support.xml revision 1.1
1<?xml version='1.0'?> 2<!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN" 3 "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd" 4[ ]> 5 6<chapter id="std.support" xreflabel="Support"> 7<?dbhtml filename="support.html"?> 8 9<chapterinfo> 10 <keywordset> 11 <keyword> 12 ISO C++ 13 </keyword> 14 <keyword> 15 library 16 </keyword> 17 </keywordset> 18</chapterinfo> 19 20<title> 21 Support 22 <indexterm><primary>Support</primary></indexterm> 23</title> 24 25 <para> 26 This part deals with the functions called and objects created 27 automatically during the course of a program's existence. 28 </para> 29 30 <para> 31 While we can't reproduce the contents of the Standard here (you 32 need to get your own copy from your nation's member body; see our 33 homepage for help), we can mention a couple of changes in what 34 kind of support a C++ program gets from the Standard Library. 35 </para> 36 37<sect1 id="std.support.types" xreflabel="Types"> 38 <?dbhtml filename="fundamental_types.html"?> 39 <title>Types</title> 40 <sect2 id="std.support.types.fundamental" xreflabel="Fundamental Types"> 41 <title>Fundamental Types</title> 42 <para> 43 C++ has the following builtin types: 44 </para> 45 <itemizedlist> 46 <listitem><para> 47 char 48 </para></listitem> 49 <listitem><para> 50 signed char 51 </para></listitem> 52 <listitem><para> 53 unsigned char 54 </para></listitem> 55 <listitem><para> 56 signed short 57 </para></listitem> 58 <listitem><para> 59 signed int 60 </para></listitem> 61 <listitem><para> 62 signed long 63 </para></listitem> 64 <listitem><para> 65 unsigned short 66 </para></listitem> 67 <listitem><para> 68 unsigned int 69 </para></listitem> 70 <listitem><para> 71 unsigned long 72 </para></listitem> 73 <listitem><para> 74 bool 75 </para></listitem> 76 <listitem><para> 77 wchar_t 78 </para></listitem> 79 <listitem><para> 80 float 81 </para></listitem> 82 <listitem><para> 83 double 84 </para></listitem> 85 <listitem><para> 86 long double 87 </para></listitem> 88 </itemizedlist> 89 90 <para> 91 These fundamental types are always available, without having to 92 include a header file. These types are exactly the same in 93 either C++ or in C. 94 </para> 95 96 <para> 97 Specializing parts of the library on these types is prohibited: 98 instead, use a POD. 99 </para> 100 101 </sect2> 102 <sect2 id="std.support.types.numeric_limits" xreflabel="Numeric Properties"> 103 <title>Numeric Properties</title> 104 105 106 <para> 107 The header <filename class="headerfile">limits</filename> defines 108 traits classes to give access to various implementation 109 defined-aspects of the fundamental types. The traits classes -- 110 fourteen in total -- are all specializations of the template class 111 <classname>numeric_limits</classname>, documented <ulink 112 url="http://gcc.gnu.org/onlinedocs/libstdc++/latest-doxygen/a00593.html">here</ulink> 113 and defined as follows: 114 </para> 115 116 <programlisting> 117 template<typename T> 118 struct class 119 { 120 static const bool is_specialized; 121 static T max() throw(); 122 static T min() throw(); 123 124 static const int digits; 125 static const int digits10; 126 static const bool is_signed; 127 static const bool is_integer; 128 static const bool is_exact; 129 static const int radix; 130 static T epsilon() throw(); 131 static T round_error() throw(); 132 133 static const int min_exponent; 134 static const int min_exponent10; 135 static const int max_exponent; 136 static const int max_exponent10; 137 138 static const bool has_infinity; 139 static const bool has_quiet_NaN; 140 static const bool has_signaling_NaN; 141 static const float_denorm_style has_denorm; 142 static const bool has_denorm_loss; 143 static T infinity() throw(); 144 static T quiet_NaN() throw(); 145 static T denorm_min() throw(); 146 147 static const bool is_iec559; 148 static const bool is_bounded; 149 static const bool is_modulo; 150 151 static const bool traps; 152 static const bool tinyness_before; 153 static const float_round_style round_style; 154 }; 155 </programlisting> 156 </sect2> 157 158 <sect2 id="std.support.types.null" xreflabel="NULL"> 159 <title>NULL</title> 160 <para> 161 The only change that might affect people is the type of 162 <constant>NULL</constant>: while it is required to be a macro, 163 the definition of that macro is <emphasis>not</emphasis> allowed 164 to be <constant>(void*)0</constant>, which is often used in C. 165 </para> 166 167 <para> 168 For <command>g++</command>, <constant>NULL</constant> is 169 <programlisting>#define</programlisting>'d to be 170 <constant>__null</constant>, a magic keyword extension of 171 <command>g++</command>. 172 </para> 173 174 <para> 175 The biggest problem of #defining <constant>NULL</constant> to be 176 something like <quote>0L</quote> is that the compiler will view 177 that as a long integer before it views it as a pointer, so 178 overloading won't do what you expect. (This is why 179 <command>g++</command> has a magic extension, so that 180 <constant>NULL</constant> is always a pointer.) 181 </para> 182 183 <para>In his book <ulink 184 url="http://www.awprofessional.com/titles/0-201-92488-9/"><emphasis>Effective 185 C++</emphasis></ulink>, Scott Meyers points out that the best way 186 to solve this problem is to not overload on pointer-vs-integer 187 types to begin with. He also offers a way to make your own magic 188 <constant>NULL</constant> that will match pointers before it 189 matches integers. 190 </para> 191 <para>See 192 <ulink url="http://www.awprofessional.com/titles/0-201-31015-5/">the 193 Effective C++ CD example</ulink> 194 </para> 195 </sect2> 196 197</sect1> 198 199<sect1 id="std.support.memory" xreflabel="Dynamic Memory"> 200 <?dbhtml filename="dynamic_memory.html"?> 201 <title>Dynamic Memory</title> 202 <para> 203 There are six flavors each of <function>new</function> and 204 <function>delete</function>, so make certain that you're using the right 205 ones. Here are quickie descriptions of <function>new</function>: 206 </para> 207 <itemizedlist> 208 <listitem><para> 209 single object form, throwing a 210 <classname>bad_alloc</classname> on errors; this is what most 211 people are used to using 212 </para></listitem> 213 <listitem><para> 214 Single object "nothrow" form, returning NULL on errors 215 </para></listitem> 216 <listitem><para> 217 Array <function>new</function>, throwing 218 <classname>bad_alloc</classname> on errors 219 </para></listitem> 220 <listitem><para> 221 Array nothrow <function>new</function>, returning 222 <constant>NULL</constant> on errors 223 </para></listitem> 224 <listitem><para> 225 Placement <function>new</function>, which does nothing (like 226 it's supposed to) 227 </para></listitem> 228 <listitem><para> 229 Placement array <function>new</function>, which also does 230 nothing 231 </para></listitem> 232 </itemizedlist> 233 <para> 234 They are distinguished by the parameters that you pass to them, like 235 any other overloaded function. The six flavors of <function>delete</function> 236 are distinguished the same way, but none of them are allowed to throw 237 an exception under any circumstances anyhow. (They match up for 238 completeness' sake.) 239 </para> 240 <para> 241 Remember that it is perfectly okay to call <function>delete</function> on a 242 NULL pointer! Nothing happens, by definition. That is not the 243 same thing as deleting a pointer twice. 244 </para> 245 <para> 246 By default, if one of the <quote>throwing <function>new</function>s</quote> can't 247 allocate the memory requested, it tosses an instance of a 248 <classname>bad_alloc</classname> exception (or, technically, some class derived 249 from it). You can change this by writing your own function (called a 250 new-handler) and then registering it with <function>set_new_handler()</function>: 251 </para> 252 <programlisting> 253 typedef void (*PFV)(void); 254 255 static char* safety; 256 static PFV old_handler; 257 258 void my_new_handler () 259 { 260 delete[] safety; 261 popup_window ("Dude, you are running low on heap memory. You 262 should, like, close some windows, or something. 263 The next time you run out, we're gonna burn!"); 264 set_new_handler (old_handler); 265 return; 266 } 267 268 int main () 269 { 270 safety = new char[500000]; 271 old_handler = set_new_handler (&my_new_handler); 272 ... 273 } 274 </programlisting> 275 <para> 276 <classname>bad_alloc</classname> is derived from the base <classname>exception</classname> 277 class defined in Sect1 19. 278 </para> 279</sect1> 280 281<sect1 id="std.support.termination" xreflabel="Termination"> 282 <?dbhtml filename="termination.html"?> 283 <title>Termination</title> 284 <sect2 id="support.termination.handlers" xreflabel="Termination Handlers"> 285 <title>Termination Handlers</title> 286 <para> 287 Not many changes here to <filename 288 class="headerfile">cstdlib</filename>. You should note that the 289 <function>abort()</function> function does not call the 290 destructors of automatic nor static objects, so if you're 291 depending on those to do cleanup, it isn't going to happen. 292 (The functions registered with <function>atexit()</function> 293 don't get called either, so you can forget about that 294 possibility, too.) 295 </para> 296 <para> 297 The good old <function>exit()</function> function can be a bit 298 funky, too, until you look closer. Basically, three points to 299 remember are: 300 </para> 301 <orderedlist> 302 <listitem> 303 <para> 304 Static objects are destroyed in reverse order of their creation. 305 </para> 306 </listitem> 307 <listitem> 308 <para> 309 Functions registered with <function>atexit()</function> are called in 310 reverse order of registration, once per registration call. 311 (This isn't actually new.) 312 </para> 313 </listitem> 314 <listitem> 315 <para> 316 The previous two actions are <quote>interleaved,</quote> that is, 317 given this pseudocode: 318 </para> 319<programlisting> 320 extern "C or C++" void f1 (void); 321 extern "C or C++" void f2 (void); 322 323 static Thing obj1; 324 atexit(f1); 325 static Thing obj2; 326 atexit(f2); 327</programlisting> 328 <para> 329 then at a call of <function>exit()</function>, 330 <varname>f2</varname> will be called, then 331 <varname>obj2</varname> will be destroyed, then 332 <varname>f1</varname> will be called, and finally 333 <varname>obj1</varname> will be destroyed. If 334 <varname>f1</varname> or <varname>f2</varname> allow an 335 exception to propagate out of them, Bad Things happen. 336 </para> 337 </listitem> 338 </orderedlist> 339 <para> 340 Note also that <function>atexit()</function> is only required to store 32 341 functions, and the compiler/library might already be using some of 342 those slots. If you think you may run out, we recommend using 343 the <function>xatexit</function>/<function>xexit</function> combination from <literal>libiberty</literal>, which has no such limit. 344 </para> 345 </sect2> 346 347 <sect2 id="support.termination.verbose" xreflabel="Verbose Terminate Handler"> 348 <?dbhtml filename="verbose_termination.html"?> 349 <title>Verbose Terminate Handler</title> 350 <para> 351 If you are having difficulty with uncaught exceptions and want a 352 little bit of help debugging the causes of the core dumps, you can 353 make use of a GNU extension, the verbose terminate handler. 354 </para> 355 356<programlisting> 357#include <exception> 358 359int main() 360{ 361 std::set_terminate(__gnu_cxx::__verbose_terminate_handler); 362 ... 363 364 throw <replaceable>anything</replaceable>; 365} 366</programlisting> 367 368 <para> 369 The <function>__verbose_terminate_handler</function> function 370 obtains the name of the current exception, attempts to demangle 371 it, and prints it to stderr. If the exception is derived from 372 <classname>exception</classname> then the output from 373 <function>what()</function> will be included. 374 </para> 375 376 <para> 377 Any replacement termination function is required to kill the 378 program without returning; this one calls abort. 379 </para> 380 381 <para> 382 For example: 383 </para> 384 385<programlisting> 386#include <exception> 387#include <stdexcept> 388 389struct argument_error : public std::runtime_error 390{ 391 argument_error(const std::string& s): std::runtime_error(s) { } 392}; 393 394int main(int argc) 395{ 396 std::set_terminate(__gnu_cxx::__verbose_terminate_handler); 397 if (argc > 5) 398 throw argument_error(<quote>argc is greater than 5!</quote>); 399 else 400 throw argc; 401} 402</programlisting> 403 404 <para> 405 With the verbose terminate handler active, this gives: 406 </para> 407 408 <screen> 409 <computeroutput> 410 % ./a.out 411 terminate called after throwing a `int' 412 Aborted 413 % ./a.out f f f f f f f f f f f 414 terminate called after throwing an instance of `argument_error' 415 what(): argc is greater than 5! 416 Aborted 417 </computeroutput> 418 </screen> 419 420 <para> 421 The 'Aborted' line comes from the call to 422 <function>abort()</function>, of course. 423 </para> 424 425 <para> 426 This is the default termination handler; nothing need be done to 427 use it. To go back to the previous <quote>silent death</quote> 428 method, simply include <filename>exception</filename> and 429 <filename>cstdlib</filename>, and call 430 </para> 431 432 <programlisting> 433 std::set_terminate(std::abort); 434 </programlisting> 435 436 <para> 437 After this, all calls to <function>terminate</function> will use 438 <function>abort</function> as the terminate handler. 439 </para> 440 441 <para> 442 Note: the verbose terminate handler will attempt to write to 443 stderr. If your application closes stderr or redirects it to an 444 inappropriate location, 445 <function>__verbose_terminate_handler</function> will behave in 446 an unspecified manner. 447 </para> 448 449 </sect2> 450</sect1> 451 452</chapter> 453