructs in your program and uses this to make 72diagnostics highlight related information. In the following somewhat 73nonsensical example you can see that you don't even need to see the original source code to 74understand what is wrong based on the Clang error. Because clang prints a 75point, you know exactly <em>which</em> plus it is complaining about. The range 76information highlights the left and right side of the plus which makes it 77immediately obvious what the compiler is talking about. 78Range information is very useful for 79cases involving precedence issues and many other cases.</p> 80 81<pre> 82 $ <span class="cmd">gcc-4.9 -fsyntax-only t.c</span> 83 t.c: In function 'int f(int, int)': 84 t.c:7:39: error: invalid operands to binary + (have 'int' and 'struct A') 85 return y + func(y ? ((SomeA.X + 40) + SomeA) / 42 + SomeA.X : SomeA.X); 86 ^ 87 $ <span class="cmd">clang -fsyntax-only t.c</span> 88 <span class="loc">t.c:7:39:</span> <span class="err">error:</span> <span class="msg">invalid operands to binary expression ('int' and 'struct A')</span> 89 <span class="snip" > return y + func(y ? ((SomeA.X + 40) + SomeA) / 42 + SomeA.X : SomeA.X);</span> 90 <span class="point"> ~~~~~~~~~~~~~~ ^ ~~~~~</span> 91</pre> 92 93<h2>Precision in Wording</h2> 94 95<p>A detail is that we have tried really hard to make the diagnostics that come 96out of clang contain exactly the pertinent information about what is wrong and 97why. In the example above, we tell you what the inferred types are for 98the left and right hand sides, and we don't repeat what is obvious from the 99point (e.g., that this is a "binary +").</p> 100 101<p>Many other examples abound. In the following example, not only do we tell you 102that there is a problem with the <tt>*</tt> 103and point to it, we say exactly why and tell you what the type is (in case it is 104a complicated subexpression, such as a call to an overloaded function). This 105sort of attention to detail makes it much easier to understand and fix problems 106quickly.</p> 107 108<pre> 109 $ <span class="cmd">gcc-4.9 -fsyntax-only t.c</span> 110 t.c:5:11: error: invalid type argument of unary '*' (have 'int') 111 return *SomeA.X; 112 ^ 113 $ <span class="cmd">clang -fsyntax-only t.c</span> 114 <span class="loc">t.c:5:11:</span> <span class="err">error:</span> <span class="msg">indirection requires pointer operand ('int' invalid)</span> 115 <span class="snip" > int y = *SomeA.X;</span> 116 <span class="point"> ^~~~~~~~</span> 117</pre> 118 119<h2>Typedef Preservation and Selective Unwrapping</h2> 120 121<p>Many programmers use high-level user defined types, typedefs, and other 122syntactic sugar to refer to types in their program. This is useful because they 123can abbreviate otherwise very long types and it is useful to preserve the 124typename in diagnostics. However, sometimes very simple typedefs can wrap 125trivial types and it is important to strip off the typedef to understand what 126is going on. Clang aims to handle both cases well.<p> 127 128<p>The following example shows where it is important to preserve 129a typedef in C.</p> 130 131<pre> 132 $ <span class="cmd">clang -fsyntax-only t.c</span> 133 <span class="loc">t.c:15:11:</span> <span class="err">error:</span> <span class="msg">can't convert between vector values of different size ('__m128' and 'int const *')</span> 134 <span class="snip"> myvec[1]/P;</span> 135 <span class="point"> ~~~~~~~~^~</span> 136</pre> 137 138<p>The following example shows where it is useful for the compiler to expose 139underlying details of a typedef. If the user was somehow confused about how the 140system "pid_t" typedef is defined, Clang helpfully displays it with "aka".</p> 141 142<pre> 143 $ <span class="cmd">clang -fsyntax-only t.c</span> 144 <span class="loc">t.c:13:9:</span> <span class="err">error:</span> <span class="msg">member reference base type 'pid_t' (aka 'int') is not a structure or union</span> 145 <span class="snip"> myvar = myvar.x;</span> 146 <span class="point"> ~~~~~ ^</span> 147</pre> 148 149<p>In C++, type preservation includes retaining any qualification written into type names. For example, if we take a small snippet of code such as: 150 151<blockquote> 152<pre> 153namespace services { 154 struct WebService { }; 155} 156namespace myapp { 157 namespace servers { 158 struct Server { }; 159 } 160} 161 162using namespace myapp; 163void addHTTPService(servers::Server const &server, ::services::WebService const *http) { 164 server += http; 165} 166</pre> 167</blockquote> 168 169<p>and then compile it, we see that Clang is both providing accurate information and is retaining the types as written by the user (e.g., "servers::Server", "::services::WebService"): 170 171<pre> 172 $ <span class="cmd">clang -fsyntax-only t.cpp</span> 173 <span class="loc">t.cpp:9:10:</span> <span class="err">error:</span> <span class="msg">invalid operands to binary expression ('servers::Server const' and '::services::WebService const *')</span> 174 <span class="snip">server += http;</span> 175 <span class="point">~~~~~~ ^ ~~~~</span> 176</pre> 177 178<p>Naturally, type preservation extends to uses of templates, and Clang retains information about how a particular template specialization (like <code>std::vector<Real></code>) was spelled within the source code. For example:</p> 179 180<pre> 181 $ <span class="cmd">clang -fsyntax-only t.cpp</span> 182 <span class="loc">t.cpp:12:7:</span> <span class="err">error:</span> <span class="msg">incompatible type assigning 'vector<Real>', expected 'std::string' (aka 'class std::basic_string<char>')</span> 183 <span class="snip">str = vec</span>; 184 <span class="point">^ ~~~</span> 185</pre> 186 187<h2>Fix-it Hints</h2> 188 189<p>"Fix-it" hints provide advice for fixing small, localized problems 190in source code. When Clang produces a diagnostic about a particular 191problem that it can work around (e.g., non-standard or redundant 192syntax, missing keywords, common mistakes, etc.), it may also provide 193specific guidance in the form of a code transformation to correct the 194problem. In the following example, Clang warns about the use of a GCC 195extension that has been considered obsolete since 1993. The underlined 196code should be removed, then replaced with the code below the 197point line (".x =" or ".y =", respectively).</p> 198 199<pre> 200 $ <span class="cmd">clang t.c</span> 201 <span class="loc">t.c:5:28:</span> <span class="warn">warning:</span> <span class="msg">use of GNU old-style field designator extension</span> 202 <span class="snip">struct point origin = { x: 0.0, y: 0.0 };</span> 203 <span class="err">~~</span> <span class="msg"><span class="point">^</span></span> 204 <span class="snip">.x = </span> 205 <span class="loc">t.c:5:36:</span> <span class="warn">warning:</span> <span class="msg">use of GNU old-style field designator extension</span> 206 <span class="snip">struct point origin = { x: 0.0, y: 0.0 };</span> 207 <span class="err">~~</span> <span class="msg"><span class="point">^</span></span> 208 <span class="snip">.y = </span> 209</pre> 210 211<p>"Fix-it" hints are most useful for 212working around common user errors and misconceptions. For example, C++ users 213commonly forget the syntax for explicit specialization of class templates, 214as in the error in the following example. Again, after describing the problem, 215Clang provides the fix--add <code>template<></code>--as part of the 216diagnostic.<p> 217 218<pre> 219 $ <span class="cmd">clang t.cpp</span> 220 <span class="loc">t.cpp:9:3:</span> <span class="err">error:</span> <span class="msg">template specialization requires 'template<>'</span> 221 struct iterator_traits<file_iterator> { 222 <span class="point">^</span> 223 <span class="snip">template<> </span> 224</pre> 225 226<h2>Template Type Diffing</h2> 227 228<p>Templates types can be long and difficult to read. More so when part of an 229error message. Instead of just printing out the type name, Clang has enough 230information to remove the common elements and highlight the differences. To 231show the template structure more clearly, the templated type can also be 232printed as an indented text tree.</p> 233 234Default: template diff with type elision 235<pre> 236<span class="loc">t.cc:4:5:</span> <span class="note">note:</span> candidate function not viable: no known conversion from 'vector<map<[...], <span class="template-highlight">float</span>>>' to 'vector<map<[...], <span class="template-highlight">double</span>>>' for 1st argument; 237</pre> 238-fno-elide-type: template diff without elision 239<pre> 240<span class="loc">t.cc:4:5:</span> <span class="note">note:</span> candidate function not viable: no known conversion from 'vector<map<int, <span class="template-highlight">float</span>>>' to 'vector<map<int, <span class="template-highlight">double</span>>>' for 1st argument; 241</pre> 242-fdiagnostics-show-template-tree: template tree printing with elision 243<pre> 244<span class="loc">t.cc:4:5:</span> <span class="note">note:</span> candidate function not viable: no known conversion for 1st argument; 245 vector< 246 map< 247 [...], 248 [<span class="template-highlight">float</span> != <span class="template-highlight">double</span>]>> 249</pre> 250-fdiagnostics-show-template-tree -fno-elide-type: template tree printing with no elision 251<pre> 252<span class="loc">t.cc:4:5:</span> <span class="note">note:</span> candidate function not viable: no known conversion for 1st argument; 253 vector< 254 map< 255 int, 256 [<span class="template-highlight">float</span> != <span class="template-highlight">double</span>]>> 257</pre> 258 259<h2>Automatic Macro Expansion</h2> 260 261<p>Many errors happen in macros that are sometimes deeply nested. With 262traditional compilers, you need to dig deep into the definition of the macro to 263understand how you got into trouble. The following simple example shows how 264Clang helps you out by automatically printing instantiation information and 265nested range information for diagnostics as they are instantiated through macros 266and also shows how some of the other pieces work in a bigger example.</p> 267 268<pre> 269 $ <span class="cmd">clang -fsyntax-only t.c</span> 270 <span class="loc">t.c:80:3:</span> <span class="err">error:</span> <span class="msg">invalid operands to binary expression ('typeof(P)' (aka 'struct mystruct') and 'typeof(F)' (aka 'float'))</span> 271 <span class="snip"> X = MYMAX(P, F);</span> 272 <span class="point"> ^~~~~~~~~~~</span> 273 <span class="loc">t.c:76:94:</span> <span class="note">note:</span> expanded from: 274 <span class="snip">#define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) __b = (B); __a < __b ? __b : __a; })</span> 275 <span class="point"> ~~~ ^ ~~~</span> 276</pre> 277 278<p>Here's another real world warning that occurs in the "window" Unix package (which 279implements the "wwopen" class of APIs):</p> 280 281<pre> 282 $ <span class="cmd">clang -fsyntax-only t.c</span> 283 <span class="loc">t.c:22:2:</span> <span class="warn">warning:</span> <span class="msg">type specifier missing, defaults to 'int'</span> 284 <span class="snip"> ILPAD();</span> 285 <span class="point"> ^</span> 286 <span class="loc">t.c:17:17:</span> <span class="note">note:</span> expanded from: 287 <span class="snip">#define ILPAD() PAD((NROW - tt.tt_row) * 10) /* 1 ms per char */</span> 288 <span class="point"> ^</span> 289 <span class="loc">t.c:14:2:</span> <span class="note">note:</span> expanded from: 290 <span class="snip"> register i; \</span> 291 <span class="point"> ^</span> 292</pre> 293 294<p>In practice, we've found that Clang's treatment of macros is actually more useful in multiply nested 295macros than in simple ones.</p> 296 297<h2>Quality of Implementation and Attention to Detail</h2> 298 299<p>Finally, we have put a lot of work polishing the little things, because 300little things add up over time and contribute to a great user experience.</p> 301 302<p>The following example shows that we recover from the simple case of 303forgetting a ; after a struct definition much better than GCC.</p> 304 305<pre> 306 $ <span class="cmd">cat t.cc</span> 307 template<class T> 308 class a {}; 309 struct b {} 310 a<int> c; 311 $ <span class="cmd">gcc-4.9 t.cc</span> 312 t.cc:4:8: error: invalid declarator before 'c' 313 a<int> c; 314 ^ 315 $ <span class="cmd">clang t.cc</span> 316 <span class="loc">t.cc:3:12:</span> <span class="err">error:</span> <span class="msg">expected ';' after struct</span> 317 <span class="snip" >struct b {}</span> 318 <span class="point"> ^</span> 319 <span class="point"> ;</span> 320</pre> 321 322<p>The following example shows that we diagnose and recover from a missing 323<tt>typename</tt> keyword well, even in complex circumstances where GCC 324cannot cope.</p> 325 326<pre> 327 $ <span class="cmd">cat t.cc</span> 328 template<class T> void f(T::type) { } 329 struct A { }; 330 void g() 331 { 332 A a; 333 f<A>(a); 334 } 335 $ <span class="cmd">gcc-4.9 t.cc</span> 336 t.cc:1:33: error: variable or field 'f' declared void 337 template<class T> void f(T::type) { } 338 ^ 339 t.cc: In function 'void g()': 340 t.cc:6:5: error: 'f' was not declared in this scope 341 f<A>(a); 342 ^ 343 t.cc:6:8: error: expected primary-expression before '>' token 344 f<A>(a); 345 ^ 346 $ <span class="cmd">clang t.cc</span> 347 <span class="loc">t.cc:1:26:</span> <span class="err">error:</span> <span class="msg">missing 'typename' prior to dependent type name 'T::type'</span> 348 <span class="snip" >template<class T> void f(T::type) { }</span> 349 <span class="point"> ^~~~~~~</span> 350 <span class="point"> typename </span> 351 <span class="loc">t.cc:6:5:</span> <span class="err">error:</span> <span class="msg">no matching function for call to 'f'</span> 352 <span class="snip" > f<A>(a);</span> 353 <span class="point"> ^~~~</span> 354 <span class="loc">t.cc:1:24:</span> <span class="note">note:</span> <span class="msg">candidate template ignored: substitution failure [with T = A]: no type named 'type' in 'A'</span> 355 <span class="snip" >template<class T> void f(T::type) { }</span> 356 <span class="point"> ^ ~~~~</span> 357</pre> 358 359 360 361<p>While each of these details is minor, we feel that they all add up to provide 362a much more polished experience.</p> 363 364</div> 365</body> 366</html> 367