![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to gcc.info-9 CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Apple XNU] / GNUtools / cc|
Return to gcc.info-9 CVS log | Up to [Apple XNU] / GNUtools / cc |
1.1 root 1: This is Info file gcc.info, produced by Makeinfo-1.54 from the input
2: file gcc.texi.
3:
4: This file documents the use and the internals of the GNU compiler.
5:
6: Published by the Free Software Foundation 675 Massachusetts Avenue
7: Cambridge, MA 02139 USA
8:
9: Copyright (C) 1988, 1989, 1992, 1993 Free Software Foundation, Inc.
10:
11: Permission is granted to make and distribute verbatim copies of this
12: manual provided the copyright notice and this permission notice are
13: preserved on all copies.
14:
15: Permission is granted to copy and distribute modified versions of
16: this manual under the conditions for verbatim copying, provided also
17: that the sections entitled "GNU General Public License" and "Protect
18: Your Freedom--Fight `Look And Feel'" are included exactly as in the
19: original, and provided that the entire resulting derived work is
20: distributed under the terms of a permission notice identical to this
21: one.
22:
23: Permission is granted to copy and distribute translations of this
24: manual into another language, under the above conditions for modified
25: versions, except that the sections entitled "GNU General Public
26: License" and "Protect Your Freedom--Fight `Look And Feel'", and this
27: permission notice, may be included in translations approved by the Free
28: Software Foundation instead of in the original English.
29:
30:
31: File: gcc.info, Node: Incompatibilities, Next: Fixed Headers, Prev: External Bugs, Up: Trouble
32:
33: Incompatibilities of GNU CC
34: ===========================
35:
36: There are several noteworthy incompatibilities between GNU C and most
37: existing (non-ANSI) versions of C. The `-traditional' option
38: eliminates many of these incompatibilities, *but not all*, by telling
39: GNU C to behave like the other C compilers.
40:
41: * GNU CC normally makes string constants read-only. If several
42: identical-looking string constants are used, GNU CC stores only one
43: copy of the string.
44:
45: One consequence is that you cannot call `mktemp' with a string
46: constant argument. The function `mktemp' always alters the string
47: its argument points to.
48:
49: Another consequence is that `sscanf' does not work on some systems
50: when passed a string constant as its format control string or
51: input. This is because `sscanf' incorrectly tries to write into
52: the string constant. Likewise `fscanf' and `scanf'.
53:
54: The best solution to these problems is to change the program to use
55: `char'-array variables with initialization strings for these
56: purposes instead of string constants. But if this is not possible,
57: you can use the `-fwritable-strings' flag, which directs GNU CC to
58: handle string constants the same way most C compilers do.
59: `-traditional' also has this effect, among others.
60:
61: * `-2147483648' is positive.
62:
63: This is because 2147483648 cannot fit in the type `int', so
64: (following the ANSI C rules) its data type is `unsigned long int'.
65: Negating this value yields 2147483648 again.
66:
67: * GNU CC does not substitute macro arguments when they appear inside
68: of string constants. For example, the following macro in GNU CC
69:
70: #define foo(a) "a"
71:
72: will produce output `"a"' regardless of what the argument A is.
73:
74: The `-traditional' option directs GNU CC to handle such cases
75: (among others) in the old-fashioned (non-ANSI) fashion.
76:
77: * When you use `setjmp' and `longjmp', the only automatic variables
78: guaranteed to remain valid are those declared `volatile'. This is
79: a consequence of automatic register allocation. Consider this
80: function:
81:
82: jmp_buf j;
83:
84: foo ()
85: {
86: int a, b;
87:
88: a = fun1 ();
89: if (setjmp (j))
90: return a;
91:
92: a = fun2 ();
93: /* `longjmp (j)' may occur in `fun3'. */
94: return a + fun3 ();
95: }
96:
97: Here `a' may or may not be restored to its first value when the
98: `longjmp' occurs. If `a' is allocated in a register, then its
99: first value is restored; otherwise, it keeps the last value stored
100: in it.
101:
102: If you use the `-W' option with the `-O' option, you will get a
103: warning when GNU CC thinks such a problem might be possible.
104:
105: The `-traditional' option directs GNU C to put variables in the
106: stack by default, rather than in registers, in functions that call
107: `setjmp'. This results in the behavior found in traditional C
108: compilers.
109:
110: * Programs that use preprocessor directives in the middle of macro
111: arguments do not work with GNU CC. For example, a program like
112: this will not work:
113:
114: foobar (
115: #define luser
116: hack)
117:
118: ANSI C does not permit such a construct. It would make sense to
119: support it when `-traditional' is used, but it is too much work to
120: implement.
121:
122: * Declarations of external variables and functions within a block
123: apply only to the block containing the declaration. In other
124: words, they have the same scope as any other declaration in the
125: same place.
126:
127: In some other C compilers, a `extern' declaration affects all the
128: rest of the file even if it happens within a block.
129:
130: The `-traditional' option directs GNU C to treat all `extern'
131: declarations as global, like traditional compilers.
132:
133: * In traditional C, you can combine `long', etc., with a typedef
134: name, as shown here:
135:
136: typedef int foo;
137: typedef long foo bar;
138:
139: In ANSI C, this is not allowed: `long' and other type modifiers
140: require an explicit `int'. Because this criterion is expressed by
141: Bison grammar rules rather than C code, the `-traditional' flag
142: cannot alter it.
143:
144: * PCC allows typedef names to be used as function parameters. The
145: difficulty described immediately above applies here too.
146:
147: * PCC allows whitespace in the middle of compound assignment
148: operators such as `+='. GNU CC, following the ANSI standard, does
149: not allow this. The difficulty described immediately above
150: applies here too.
151:
152: * GNU CC complains about unterminated character constants inside of
153: preprocessor conditionals that fail. Some programs have English
154: comments enclosed in conditionals that are guaranteed to fail; if
155: these comments contain apostrophes, GNU CC will probably report an
156: error. For example, this code would produce an error:
157:
158: #if 0
159: You can't expect this to work.
160: #endif
161:
162: The best solution to such a problem is to put the text into an
163: actual C comment delimited by `/*...*/'. However, `-traditional'
164: suppresses these error messages.
165:
166: * Many user programs contain the declaration `long time ();'. In the
167: past, the system header files on many systems did not actually
168: declare `time', so it did not matter what type your program
169: declared it to return. But in systems with ANSI C headers, `time'
170: is declared to return `time_t', and if that is not the same as
171: `long', then `long time ();' is erroneous.
172:
173: The solution is to change your program to use `time_t' as the
174: return type of `time'.
175:
176: * When compiling functions that return `float', PCC converts it to a
177: double. GNU CC actually returns a `float'. If you are concerned
178: with PCC compatibility, you should declare your functions to return
179: `double'; you might as well say what you mean.
180:
181: * When compiling functions that return structures or unions, GNU CC
182: output code normally uses a method different from that used on most
183: versions of Unix. As a result, code compiled with GNU CC cannot
184: call a structure-returning function compiled with PCC, and vice
185: versa.
186:
187: The method used by GNU CC is as follows: a structure or union
188: which is 1, 2, 4 or 8 bytes long is returned like a scalar. A
189: structure or union with any other size is stored into an address
190: supplied by the caller (usually in a special, fixed register, but
191: on some machines it is passed on the stack). The
192: machine-description macros `STRUCT_VALUE' and
193: `STRUCT_INCOMING_VALUE' tell GNU CC where to pass this address.
194:
195: By contrast, PCC on most target machines returns structures and
196: unions of any size by copying the data into an area of static
197: storage, and then returning the address of that storage as if it
198: were a pointer value. The caller must copy the data from that
199: memory area to the place where the value is wanted. GNU CC does
200: not use this method because it is slower and nonreentrant.
201:
202: On some newer machines, PCC uses a reentrant convention for all
203: structure and union returning. GNU CC on most of these machines
204: uses a compatible convention when returning structures and unions
205: in memory, but still returns small structures and unions in
206: registers.
207:
208: You can tell GNU CC to use a compatible convention for all
209: structure and union returning with the option
210: `-fpcc-struct-return'.
211:
212: * GNU C complains about program fragments such as `0x74ae-0x4000'
213: which appear to be two hexadecimal constants separated by the minus
214: operator. Actually, this string is a single "preprocessing token".
215: Each such token must correspond to one token in C. Since this
216: does not, GNU C prints an error message. Although it may appear
217: obvious that what is meant is an operator and two values, the ANSI
218: C standard specifically requires that this be treated as erroneous.
219:
220: A "preprocessing token" is a "preprocessing number" if it begins
221: with a digit and is followed by letters, underscores, digits,
222: periods and `e+', `e-', `E+', or `E-' character sequences.
223:
224: To make the above program fragment valid, place whitespace in
225: front of the minus sign. This whitespace will end the
226: preprocessing number.
227:
228:
229: File: gcc.info, Node: Fixed Headers, Next: Disappointments, Prev: Incompatibilities, Up: Trouble
230:
231: Fixed Header Files
232: ==================
233:
234: GNU CC needs to install corrected versions of some system header
235: files. This is because most target systems have some header files that
236: won't work with GNU CC unless they are changed. Some have bugs, some
237: are incompatible with ANSI C, and some depend on special features of
238: other compilers.
239:
240: Installing GNU CC automatically creates and installs the fixed header
241: files, by running a program called `fixincludes' (or for certain
242: targets an alternative such as `fixinc.svr4'). Normally, you don't
243: need to pay attention to this. But there are cases where it doesn't do
244: the right thing automatically.
245:
246: * If you update the system's header files, such as by installing a
247: new system version, the fixed header files of GNU CC are not
248: automatically updated. The easiest way to update them is to
249: reinstall GNU CC. (If you want to be clever, look in the makefile
250: and you can find a shortcut.)
251:
252: * On some systems, in particular SunOS 4, header file directories
253: contain machine-specific symbolic links in certain places. This
254: makes it possible to share most of the header files among hosts
255: running the same version of SunOS 4 on different machine models.
256:
257: The programs that fix the header files do not understand this
258: special way of using symbolic links; therefore, the directory of
259: fixed header files is good only for the machine model used to
260: build it.
261:
262: In SunOS 4, only programs that look inside the kernel will notice
263: the difference between machine models. Therefore, for most
264: purposes, you need not be concerned about this.
265:
266: It is possible to make separate sets of fixed header files for the
267: different machine models, and arrange a structure of symbolic
268: links so as to use the proper set, but you'll have to do this by
269: hand.
270:
271:
272: File: gcc.info, Node: Disappointments, Next: C++ Misunderstandings, Prev: Fixed Headers, Up: Trouble
273:
274: Disappointments and Misunderstandings
275: =====================================
276:
277: These problems are perhaps regrettable, but we don't know any
278: practical way around them.
279:
280: * Certain local variables aren't recognized by debuggers when you
281: compile with optimization.
282:
283: This occurs because sometimes GNU CC optimizes the variable out of
284: existence. There is no way to tell the debugger how to compute the
285: value such a variable "would have had", and it is not clear that
286: would be desirable anyway. So GNU CC simply does not mention the
287: eliminated variable when it writes debugging information.
288:
289: You have to expect a certain amount of disagreement between the
290: executable and your source code, when you use optimization.
291:
292: * Users often think it is a bug when GNU CC reports an error for code
293: like this:
294:
295: int foo (struct mumble *);
296:
297: struct mumble { ... };
298:
299: int foo (struct mumble *x)
300: { ... }
301:
302: This code really is erroneous, because the scope of `struct
303: mumble' in the prototype is limited to the argument list
304: containing it. It does not refer to the `struct mumble' defined
305: with file scope immediately below--they are two unrelated types
306: with similar names in different scopes.
307:
308: But in the definition of `foo', the file-scope type is used
309: because that is available to be inherited. Thus, the definition
310: and the prototype do not match, and you get an error.
311:
312: This behavior may seem silly, but it's what the ANSI standard
313: specifies. It is easy enough for you to make your code work by
314: moving the definition of `struct mumble' above the prototype.
315: It's not worth being incompatible with ANSI C just to avoid an
316: error for the example shown above.
317:
318: * Accesses to bitfields even in volatile objects works by accessing
319: larger objects, such as a byte or a word. You cannot rely on what
320: size of object is accessed in order to read or write the bitfield;
321: it may even vary for a given bitfield according to the precise
322: usage.
323:
324: If you care about controlling the amount of memory that is
325: accessed, use volatile but do not use bitfields.
326:
327: * GNU CC comes with shell scripts to fix certain known problems in
328: system header files. They install corrected copies of various
329: header files in a special directory where only GNU CC will
330: normally look for them. The scripts adapt to various systems by
331: searching all the system header files for the problem cases that
332: we know about.
333:
334: If new system header files are installed, nothing automatically
335: arranges to update the corrected header files. You will have to
336: reinstall GNU CC to fix the new header files. More specifically,
337: go to the build directory and delete the files `stmp-fixinc' and
338: `stmp-headers', and the subdirectory `include'; then do `make
339: install' again.
340:
341: * On 68000 systems, you can get paradoxical results if you test the
342: precise values of floating point numbers. For example, you can
343: find that a floating point value which is not a NaN is not equal
344: to itself. This results from the fact that the the floating point
345: registers hold a few more bits of precision than fit in a `double'
346: in memory. Compiled code moves values between memory and floating
347: point registers at its convenience, and moving them into memory
348: truncates them.
349:
350: You can partially avoid this problem by using the `-ffloat-store'
351: option (*note Optimize Options::.).
352:
353: * On the MIPS, variable argument functions using `varargs.h' cannot
354: have a floating point value for the first argument. The reason
355: for this is that in the absence of a prototype in scope, if the
356: first argument is a floating point, it is passed in a floating
357: point register, rather than an integer register.
358:
359: If the code is rewritten to use the ANSI standard `stdarg.h'
360: method of variable arguments, and the prototype is in scope at the
361: time of the call, everything will work fine.
362:
363:
364: File: gcc.info, Node: C++ Misunderstandings, Next: Protoize Caveats, Prev: Disappointments, Up: Trouble
365:
366: Common Misunderstandings with GNU C++
367: =====================================
368:
369: C++ is a complex language and an evolving one, and its standard
370: definition (the ANSI C++ draft standard) is also evolving. As a result,
371: your C++ compiler may occasionally surprise you, even when its behavior
372: is correct. This section discusses some areas that frequently give
373: rise to questions of this sort.
374:
375: * Menu:
376:
377: * Static Definitions:: Static member declarations are not definitions
378: * Temporaries:: Temporaries may vanish before you expect
379:
380:
381: File: gcc.info, Node: Static Definitions, Next: Temporaries, Up: C++ Misunderstandings
382:
383: Declare *and* Define Static Members
384: -----------------------------------
385:
386: When a class has static data members, it is not enough to *declare*
387: the static member; you must also *define* it. For example:
388:
389: class Foo
390: {
391: ...
392: void method();
393: static int bar;
394: };
395:
396: This declaration only establishes that the class `Foo' has an `int'
397: named `Foo::bar', and a member function named `Foo::method'. But you
398: still need to define *both* `method' and `bar' elsewhere. According to
399: the draft ANSI standard, you must supply an initializer in one (and
400: only one) source file, such as:
401:
402: int Foo::bar = 0;
403:
404: Other C++ compilers may not correctly implement the standard
405: behavior. As a result, when you switch to `g++' from one of these
406: compilers, you may discover that a program that appeared to work
407: correctly in fact does not conform to the standard: `g++' reports as
408: undefined symbols any static data members that lack definitions.
409:
410:
411: File: gcc.info, Node: Temporaries, Prev: Static Definitions, Up: C++ Misunderstandings
412:
413: Temporaries May Vanish Before You Expect
414: ----------------------------------------
415:
416: It is dangerous to use pointers or references to *portions* of a
417: temporary object. The compiler may very well delete the object before
418: you expect it to, leaving a pointer to garbage. The most common place
419: where this problem crops up is in classes like the libg++ `String'
420: class, that define a conversion function to type `char *' or `const
421: char *'. However, any class that returns a pointer to some internal
422: structure is potentially subject to this problem.
423:
424: For example, a program may use a function `strfunc' that returns
425: `String' objects, and another function `charfunc' that operates on
426: pointers to `char':
427:
428: String strfunc ();
429: void charfunc (const char *);
430:
431: In this situation, it may seem natural to write
432: `charfunc (strfunc ());' based on the knowledge that class `String' has
433: an explicit conversion to `char' pointers. However, what really
434: happens is akin to `charfunc (strfunc ().convert ());', where the
435: `convert' method is a function to do the same data conversion normally
436: performed by a cast. Since the last use of the temporary `String'
437: object is the call to the conversion function, the compiler may delete
438: that object before actually calling `charfunc'. The compiler has no
439: way of knowing that deleting the `String' object will invalidate the
440: pointer. The pointer then points to garbage, so that by the time
441: `charfunc' is called, it gets an invalid argument.
442:
443: Code like this may run successfully under some other compilers,
444: especially those that delete temporaries relatively late. However, the
445: GNU C++ behavior is also standard-conformant, so if your program depends
446: on late destruction of temporaries it is not portable.
447:
448: If you think this is surprising, you should be aware that the ANSI
449: C++ committee continues to debate the lifetime-of-temporaries problem.
450:
451: For now, at least, the safe way to write such code is to give the
452: temporary a name, which forces it to remain until the end of the scope
453: of the name. For example:
454:
455: String& tmp = strfunc ();
456: charfunc (tmp);
457:
458:
459: File: gcc.info, Node: Protoize Caveats, Next: Non-bugs, Prev: C++ Misunderstandings, Up: Trouble
460:
461: Caveats of using `protoize'
462: ===========================
463:
464: The conversion programs `protoize' and `unprotoize' can sometimes
465: change a source file in a way that won't work unless you rearrange it.
466:
467: * `protoize' can insert references to a type name or type tag before
468: the definition, or in a file where they are not defined.
469:
470: If this happens, compiler error messages should show you where the
471: new references are, so fixing the file by hand is straightforward.
472:
473: * There are some C constructs which `protoize' cannot figure out.
474: For example, it can't determine argument types for declaring a
475: pointer-to-function variable; this you must do by hand. `protoize'
476: inserts a comment containing `???' each time it finds such a
477: variable; so you can find all such variables by searching for this
478: string. ANSI C does not require declaring the argument types of
479: pointer-to-function types.
480:
481: * Using `unprotoize' can easily introduce bugs. If the program
482: relied on prototypes to bring about conversion of arguments, these
483: conversions will not take place in the program without prototypes.
484: One case in which you can be sure `unprotoize' is safe is when you
485: are removing prototypes that were made with `protoize'; if the
486: program worked before without any prototypes, it will work again
487: without them.
488:
489: You can find all the places where this problem might occur by
490: compiling the program with the `-Wconversion' option. It prints a
491: warning whenever an argument is converted.
492:
493: * Both conversion programs can be confused if there are macro calls
494: in and around the text to be converted. In other words, the
495: standard syntax for a declaration or definition must not result
496: from expanding a macro. This problem is inherent in the design of
497: C and cannot be fixed. If only a few functions have confusing
498: macro calls, you can easily convert them manually.
499:
500: * `protoize' cannot get the argument types for a function whose
501: definition was not actually compiled due to preprocessor
502: conditionals. When this happens, `protoize' changes nothing in
503: regard to such a function. `protoize' tries to detect such
504: instances and warn about them.
505:
506: You can generally work around this problem by using `protoize' step
507: by step, each time specifying a different set of `-D' options for
508: compilation, until all of the functions have been converted.
509: There is no automatic way to verify that you have got them all,
510: however.
511:
512: * Confusion may result if there is an occasion to convert a function
513: declaration or definition in a region of source code where there
514: is more than one formal parameter list present. Thus, attempts to
515: convert code containing multiple (conditionally compiled) versions
516: of a single function header (in the same vicinity) may not produce
517: the desired (or expected) results.
518:
519: If you plan on converting source files which contain such code, it
520: is recommended that you first make sure that each conditionally
521: compiled region of source code which contains an alternative
522: function header also contains at least one additional follower
523: token (past the final right parenthesis of the function header).
524: This should circumvent the problem.
525:
526: * `unprotoize' can become confused when trying to convert a function
527: definition or declaration which contains a declaration for a
528: pointer-to-function formal argument which has the same name as the
529: function being defined or declared. We recommand you avoid such
530: choices of formal parameter names.
531:
532: * You might also want to correct some of the indentation by hand and
533: break long lines. (The conversion programs don't write lines
534: longer than eighty characters in any case.)
535:
536:
537: File: gcc.info, Node: Non-bugs, Next: Warnings and Errors, Prev: Protoize Caveats, Up: Trouble
538:
539: Certain Changes We Don't Want to Make
540: =====================================
541:
542: This section lists changes that people frequently request, but which
543: we do not make because we think GNU CC is better without them.
544:
545: * Checking the number and type of arguments to a function which has
546: an old-fashioned definition and no prototype.
547:
548: Such a feature would work only occasionally--only for calls that
549: appear in the same file as the called function, following the
550: definition. The only way to check all calls reliably is to add a
551: prototype for the function. But adding a prototype eliminates the
552: motivation for this feature. So the feature is not worthwhile.
553:
554: * Warning about using an expression whose type is signed as a shift
555: count.
556:
557: Shift count operands are probably signed more often than unsigned.
558: Warning about this would cause far more annoyance than good.
559:
560: * Warning about assigning a signed value to an unsigned variable.
561:
562: Such assignments must be very common; warning about them would
563: cause more annoyance than good.
564:
565: * Warning about unreachable code.
566:
567: It's very common to have unreachable code in machine-generated
568: programs. For example, this happens normally in some files of GNU
569: C itself.
570:
571: * Warning when a non-void function value is ignored.
572:
573: Coming as I do from a Lisp background, I balk at the idea that
574: there is something dangerous about discarding a value. There are
575: functions that return values which some callers may find useful;
576: it makes no sense to clutter the program with a cast to `void'
577: whenever the value isn't useful.
578:
579: * Assuming (for optimization) that the address of an external symbol
580: is never zero.
581:
582: This assumption is false on certain systems when `#pragma weak' is
583: used.
584:
585: * Making `-fshort-enums' the default.
586:
587: This would cause storage layout to be incompatible with most other
588: C compilers. And it doesn't seem very important, given that you
589: can get the same result in other ways. The case where it matters
590: most is when the enumeration-valued object is inside a structure,
591: and in that case you can specify a field width explicitly.
592:
593: * Making bitfields unsigned by default on particular machines where
594: "the ABI standard" says to do so.
595:
596: The ANSI C standard leaves it up to the implementation whether a
597: bitfield declared plain `int' is signed or not. This in effect
598: creates two alternative dialects of C.
599:
600: The GNU C compiler supports both dialects; you can specify the
601: signed dialect with `-fsigned-bitfields' and the unsigned dialect
602: with `-funsigned-bitfields'. However, this leaves open the
603: question of which dialect to use by default.
604:
605: Currently, the preferred dialect makes plain bitfields signed,
606: because this is simplest. Since `int' is the same as `signed int'
607: in every other context, it is cleanest for them to be the same in
608: bitfields as well.
609:
610: Some computer manufacturers have published Application Binary
611: Interface standards which specify that plain bitfields should be
612: unsigned. It is a mistake, however, to say anything about this
613: issue in an ABI. This is because the handling of plain bitfields
614: distinguishes two dialects of C. Both dialects are meaningful on
615: every type of machine. Whether a particular object file was
616: compiled using signed bitfields or unsigned is of no concern to
617: other object files, even if they access the same bitfields in the
618: same data structures.
619:
620: A given program is written in one or the other of these two
621: dialects. The program stands a chance to work on most any machine
622: if it is compiled with the proper dialect. It is unlikely to work
623: at all if compiled with the wrong dialect.
624:
625: Many users appreciate the GNU C compiler because it provides an
626: environment that is uniform across machines. These users would be
627: inconvenienced if the compiler treated plain bitfields differently
628: on certain machines.
629:
630: Occasionally users write programs intended only for a particular
631: machine type. On these occasions, the users would benefit if the
632: GNU C compiler were to support by default the same dialect as the
633: other compilers on that machine. But such applications are rare.
634: And users writing a program to run on more than one type of
635: machine cannot possibly benefit from this kind of compatibility.
636:
637: This is why GNU CC does and will treat plain bitfields in the same
638: fashion on all types of machines (by default).
639:
640: There are some arguments for making bitfields unsigned by default
641: on all machines. If, for example, this becomes a universal de
642: facto standard, it would make sense for GNU CC to go along with
643: it. This is something to be considered in the future.
644:
645: (Of course, users strongly concerned about portability should
646: indicate explicitly in each bitfield whether it is signed or not.
647: In this way, they write programs which have the same meaning in
648: both C dialects.)
649:
650: * Undefining `__STDC__' when `-ansi' is not used.
651:
652: Currently, GNU CC defines `__STDC__' as long as you don't use
653: `-traditional'. This provides good results in practice.
654:
655: Programmers normally use conditionals on `__STDC__' to ask whether
656: it is safe to use certain features of ANSI C, such as function
657: prototypes or ANSI token concatenation. Since plain `gcc' supports
658: all the features of ANSI C, the correct answer to these questions
659: is "yes".
660:
661: Some users try to use `__STDC__' to check for the availability of
662: certain library facilities. This is actually incorrect usage in
663: an ANSI C program, because the ANSI C standard says that a
664: conforming freestanding implementation should define `__STDC__'
665: even though it does not have the library facilities. `gcc -ansi
666: -pedantic' is a conforming freestanding implementation, and it is
667: therefore required to define `__STDC__', even though it does not
668: come with an ANSI C library.
669:
670: Sometimes people say that defining `__STDC__' in a compiler that
671: does not completely conform to the ANSI C standard somehow
672: violates the standard. This is illogical. The standard is a
673: standard for compilers that claim to support ANSI C, such as `gcc
674: -ansi'--not for other compilers such as plain `gcc'. Whatever the
675: ANSI C standard says is relevant to the design of plain `gcc'
676: without `-ansi' only for pragmatic reasons, not as a requirement.
677:
678: * Undefining `__STDC__' in C++.
679:
680: Programs written to compile with C++-to-C translators get the
681: value of `__STDC__' that goes with the C compiler that is
682: subsequently used. These programs must test `__STDC__' to
683: determine what kind of C preprocessor that compiler uses: whether
684: they should concatenate tokens in the ANSI C fashion or in the
685: traditional fashion.
686:
687: These programs work properly with GNU C++ if `__STDC__' is defined.
688: They would not work otherwise.
689:
690: In addition, many header files are written to provide prototypes
691: in ANSI C but not in traditional C. Many of these header files
692: can work without change in C++ provided `__STDC__' is defined. If
693: `__STDC__' is not defined, they will all fail, and will all need
694: to be changed to test explicitly for C++ as well.
695:
696: * Deleting "empty" loops.
697:
698: GNU CC does not delete "empty" loops because the most likely reason
699: you would put one in a program is to have a delay. Deleting them
700: will not make real programs run any faster, so it would be
701: pointless.
702:
703: It would be different if optimization of a nonempty loop could
704: produce an empty one. But this generally can't happen.
705:
706: * Making side effects happen in the same order as in some other
707: compiler.
708:
709: It is never safe to depend on the order of evaluation of side
710: effects. For example, a function call like this may very well
711: behave differently from one compiler to another:
712:
713: void func (int, int);
714:
715: int i = 2;
716: func (i++, i++);
717:
718: There is no guarantee (in either the C or the C++ standard language
719: definitions) that the increments will be evaluated in any
720: particular order. Either increment might happen first. `func'
721: might get the arguments `3, 4', or it might get `4, 3', or even
722: `3, 3'.
723:
724: * Using the "canonical" form of the target configuration name as the
725: directory for installation.
726:
727: This would be an improvement in some respects, but it would also
728: cause problems. For one thing, users might expect to use in the
729: `-b' option the same name specified at installation; if
730: installation used the canonical form, that would not work. What's
731: more, the canonical name might be too long for certain file
732: systems.
733:
734: We suggest you make a link to the installation directory under the
735: canonical name, if you want to use that name in the `-b' option.
736:
737:
738: File: gcc.info, Node: Warnings and Errors, Prev: Non-bugs, Up: Trouble
739:
740: Warning Messages and Error Messages
741: ===================================
742:
743: The GNU compiler can produce two kinds of diagnostics: errors and
744: warnings. Each kind has a different purpose:
745:
746: *Errors* report problems that make it impossible to compile your
747: program. GNU CC reports errors with the source file name and line
748: number where the problem is apparent.
749:
750: *Warnings* report other unusual conditions in your code that *may*
751: indicate a problem, although compilation can (and does) proceed.
752: Warning messages also report the source file name and line number,
753: but include the text `warning:' to distinguish them from error
754: messages.
755:
756: Warnings may indicate danger points where you should check to make
757: sure that your program really does what you intend; or the use of
758: obsolete features; or the use of nonstandard features of GNU C or C++.
759: Many warnings are issued only if you ask for them, with one of the `-W'
760: options (for instance, `-Wall' requests a variety of useful warnings).
761:
762: GNU CC always tries to compile your program if possible; it never
763: gratuituously rejects a program whose meaning is clear merely because
764: (for instance) it fails to conform to a standard. In some cases,
765: however, the C and C++ standards specify that certain extensions are
766: forbidden, and a diagnostic *must* be issued by a conforming compiler.
767: The `-pedantic' option tells GNU CC to issue warnings in such cases;
768: `-pedantic-errors' says to make them errors instead. This does not
769: mean that *all* non-ANSI constructs get warnings or errors.
770:
771: *Note Options to Request or Suppress Warnings: Warning Options, for
772: more detail on these and related command-line options.
773:
774:
775: File: gcc.info, Node: Bugs, Next: Service, Prev: Trouble, Up: Top
776:
777: Reporting Bugs
778: **************
779:
780: Your bug reports play an essential role in making GNU CC reliable.
781:
782: When you encounter a problem, the first thing to do is to see if it
783: is already known. *Note Trouble::. If it isn't known, then you should
784: report the problem.
785:
786: Reporting a bug may help you by bringing a solution to your problem,
787: or it may not. (If it does not, look in the service directory; see
788: *Note Service::.) In any case, the principal function of a bug report
789: is to help the entire community by making the next version of GNU CC
790: work better. Bug reports are your contribution to the maintenance of
791: GNU CC.
792:
793: Since the maintainers are very overloaded, we cannot respond to every
794: bug report. However, if the bug has not been fixed, we are likely to
795: send you a patch and ask you to tell us whether it works.
796:
797: In order for a bug report to serve its purpose, you must include the
798: information that makes for fixing the bug.
799:
800: * Menu:
801:
802: * Criteria: Bug Criteria. Have you really found a bug?
803: * Where: Bug Lists. Where to send your bug report.
804: * Reporting: Bug Reporting. How to report a bug effectively.
805: * Patches: Sending Patches. How to send a patch for GNU CC.
806: * Known: Trouble. Known problems.
807: * Help: Service. Where to ask for help.
808:
809:
810: File: gcc.info, Node: Bug Criteria, Next: Bug Lists, Up: Bugs
811:
812: Have You Found a Bug?
813: =====================
814:
815: If you are not sure whether you have found a bug, here are some
816: guidelines:
817:
818: * If the compiler gets a fatal signal, for any input whatever, that
819: is a compiler bug. Reliable compilers never crash.
820:
821: * If the compiler produces invalid assembly code, for any input
822: whatever (except an `asm' statement), that is a compiler bug,
823: unless the compiler reports errors (not just warnings) which would
824: ordinarily prevent the assembler from being run.
825:
826: * If the compiler produces valid assembly code that does not
827: correctly execute the input source code, that is a compiler bug.
828:
829: However, you must double-check to make sure, because you may have
830: run into an incompatibility between GNU C and traditional C (*note
831: Incompatibilities::.). These incompatibilities might be considered
832: bugs, but they are inescapable consequences of valuable features.
833:
834: Or you may have a program whose behavior is undefined, which
835: happened by chance to give the desired results with another C or
836: C++ compiler.
837:
838: For example, in many nonoptimizing compilers, you can write `x;'
839: at the end of a function instead of `return x;', with the same
840: results. But the value of the function is undefined if `return'
841: is omitted; it is not a bug when GNU CC produces different results.
842:
843: Problems often result from expressions with two increment
844: operators, as in `f (*p++, *p++)'. Your previous compiler might
845: have interpreted that expression the way you intended; GNU CC might
846: interpret it another way. Neither compiler is wrong. The bug is
847: in your code.
848:
849: After you have localized the error to a single source line, it
850: should be easy to check for these things. If your program is
851: correct and well defined, you have found a compiler bug.
852:
853: * If the compiler produces an error message for valid input, that is
854: a compiler bug.
855:
856: * If the compiler does not produce an error message for invalid
857: input, that is a compiler bug. However, you should note that your
858: idea of "invalid input" might be my idea of "an extension" or
859: "support for traditional practice".
860:
861: * If you are an experienced user of C or C++ compilers, your
862: suggestions for improvement of GNU CC or GNU C++ are welcome in
863: any case.
864:
865:
866: File: gcc.info, Node: Bug Lists, Next: Bug Reporting, Prev: Bug Criteria, Up: Bugs
867:
868: Where to Report Bugs
869: ====================
870:
871: Send bug reports for GNU C to one of these addresses:
872:
873: [email protected]
874: {ucbvax|mit-eddie|uunet}!prep.ai.mit.edu!bug-gcc
875:
876: Send bug reports for GNU C++ to one of these addresses:
877:
878: [email protected]
879: {ucbvax|mit-eddie|uunet}!prep.ai.mit.edu!bug-g++
880:
881: If your bug involves the C++ class library libg++, send mail to
882: `[email protected]'. If you're not sure, you can send the
883: bug report to both lists.
884:
885: *Do not send bug reports to the mailing list `help-gcc', or to the
886: newsgroup `gnu.gcc.help'.* Most users of GNU CC do not want to receive
887: bug reports. Those that do, have asked to be on `bug-gcc' and/or
888: `bug-g++'.
889:
890: The mailing lists `bug-gcc' and `bug-g++' both have newsgroups which
891: serve as repeaters: `gnu.gcc.bug' and `gnu.g++.bug'. Each mailing list
892: and its newsgroup carry exactly the same messages.
893:
894: Often people think of posting bug reports to the newsgroup instead of
895: mailing them. This appears to work, but it has one problem which can be
896: crucial: a newsgroup posting does not contain a mail path back to the
897: sender. Thus, if maintainers need more information, they may be unable
898: to reach you. For this reason, you should always send bug reports by
899: mail to the proper mailing list.
900:
901: As a last resort, send bug reports on paper to:
902:
903: GNU Compiler Bugs
904: Free Software Foundation
905: 675 Mass Ave
906: Cambridge, MA 02139
907:
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.