1 This is Info file gcc.info, produced by Makeinfo version 1.68 from the
4 This file documents the use and the internals of the GNU compiler.
6 Published by the Free Software Foundation 59 Temple Place - Suite 330
7 Boston, MA 02111-1307 USA
9 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998
10 Free Software Foundation, Inc.
12 Permission is granted to make and distribute verbatim copies of this
13 manual provided the copyright notice and this permission notice are
14 preserved on all copies.
16 Permission is granted to copy and distribute modified versions of
17 this manual under the conditions for verbatim copying, provided also
18 that the sections entitled "GNU General Public License," "Funding for
19 Free Software," and "Protect Your Freedom--Fight `Look And Feel'" are
20 included exactly as in the original, and provided that the entire
21 resulting derived work is distributed under the terms of a permission
22 notice identical to this one.
24 Permission is granted to copy and distribute translations of this
25 manual into another language, under the above conditions for modified
26 versions, except that the sections entitled "GNU General Public
27 License," "Funding for Free Software," and "Protect Your Freedom--Fight
28 `Look And Feel'", and this permission notice, may be included in
29 translations approved by the Free Software Foundation instead of in the
33 File: gcc.info, Node: C++ Signatures, Prev: Template Instantiation, Up: C++ Extensions
35 Type Abstraction using Signatures
36 =================================
38 In GNU C++, you can use the keyword `signature' to define a
39 completely abstract class interface as a datatype. You can connect this
40 abstraction with actual classes using signature pointers. If you want
41 to use signatures, run the GNU compiler with the `-fhandle-signatures'
42 command-line option. (With this option, the compiler reserves a second
43 keyword `sigof' as well, for a future extension.)
45 Roughly, signatures are type abstractions or interfaces of classes.
46 Some other languages have similar facilities. C++ signatures are
47 related to ML's signatures, Haskell's type classes, definition modules
48 in Modula-2, interface modules in Modula-3, abstract types in Emerald,
49 type modules in Trellis/Owl, categories in Scratchpad II, and types in
50 POOL-I. For a more detailed discussion of signatures, see `Signatures:
51 A Language Extension for Improving Type Abstraction and Subtype
52 Polymorphism in C++' by Gerald Baumgartner and Vincent F. Russo (Tech
53 report CSD-TR-95-051, Dept. of Computer Sciences, Purdue University,
54 August 1995, a slightly improved version appeared in
55 *Software--Practice & Experience*, 25(8), pp. 863-889, August 1995).
56 You can get the tech report by anonymous FTP from `ftp.cs.purdue.edu'
57 in `pub/gb/Signature-design.ps.gz'.
59 Syntactically, a signature declaration is a collection of member
60 function declarations and nested type declarations. For example, this
61 signature declaration defines a new abstract type `S' with member
62 functions `int foo ()' and `int bar (int)':
70 Since signature types do not include implementation definitions, you
71 cannot write an instance of a signature directly. Instead, you can
72 define a pointer to any class that contains the required interfaces as a
73 "signature pointer". Such a class "implements" the signature type.
75 To use a class as an implementation of `S', you must ensure that the
76 class has public member functions `int foo ()' and `int bar (int)'.
77 The class can have other member functions as well, public or not; as
78 long as it offers what's declared in the signature, it is suitable as
79 an implementation of that signature type.
81 For example, suppose that `C' is a class that meets the requirements
82 of signature `S' (`C' "conforms to" `S'). Then
87 defines a signature pointer `p' and initializes it to point to an
88 object of type `C'. The member function call `int i = p->foo ();'
89 executes `obj.foo ()'.
91 Abstract virtual classes provide somewhat similar facilities in
92 standard C++. There are two main advantages to using signatures
95 1. Subtyping becomes independent from inheritance. A class or
96 signature type `T' is a subtype of a signature type `S'
97 independent of any inheritance hierarchy as long as all the member
98 functions declared in `S' are also found in `T'. So you can
99 define a subtype hierarchy that is completely independent from any
100 inheritance (implementation) hierarchy, instead of being forced to
101 use types that mirror the class inheritance hierarchy.
103 2. Signatures allow you to work with existing class hierarchies as
104 implementations of a signature type. If those class hierarchies
105 are only available in compiled form, you're out of luck with
106 abstract virtual classes, since an abstract virtual class cannot
107 be retrofitted on top of existing class hierarchies. So you would
108 be required to write interface classes as subtypes of the abstract
111 There is one more detail about signatures. A signature declaration
112 can contain member function *definitions* as well as member function
113 declarations. A signature member function with a full definition is
114 called a *default implementation*; classes need not contain that
115 particular interface in order to conform. For example, a class `C' can
116 conform to the signature
121 int f0 () { return f (0); };
124 whether or not `C' implements the member function `int f0 ()'. If you
125 define `C::f0', that definition takes precedence; otherwise, the
126 default implementation `S::f0' applies.
129 File: gcc.info, Node: Gcov, Next: Trouble, Prev: C++ Extensions, Up: Top
131 `gcov': a Test Coverage Program
132 *******************************
134 `gcov' is a tool you can use in conjunction with GNU CC to test code
135 coverage in your programs.
137 This chapter describes version 1.5 of `gcov'.
141 * Gcov Intro:: Introduction to gcov.
142 * Invoking Gcov:: How to use gcov.
143 * Gcov and Optimization:: Using gcov with GCC optimization.
144 * Gcov Data Files:: The files used by gcov.
147 File: gcc.info, Node: Gcov Intro, Next: Invoking Gcov, Up: Gcov
149 Introduction to `gcov'
150 ======================
152 `gcov' is a test coverage program. Use it in concert with GNU CC to
153 analyze your programs to help create more efficient, faster running
154 code. You can use `gcov' as a profiling tool to help discover where
155 your optimization efforts will best affect your code. You can also use
156 `gcov' along with the other profiling tool, `gprof', to assess which
157 parts of your code use the greatest amount of computing time.
159 Profiling tools help you analyze your code's performance. Using a
160 profiler such as `gcov' or `gprof', you can find out some basic
161 performance statistics, such as:
163 * how often each line of code executes
165 * what lines of code are actually executed
167 * how much computing time each section of code uses
169 Once you know these things about how your code works when compiled,
170 you can look at each module to see which modules should be optimized.
171 `gcov' helps you determine where to work on optimization.
173 Software developers also use coverage testing in concert with
174 testsuites, to make sure software is actually good enough for a release.
175 Testsuites can verify that a program works as expected; a coverage
176 program tests to see how much of the program is exercised by the
177 testsuite. Developers can then determine what kinds of test cases need
178 to be added to the testsuites to create both better testing and a better
181 You should compile your code without optimization if you plan to use
182 `gcov' because the optimization, by combining some lines of code into
183 one function, may not give you as much information as you need to look
184 for `hot spots' where the code is using a great deal of computer time.
185 Likewise, because `gcov' accumulates statistics by line (at the lowest
186 resolution), it works best with a programming style that places only
187 one statement on each line. If you use complicated macros that expand
188 to loops or to other control structures, the statistics are less
189 helpful--they only report on the line where the macro call appears. If
190 your complex macros behave like functions, you can replace them with
191 inline functions to solve this problem.
193 `gcov' creates a logfile called `SOURCEFILE.gcov' which indicates
194 how many times each line of a source file `SOURCEFILE.c' has executed.
195 You can use these logfiles along with `gprof' to aid in fine-tuning the
196 performance of your programs. `gprof' gives timing information you can
197 use along with the information you get from `gcov'.
199 `gcov' works only on code compiled with GNU CC. It is not
200 compatible with any other profiling or test coverage mechanism.
203 File: gcc.info, Node: Invoking Gcov, Next: Gcov and Optimization, Prev: Gcov Intro, Up: Gcov
208 gcov [-b] [-v] [-n] [-l] [-f] [-o directory] SOURCEFILE
211 Write branch frequencies to the output file, and write branch
212 summary info to the standard output. This option allows you to
213 see how often each branch in your program was taken.
216 Display the `gcov' version number (on the standard error stream).
219 Do not create the `gcov' output file.
222 Create long file names for included source files. For example, if
223 the header file `x.h' contains code, and was included in the file
224 `a.c', then running `gcov' on the file `a.c' will produce an
225 output file called `a.c.x.h.gcov' instead of `x.h.gcov'. This can
226 be useful if `x.h' is included in multiple source files.
229 Output summaries for each function in addition to the file level
233 The directory where the object files live. Gcov will search for
234 `.bb', `.bbg', and `.da' files in this directory.
236 When using `gcov', you must first compile your program with two
237 special GNU CC options: `-fprofile-arcs -ftest-coverage'. This tells
238 the compiler to generate additional information needed by gcov
239 (basically a flow graph of the program) and also includes additional
240 code in the object files for generating the extra profiling information
241 needed by gcov. These additional files are placed in the directory
242 where the source code is located.
244 Running the program will cause profile output to be generated. For
245 each source file compiled with -fprofile-arcs, an accompanying `.da'
246 file will be placed in the source directory.
248 Running `gcov' with your program's source file names as arguments
249 will now produce a listing of the code along with frequency of execution
250 for each line. For example, if your program is called `tmp.c', this is
251 what you see when you use the basic `gcov' facility:
253 $ gcc -fprofile-arcs -ftest-coverage tmp.c
256 87.50% of 8 source lines executed in file tmp.c
259 The file `tmp.c.gcov' contains output from `gcov'. Here is a sample:
267 11 for (i = 0; i < 10; i++)
271 ###### printf ("Failure\n");
273 1 printf ("Success\n");
276 When you use the `-b' option, your output looks like this:
279 87.50% of 8 source lines executed in file tmp.c
280 80.00% of 5 branches executed in file tmp.c
281 80.00% of 5 branches taken at least once in file tmp.c
282 50.00% of 2 calls executed in file tmp.c
285 Here is a sample of a resulting `tmp.c.gcov' file:
293 11 for (i = 0; i < 10; i++)
295 branch 1 taken = 100%
296 branch 2 taken = 100%
300 branch 0 taken = 100%
301 ###### printf ("Failure\n");
302 call 0 never executed
303 branch 1 never executed
305 1 printf ("Success\n");
306 call 0 returns = 100%
309 For each basic block, a line is printed after the last line of the
310 basic block describing the branch or call that ends the basic block.
311 There can be multiple branches and calls listed for a single source
312 line if there are multiple basic blocks that end on that line. In this
313 case, the branches and calls are each given a number. There is no
314 simple way to map these branches and calls back to source constructs.
315 In general, though, the lowest numbered branch or call will correspond
316 to the leftmost construct on the source line.
318 For a branch, if it was executed at least once, then a percentage
319 indicating the number of times the branch was taken divided by the
320 number of times the branch was executed will be printed. Otherwise, the
321 message "never executed" is printed.
323 For a call, if it was executed at least once, then a percentage
324 indicating the number of times the call returned divided by the number
325 of times the call was executed will be printed. This will usually be
326 100%, but may be less for functions call `exit' or `longjmp', and thus
327 may not return everytime they are called.
329 The execution counts are cumulative. If the example program were
330 executed again without removing the `.da' file, the count for the
331 number of times each line in the source was executed would be added to
332 the results of the previous run(s). This is potentially useful in
333 several ways. For example, it could be used to accumulate data over a
334 number of program runs as part of a test verification suite, or to
335 provide more accurate long-term information over a large number of
338 The data in the `.da' files is saved immediately before the program
339 exits. For each source file compiled with -fprofile-arcs, the profiling
340 code first attempts to read in an existing `.da' file; if the file
341 doesn't match the executable (differing number of basic block counts) it
342 will ignore the contents of the file. It then adds in the new execution
343 counts and finally writes the data to the file.
346 File: gcc.info, Node: Gcov and Optimization, Next: Gcov Data Files, Prev: Invoking Gcov, Up: Gcov
348 Using `gcov' with GCC Optimization
349 ==================================
351 If you plan to use `gcov' to help optimize your code, you must first
352 compile your program with two special GNU CC options: `-fprofile-arcs
353 -ftest-coverage'. Aside from that, you can use any other GNU CC
354 options; but if you want to prove that every single line in your
355 program was executed, you should not compile with optimization at the
356 same time. On some machines the optimizer can eliminate some simple
357 code lines by combining them with other lines. For example, code like
365 can be compiled into one instruction on some machines. In this case,
366 there is no way for `gcov' to calculate separate execution counts for
367 each line because there isn't separate code for each line. Hence the
368 `gcov' output looks like this if you compiled the program with
376 The output shows that this block of code, combined by optimization,
377 executed 100 times. In one sense this result is correct, because there
378 was only one instruction representing all four of these lines. However,
379 the output does not indicate how many times the result was 0 and how
380 many times the result was 1.
383 File: gcc.info, Node: Gcov Data Files, Prev: Gcov and Optimization, Up: Gcov
385 Brief description of `gcov' data files
386 ======================================
388 `gcov' uses three files for doing profiling. The names of these
389 files are derived from the original *source* file by substituting the
390 file suffix with either `.bb', `.bbg', or `.da'. All of these files
391 are placed in the same directory as the source file, and contain data
392 stored in a platform-independent method.
394 The `.bb' and `.bbg' files are generated when the source file is
395 compiled with the GNU CC `-ftest-coverage' option. The `.bb' file
396 contains a list of source files (including headers), functions within
397 those files, and line numbers corresponding to each basic block in the
400 The `.bb' file format consists of several lists of 4-byte integers
401 which correspond to the line numbers of each basic block in the file.
402 Each list is terminated by a line number of 0. A line number of -1 is
403 used to designate that the source file name (padded to a 4-byte
404 boundary and followed by another -1) follows. In addition, a line
405 number of -2 is used to designate that the name of a function (also
406 padded to a 4-byte boundary and followed by a -2) follows.
408 The `.bbg' file is used to reconstruct the program flow graph for
409 the source file. It contains a list of the program flow arcs (possible
410 branches taken from one basic block to another) for each function which,
411 in combination with the `.bb' file, enables gcov to reconstruct the
414 In the `.bbg' file, the format is:
415 number of basic blocks for function #0 (4-byte number)
416 total number of arcs for function #0 (4-byte number)
417 count of arcs in basic block #0 (4-byte number)
418 destination basic block of arc #0 (4-byte number)
419 flag bits (4-byte number)
420 destination basic block of arc #1 (4-byte number)
421 flag bits (4-byte number)
423 destination basic block of arc #N (4-byte number)
424 flag bits (4-byte number)
425 count of arcs in basic block #1 (4-byte number)
426 destination basic block of arc #0 (4-byte number)
427 flag bits (4-byte number)
430 A -1 (stored as a 4-byte number) is used to separate each function's
431 list of basic blocks, and to verify that the file has been read
434 The `.da' file is generated when a program containing object files
435 built with the GNU CC `-fprofile-arcs' option is executed. A separate
436 `.da' file is created for each source file compiled with this option,
437 and the name of the `.da' file is stored as an absolute pathname in the
438 resulting object file. This path name is derived from the source file
439 name by substituting a `.da' suffix.
441 The format of the `.da' file is fairly simple. The first 8-byte
442 number is the number of counts in the file, followed by the counts
443 (stored as 8-byte numbers). Each count corresponds to the number of
444 times each arc in the program is executed. The counts are cumulative;
445 each time the program is executed, it attemps to combine the existing
446 `.da' files with the new counts for this invocation of the program. It
447 ignores the contents of any `.da' files whose number of arcs doesn't
448 correspond to the current program, and merely overwrites them instead.
450 All three of these files use the functions in `gcov-io.h' to store
451 integers; the functions in this header provide a machine-independent
452 mechanism for storing and retrieving data from a stream.
455 File: gcc.info, Node: Trouble, Next: Bugs, Prev: Gcov, Up: Top
457 Known Causes of Trouble with GNU CC
458 ***********************************
460 This section describes known problems that affect users of GNU CC.
461 Most of these are not GNU CC bugs per se--if they were, we would fix
462 them. But the result for a user may be like the result of a bug.
464 Some of these problems are due to bugs in other software, some are
465 missing features that are too much work to add, and some are places
466 where people's opinions differ as to what is best.
470 * Actual Bugs:: Bugs we will fix later.
471 * Installation Problems:: Problems that manifest when you install GNU CC.
472 * Cross-Compiler Problems:: Common problems of cross compiling with GNU CC.
473 * Interoperation:: Problems using GNU CC with other compilers,
474 and with certain linkers, assemblers and debuggers.
475 * External Bugs:: Problems compiling certain programs.
476 * Incompatibilities:: GNU CC is incompatible with traditional C.
477 * Fixed Headers:: GNU C uses corrected versions of system header files.
478 This is necessary, but doesn't always work smoothly.
479 * Standard Libraries:: GNU C uses the system C library, which might not be
480 compliant with the ISO/ANSI C standard.
481 * Disappointments:: Regrettable things we can't change, but not quite bugs.
482 * C++ Misunderstandings:: Common misunderstandings with GNU C++.
483 * Protoize Caveats:: Things to watch out for when using `protoize'.
484 * Non-bugs:: Things we think are right, but some others disagree.
485 * Warnings and Errors:: Which problems in your code get warnings,
486 and which get errors.
489 File: gcc.info, Node: Actual Bugs, Next: Installation Problems, Up: Trouble
491 Actual Bugs We Haven't Fixed Yet
492 ================================
494 * The `fixincludes' script interacts badly with automounters; if the
495 directory of system header files is automounted, it tends to be
496 unmounted while `fixincludes' is running. This would seem to be a
497 bug in the automounter. We don't know any good way to work around
500 * The `fixproto' script will sometimes add prototypes for the
501 `sigsetjmp' and `siglongjmp' functions that reference the
502 `jmp_buf' type before that type is defined. To work around this,
503 edit the offending file and place the typedef in front of the
506 * There are several obscure case of mis-using struct, union, and
507 enum tags that are not detected as errors by the compiler.
509 * When `-pedantic-errors' is specified, GNU C will incorrectly give
510 an error message when a function name is specified in an expression
511 involving the comma operator.
513 * Loop unrolling doesn't work properly for certain C++ programs.
514 This is a bug in the C++ front end. It sometimes emits incorrect
515 debug info, and the loop unrolling code is unable to recover from
519 File: gcc.info, Node: Installation Problems, Next: Cross-Compiler Problems, Prev: Actual Bugs, Up: Trouble
521 Installation Problems
522 =====================
524 This is a list of problems (and some apparent problems which don't
525 really mean anything is wrong) that show up during installation of GNU
528 * On certain systems, defining certain environment variables such as
529 `CC' can interfere with the functioning of `make'.
531 * If you encounter seemingly strange errors when trying to build the
532 compiler in a directory other than the source directory, it could
533 be because you have previously configured the compiler in the
534 source directory. Make sure you have done all the necessary
535 preparations. *Note Other Dir::.
537 * If you build GNU CC on a BSD system using a directory stored in a
538 System V file system, problems may occur in running `fixincludes'
539 if the System V file system doesn't support symbolic links. These
540 problems result in a failure to fix the declaration of `size_t' in
541 `sys/types.h'. If you find that `size_t' is a signed type and
542 that type mismatches occur, this could be the cause.
544 The solution is not to use such a directory for building GNU CC.
546 * In previous versions of GNU CC, the `gcc' driver program looked for
547 `as' and `ld' in various places; for example, in files beginning
548 with `/usr/local/lib/gcc-'. GNU CC version 2 looks for them in
549 the directory `/usr/local/lib/gcc-lib/TARGET/VERSION'.
551 Thus, to use a version of `as' or `ld' that is not the system
552 default, for example `gas' or GNU `ld', you must put them in that
553 directory (or make links to them from that directory).
555 * Some commands executed when making the compiler may fail (return a
556 non-zero status) and be ignored by `make'. These failures, which
557 are often due to files that were not found, are expected, and can
560 * It is normal to have warnings in compiling certain files about
561 unreachable code and about enumeration type clashes. These files'
562 names begin with `insn-'. Also, `real.c' may get some warnings
565 * Sometimes `make' recompiles parts of the compiler when installing
566 the compiler. In one case, this was traced down to a bug in
567 `make'. Either ignore the problem or switch to GNU Make.
569 * If you have installed a program known as purify, you may find that
570 it causes errors while linking `enquire', which is part of building
571 GNU CC. The fix is to get rid of the file `real-ld' which purify
572 installs--so that GNU CC won't try to use it.
574 * On GNU/Linux SLS 1.01, there is a problem with `libc.a': it does
575 not contain the obstack functions. However, GNU CC assumes that
576 the obstack functions are in `libc.a' when it is the GNU C
577 library. To work around this problem, change the
578 `__GNU_LIBRARY__' conditional around line 31 to `#if 1'.
580 * On some 386 systems, building the compiler never finishes because
581 `enquire' hangs due to a hardware problem in the motherboard--it
582 reports floating point exceptions to the kernel incorrectly. You
583 can install GNU CC except for `float.h' by patching out the
584 command to run `enquire'. You may also be able to fix the problem
585 for real by getting a replacement motherboard. This problem was
586 observed in Revision E of the Micronics motherboard, and is fixed
587 in Revision F. It has also been observed in the MYLEX MXA-33
590 If you encounter this problem, you may also want to consider
591 removing the FPU from the socket during the compilation.
592 Alternatively, if you are running SCO Unix, you can reboot and
593 force the FPU to be ignored. To do this, type `hd(40)unix auto
596 * On some 386 systems, GNU CC crashes trying to compile `enquire.c'.
597 This happens on machines that don't have a 387 FPU chip. On 386
598 machines, the system kernel is supposed to emulate the 387 when you
599 don't have one. The crash is due to a bug in the emulator.
601 One of these systems is the Unix from Interactive Systems: 386/ix.
602 On this system, an alternate emulator is provided, and it does
603 work. To use it, execute this command as super-user:
605 ln /etc/emulator.rel1 /etc/emulator
607 and then reboot the system. (The default emulator file remains
608 present under the name `emulator.dflt'.)
610 Try using `/etc/emulator.att', if you have such a problem on the
613 Another system which has this problem is Esix. We don't know
614 whether it has an alternate emulator that works.
616 On NetBSD 0.8, a similar problem manifests itself as these error
619 enquire.c: In function `fprop':
620 enquire.c:2328: floating overflow
622 * On SCO systems, when compiling GNU CC with the system's compiler,
623 do not use `-O'. Some versions of the system's compiler miscompile
626 * Sometimes on a Sun 4 you may observe a crash in the program
627 `genflags' or `genoutput' while building GNU CC. This is said to
628 be due to a bug in `sh'. You can probably get around it by running
629 `genflags' or `genoutput' manually and then retrying the `make'.
631 * On Solaris 2, executables of GNU CC version 2.0.2 are commonly
632 available, but they have a bug that shows up when compiling current
633 versions of GNU CC: undefined symbol errors occur during assembly
636 The solution is to compile the current version of GNU CC without
637 `-g'. That makes a working compiler which you can use to recompile
640 * Solaris 2 comes with a number of optional OS packages. Some of
641 these packages are needed to use GNU CC fully. If you did not
642 install all optional packages when installing Solaris, you will
643 need to verify that the packages that GNU CC needs are installed.
645 To check whether an optional package is installed, use the
646 `pkginfo' command. To add an optional package, use the `pkgadd'
647 command. For further details, see the Solaris documentation.
649 For Solaris 2.0 and 2.1, GNU CC needs six packages: `SUNWarc',
650 `SUNWbtool', `SUNWesu', `SUNWhea', `SUNWlibm', and `SUNWtoo'.
652 For Solaris 2.2, GNU CC needs an additional seventh package:
655 * On Solaris 2, trying to use the linker and other tools in
656 `/usr/ucb' to install GNU CC has been observed to cause trouble.
657 For example, the linker may hang indefinitely. The fix is to
658 remove `/usr/ucb' from your `PATH'.
660 * If you use the 1.31 version of the MIPS assembler (such as was
661 shipped with Ultrix 3.1), you will need to use the
662 -fno-delayed-branch switch when optimizing floating point code.
663 Otherwise, the assembler will complain when the GCC compiler fills
664 a branch delay slot with a floating point instruction, such as
667 * If on a MIPS system you get an error message saying "does not have
668 gp sections for all it's [sic] sectons [sic]", don't worry about
669 it. This happens whenever you use GAS with the MIPS linker, but
670 there is not really anything wrong, and it is okay to use the
671 output file. You can stop such warnings by installing the GNU
674 It would be nice to extend GAS to produce the gp tables, but they
675 are optional, and there should not be a warning about their
678 * In Ultrix 4.0 on the MIPS machine, `stdio.h' does not work with GNU
679 CC at all unless it has been fixed with `fixincludes'. This causes
680 problems in building GNU CC. Once GNU CC is installed, the
683 To work around this problem, when making the stage 1 compiler,
684 specify this option to Make:
686 GCC_FOR_TARGET="./xgcc -B./ -I./include"
688 When making stage 2 and stage 3, specify this option:
690 CFLAGS="-g -I./include"
692 * Users have reported some problems with version 2.0 of the MIPS
693 compiler tools that were shipped with Ultrix 4.1. Version 2.10
694 which came with Ultrix 4.2 seems to work fine.
696 Users have also reported some problems with version 2.20 of the
697 MIPS compiler tools that were shipped with RISC/os 4.x. The
698 earlier version 2.11 seems to work fine.
700 * Some versions of the MIPS linker will issue an assertion failure
701 when linking code that uses `alloca' against shared libraries on
702 RISC-OS 5.0, and DEC's OSF/1 systems. This is a bug in the
703 linker, that is supposed to be fixed in future revisions. To
704 protect against this, GNU CC passes `-non_shared' to the linker
705 unless you pass an explicit `-shared' or `-call_shared' switch.
707 * On System V release 3, you may get this error message while
710 ld fatal: failed to write symbol name SOMETHING
711 in strings table for file WHATEVER
713 This probably indicates that the disk is full or your ULIMIT won't
714 allow the file to be as large as it needs to be.
716 This problem can also result because the kernel parameter `MAXUMEM'
717 is too small. If so, you must regenerate the kernel and make the
718 value much larger. The default value is reported to be 1024; a
719 value of 32768 is said to work. Smaller values may also work.
721 * On System V, if you get an error like this,
723 /usr/local/lib/bison.simple: In function `yyparse':
724 /usr/local/lib/bison.simple:625: virtual memory exhausted
726 that too indicates a problem with disk space, ULIMIT, or `MAXUMEM'.
728 * Current GNU CC versions probably do not work on version 2 of the
729 NeXT operating system.
731 * On NeXTStep 3.0, the Objective C compiler does not work, due,
732 apparently, to a kernel bug that it happens to trigger. This
733 problem does not happen on 3.1.
735 * On the Tower models 4N0 and 6N0, by default a process is not
736 allowed to have more than one megabyte of memory. GNU CC cannot
737 compile itself (or many other programs) with `-O' in that much
740 To solve this problem, reconfigure the kernel adding the following
741 line to the configuration file:
745 * On HP 9000 series 300 or 400 running HP-UX release 8.0, there is a
746 bug in the assembler that must be fixed before GNU CC can be
747 built. This bug manifests itself during the first stage of
748 compilation, while building `libgcc2.a':
751 cc1: warning: `-g' option not supported on this version of GCC
752 cc1: warning: `-g1' option not supported on this version of GCC
753 ./xgcc: Internal compiler error: program as got fatal signal 11
755 A patched version of the assembler is available by anonymous ftp
756 from `altdorf.ai.mit.edu' as the file
757 `archive/cph/hpux-8.0-assembler'. If you have HP software support,
758 the patch can also be obtained directly from HP, as described in
761 This is the patched assembler, to patch SR#1653-010439, where
762 the assembler aborts on floating point constants.
764 The bug is not really in the assembler, but in the shared
765 library version of the function "cvtnum(3c)". The bug on
766 "cvtnum(3c)" is SR#4701-078451. Anyway, the attached
767 assembler uses the archive library version of "cvtnum(3c)"
768 and thus does not exhibit the bug.
770 This patch is also known as PHCO_4484.
772 * On HP-UX version 8.05, but not on 8.07 or more recent versions,
773 the `fixproto' shell script triggers a bug in the system shell.
774 If you encounter this problem, upgrade your operating system or
775 use BASH (the GNU shell) to run `fixproto'.
777 * Some versions of the Pyramid C compiler are reported to be unable
778 to compile GNU CC. You must use an older version of GNU CC for
779 bootstrapping. One indication of this problem is if you get a
780 crash when GNU CC compiles the function `muldi3' in file
783 You may be able to succeed by getting GNU CC version 1, installing
784 it, and using it to compile GNU CC version 2. The bug in the
785 Pyramid C compiler does not seem to affect GNU CC version 1.
787 * There may be similar problems on System V Release 3.1 on 386
790 * On the Intel Paragon (an i860 machine), if you are using operating
791 system version 1.0, you will get warnings or errors about
792 redefinition of `va_arg' when you build GNU CC.
794 If this happens, then you need to link most programs with the
795 library `iclib.a'. You must also modify `stdio.h' as follows:
798 #if defined(__i860__) && !defined(_VA_LIST)
807 extern int vprintf(const char *, va_list );
808 extern int vsprintf(char *, const char *, va_list );
815 These problems don't exist in operating system version 1.1.
817 * On the Altos 3068, programs compiled with GNU CC won't work unless
818 you fix a kernel bug. This happens using system versions V.2.2
819 1.0gT1 and V.2.2 1.0e and perhaps later versions as well. See the
822 * You will get several sorts of compilation and linking errors on the
823 we32k if you don't follow the special instructions. *Note
826 * A bug in the HP-UX 8.05 (and earlier) shell will cause the fixproto
827 program to report an error of the form:
829 ./fixproto: sh internal 1K buffer overflow
831 To fix this, change the first line of the fixproto script to look
837 File: gcc.info, Node: Cross-Compiler Problems, Next: Interoperation, Prev: Installation Problems, Up: Trouble
839 Cross-Compiler Problems
840 =======================
842 You may run into problems with cross compilation on certain machines,
845 * Cross compilation can run into trouble for certain machines because
846 some target machines' assemblers require floating point numbers to
847 be written as *integer* constants in certain contexts.
849 The compiler writes these integer constants by examining the
850 floating point value as an integer and printing that integer,
851 because this is simple to write and independent of the details of
852 the floating point representation. But this does not work if the
853 compiler is running on a different machine with an incompatible
854 floating point format, or even a different byte-ordering.
856 In addition, correct constant folding of floating point values
857 requires representing them in the target machine's format. (The C
858 standard does not quite require this, but in practice it is the
861 It is now possible to overcome these problems by defining macros
862 such as `REAL_VALUE_TYPE'. But doing so is a substantial amount of
863 work for each target machine. *Note Cross-compilation::.
865 * At present, the program `mips-tfile' which adds debug support to
866 object files on MIPS systems does not work in a cross compile
870 File: gcc.info, Node: Interoperation, Next: External Bugs, Prev: Cross-Compiler Problems, Up: Trouble
875 This section lists various difficulties encountered in using GNU C or
876 GNU C++ together with other compilers or with the assemblers, linkers,
877 libraries and debuggers on certain systems.
879 * Objective C does not work on the RS/6000.
881 * GNU C++ does not do name mangling in the same way as other C++
882 compilers. This means that object files compiled with one compiler
883 cannot be used with another.
885 This effect is intentional, to protect you from more subtle
886 problems. Compilers differ as to many internal details of C++
887 implementation, including: how class instances are laid out, how
888 multiple inheritance is implemented, and how virtual function
889 calls are handled. If the name encoding were made the same, your
890 programs would link against libraries provided from other
891 compilers--but the programs would then crash when run.
892 Incompatible libraries are then detected at link time, rather than
895 * Older GDB versions sometimes fail to read the output of GNU CC
896 version 2. If you have trouble, get GDB version 4.4 or later.
898 * DBX rejects some files produced by GNU CC, though it accepts
899 similar constructs in output from PCC. Until someone can supply a
900 coherent description of what is valid DBX input and what is not,
901 there is nothing I can do about these problems. You are on your
904 * The GNU assembler (GAS) does not support PIC. To generate PIC
905 code, you must use some other assembler, such as `/bin/as'.
907 * On some BSD systems, including some versions of Ultrix, use of
908 profiling causes static variable destructors (currently used only
909 in C++) not to be run.
911 * Use of `-I/usr/include' may cause trouble.
913 Many systems come with header files that won't work with GNU CC
914 unless corrected by `fixincludes'. The corrected header files go
915 in a new directory; GNU CC searches this directory before
916 `/usr/include'. If you use `-I/usr/include', this tells GNU CC to
917 search `/usr/include' earlier on, before the corrected headers.
918 The result is that you get the uncorrected header files.
920 Instead, you should use these options (when compiling C programs):
922 -I/usr/local/lib/gcc-lib/TARGET/VERSION/include -I/usr/include
924 For C++ programs, GNU CC also uses a special directory that
925 defines C++ interfaces to standard C subroutines. This directory
926 is meant to be searched *before* other standard include
927 directories, so that it takes precedence. If you are compiling
928 C++ programs and specifying include directories explicitly, use
929 this option first, then the two options above:
931 -I/usr/local/lib/g++-include
933 * On some SGI systems, when you use `-lgl_s' as an option, it gets
934 translated magically to `-lgl_s -lX11_s -lc_s'. Naturally, this
935 does not happen when you use GNU CC. You must specify all three
938 * On a Sparc, GNU CC aligns all values of type `double' on an 8-byte
939 boundary, and it expects every `double' to be so aligned. The Sun
940 compiler usually gives `double' values 8-byte alignment, with one
941 exception: function arguments of type `double' may not be aligned.
943 As a result, if a function compiled with Sun CC takes the address
944 of an argument of type `double' and passes this pointer of type
945 `double *' to a function compiled with GNU CC, dereferencing the
946 pointer may cause a fatal signal.
948 One way to solve this problem is to compile your entire program
949 with GNU CC. Another solution is to modify the function that is
950 compiled with Sun CC to copy the argument into a local variable;
951 local variables are always properly aligned. A third solution is
952 to modify the function that uses the pointer to dereference it via
953 the following function `access_double' instead of directly with
957 access_double (double *unaligned_ptr)
959 union d2i { double d; int i[2]; };
961 union d2i *p = (union d2i *) unaligned_ptr;
970 Storing into the pointer can be done likewise with the same union.
972 * On Solaris, the `malloc' function in the `libmalloc.a' library may
973 allocate memory that is only 4 byte aligned. Since GNU CC on the
974 Sparc assumes that doubles are 8 byte aligned, this may result in a
975 fatal signal if doubles are stored in memory allocated by the
976 `libmalloc.a' library.
978 The solution is to not use the `libmalloc.a' library. Use instead
979 `malloc' and related functions from `libc.a'; they do not have
982 * Sun forgot to include a static version of `libdl.a' with some
983 versions of SunOS (mainly 4.1). This results in undefined symbols
984 when linking static binaries (that is, if you use `-static'). If
985 you see undefined symbols `_dlclose', `_dlsym' or `_dlopen' when
986 linking, compile and link against the file `mit/util/misc/dlsym.c'
987 from the MIT version of X windows.
989 * The 128-bit long double format that the Sparc port supports
990 currently works by using the architecturally defined quad-word
991 floating point instructions. Since there is no hardware that
992 supports these instructions they must be emulated by the operating
993 system. Long doubles do not work in Sun OS versions 4.0.3 and
994 earlier, because the kernel emulator uses an obsolete and
995 incompatible format. Long doubles do not work in Sun OS version
996 4.1.1 due to a problem in a Sun library. Long doubles do work on
997 Sun OS versions 4.1.2 and higher, but GNU CC does not enable them
998 by default. Long doubles appear to work in Sun OS 5.x (Solaris
1001 * On HP-UX version 9.01 on the HP PA, the HP compiler `cc' does not
1002 compile GNU CC correctly. We do not yet know why. However, GNU CC
1003 compiled on earlier HP-UX versions works properly on HP-UX 9.01
1004 and can compile itself properly on 9.01.
1006 * On the HP PA machine, ADB sometimes fails to work on functions
1007 compiled with GNU CC. Specifically, it fails to work on functions
1008 that use `alloca' or variable-size arrays. This is because GNU CC
1009 doesn't generate HP-UX unwind descriptors for such functions. It
1010 may even be impossible to generate them.
1012 * Debugging (`-g') is not supported on the HP PA machine, unless you
1013 use the preliminary GNU tools (*note Installation::.).
1015 * Taking the address of a label may generate errors from the HP-UX
1016 PA assembler. GAS for the PA does not have this problem.
1018 * Using floating point parameters for indirect calls to static
1019 functions will not work when using the HP assembler. There simply
1020 is no way for GCC to specify what registers hold arguments for
1021 static functions when using the HP assembler. GAS for the PA does
1022 not have this problem.
1024 * In extremely rare cases involving some very large functions you may
1025 receive errors from the HP linker complaining about an out of
1026 bounds unconditional branch offset. This used to occur more often
1027 in previous versions of GNU CC, but is now exceptionally rare. If
1028 you should run into it, you can work around by making your
1031 * GNU CC compiled code sometimes emits warnings from the HP-UX
1032 assembler of the form:
1034 (warning) Use of GR3 when
1035 frame >= 8192 may cause conflict.
1037 These warnings are harmless and can be safely ignored.
1039 * The current version of the assembler (`/bin/as') for the RS/6000
1040 has certain problems that prevent the `-g' option in GCC from
1041 working. Note that `Makefile.in' uses `-g' by default when
1042 compiling `libgcc2.c'.
1044 IBM has produced a fixed version of the assembler. The upgraded
1045 assembler unfortunately was not included in any of the AIX 3.2
1046 update PTF releases (3.2.2, 3.2.3, or 3.2.3e). Users of AIX 3.1
1047 should request PTF U403044 from IBM and users of AIX 3.2 should
1048 request PTF U416277. See the file `README.RS6000' for more
1049 details on these updates.
1051 You can test for the presense of a fixed assembler by using the
1056 If the command exits normally, the assembler fix already is
1057 installed. If the assembler complains that "-u" is an unknown
1058 flag, you need to order the fix.
1060 * On the IBM RS/6000, compiling code of the form
1068 will cause the linker to report an undefined symbol `foo'.
1069 Although this behavior differs from most other systems, it is not a
1070 bug because redefining an `extern' variable as `static' is
1071 undefined in ANSI C.
1073 * AIX on the RS/6000 provides support (NLS) for environments outside
1074 of the United States. Compilers and assemblers use NLS to support
1075 locale-specific representations of various objects including
1076 floating-point numbers ("." vs "," for separating decimal
1077 fractions). There have been problems reported where the library
1078 linked with GCC does not produce the same floating-point formats
1079 that the assembler accepts. If you have this problem, set the
1080 LANG environment variable to "C" or "En_US".
1082 * Even if you specify `-fdollars-in-identifiers', you cannot
1083 successfully use `$' in identifiers on the RS/6000 due to a
1084 restriction in the IBM assembler. GAS supports these identifiers.
1086 * On the RS/6000, XLC version 1.3.0.0 will miscompile `jump.c'. XLC
1087 version 1.3.0.1 or later fixes this problem. You can obtain
1088 XLC-1.3.0.2 by requesting PTF 421749 from IBM.
1090 * There is an assembler bug in versions of DG/UX prior to 5.4.2.01
1091 that occurs when the `fldcr' instruction is used. GNU CC uses
1092 `fldcr' on the 88100 to serialize volatile memory references. Use
1093 the option `-mno-serialize-volatile' if your version of the
1094 assembler has this bug.
1096 * On VMS, GAS versions 1.38.1 and earlier may cause spurious warning
1097 messages from the linker. These warning messages complain of
1098 mismatched psect attributes. You can ignore them. *Note VMS
1101 * On NewsOS version 3, if you include both of the files `stddef.h'
1102 and `sys/types.h', you get an error because there are two typedefs
1103 of `size_t'. You should change `sys/types.h' by adding these
1104 lines around the definition of `size_t':
1111 * On the Alliant, the system's own convention for returning
1112 structures and unions is unusual, and is not compatible with GNU
1113 CC no matter what options are used.
1115 * On the IBM RT PC, the MetaWare HighC compiler (hc) uses a different
1116 convention for structure and union returning. Use the option
1117 `-mhc-struct-return' to tell GNU CC to use a convention compatible
1120 * On Ultrix, the Fortran compiler expects registers 2 through 5 to
1121 be saved by function calls. However, the C compiler uses
1122 conventions compatible with BSD Unix: registers 2 through 5 may be
1123 clobbered by function calls.
1125 GNU CC uses the same convention as the Ultrix C compiler. You can
1126 use these options to produce code compatible with the Fortran
1129 -fcall-saved-r2 -fcall-saved-r3 -fcall-saved-r4 -fcall-saved-r5
1131 * On the WE32k, you may find that programs compiled with GNU CC do
1132 not work with the standard shared C library. You may need to link
1133 with the ordinary C compiler. If you do so, you must specify the
1136 -L/usr/local/lib/gcc-lib/we32k-att-sysv/2.7.1 -lgcc -lc_s
1138 The first specifies where to find the library `libgcc.a' specified
1139 with the `-lgcc' option.
1141 GNU CC does linking by invoking `ld', just as `cc' does, and there
1142 is no reason why it *should* matter which compilation program you
1143 use to invoke `ld'. If someone tracks this problem down, it can
1144 probably be fixed easily.
1146 * On the Alpha, you may get assembler errors about invalid syntax as
1147 a result of floating point constants. This is due to a bug in the
1148 C library functions `ecvt', `fcvt' and `gcvt'. Given valid
1149 floating point numbers, they sometimes print `NaN'.
1151 * On Irix 4.0.5F (and perhaps in some other versions), an assembler
1152 bug sometimes reorders instructions incorrectly when optimization
1153 is turned on. If you think this may be happening to you, try
1154 using the GNU assembler; GAS version 2.1 supports ECOFF on Irix.
1156 Or use the `-noasmopt' option when you compile GNU CC with itself,
1157 and then again when you compile your program. (This is a temporary
1158 kludge to turn off assembler optimization on Irix.) If this
1159 proves to be what you need, edit the assembler spec in the file
1160 `specs' so that it unconditionally passes `-O0' to the assembler,
1161 and never passes `-O2' or `-O3'.