source: trunk/third/gcc/gcc.info-11 @ 11288

This is Info file gcc.info, produced by Makeinfo version 1.67 from the
input file gcc.texi.

   This file documents the use and the internals of the GNU compiler.

   Published by the Free Software Foundation 59 Temple Place - Suite 330
Boston, MA 02111-1307 USA

   Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998
Free Software Foundation, Inc.

   Permission is granted to make and distribute verbatim copies of this
manual provided the copyright notice and this permission notice are
preserved on all copies.

   Permission is granted to copy and distribute modified versions of
this manual under the conditions for verbatim copying, provided also
that the sections entitled "GNU General Public License," "Funding for
Free Software," and "Protect Your Freedom--Fight `Look And Feel'" are
included exactly as in the original, and provided that the entire
resulting derived work is distributed under the terms of a permission
notice identical to this one.

   Permission is granted to copy and distribute translations of this
manual into another language, under the above conditions for modified
versions, except that the sections entitled "GNU General Public
License," "Funding for Free Software," and "Protect Your Freedom--Fight
`Look And Feel'", and this permission notice, may be included in
translations approved by the Free Software Foundation instead of in the
original English.


File: gcc.info,  Node: C++ Signatures,  Prev: Template Instantiation,  Up: C++ Extensions

Type Abstraction using Signatures
=================================

   In GNU C++, you can use the keyword `signature' to define a
completely abstract class interface as a datatype.  You can connect this
abstraction with actual classes using signature pointers.  If you want
to use signatures, run the GNU compiler with the `-fhandle-signatures'
command-line option.  (With this option, the compiler reserves a second
keyword `sigof' as well, for a future extension.)

   Roughly, signatures are type abstractions or interfaces of classes.
Some other languages have similar facilities.  C++ signatures are
related to ML's signatures, Haskell's type classes, definition modules
in Modula-2, interface modules in Modula-3, abstract types in Emerald,
type modules in Trellis/Owl, categories in Scratchpad II, and types in
POOL-I.  For a more detailed discussion of signatures, see `Signatures:
A Language Extension for Improving Type Abstraction and Subtype
Polymorphism in C++' by Gerald Baumgartner and Vincent F. Russo (Tech
report CSD-TR-95-051, Dept. of Computer Sciences, Purdue University,
August 1995, a slightly improved version appeared in
*Software--Practice & Experience*, 25(8), pp. 863-889, August 1995).
You can get the tech report by anonymous FTP from `ftp.cs.purdue.edu'
in `pub/gb/Signature-design.ps.gz'.

   Syntactically, a signature declaration is a collection of member
function declarations and nested type declarations.  For example, this
signature declaration defines a new abstract type `S' with member
functions `int foo ()' and `int bar (int)':

     signature S
     {
       int foo ();
       int bar (int);
     };

   Since signature types do not include implementation definitions, you
cannot write an instance of a signature directly.  Instead, you can
define a pointer to any class that contains the required interfaces as a
"signature pointer".  Such a class "implements" the signature type.

   To use a class as an implementation of `S', you must ensure that the
class has public member functions `int foo ()' and `int bar (int)'.
The class can have other member functions as well, public or not; as
long as it offers what's declared in the signature, it is suitable as
an implementation of that signature type.

   For example, suppose that `C' is a class that meets the requirements
of signature `S' (`C' "conforms to" `S').  Then

     C obj;
     S * p = &obj;

defines a signature pointer `p' and initializes it to point to an
object of type `C'.  The member function call `int i = p->foo ();'
executes `obj.foo ()'.

   Abstract virtual classes provide somewhat similar facilities in
standard C++.  There are two main advantages to using signatures
instead:

  1. Subtyping becomes independent from inheritance.  A class or
     signature type `T' is a subtype of a signature type `S'
     independent of any inheritance hierarchy as long as all the member
     functions declared in `S' are also found in `T'.  So you can
     define a subtype hierarchy that is completely independent from any
     inheritance (implementation) hierarchy, instead of being forced to
     use types that mirror the class inheritance hierarchy.

  2. Signatures allow you to work with existing class hierarchies as
     implementations of a signature type.  If those class hierarchies
     are only available in compiled form, you're out of luck with
     abstract virtual classes, since an abstract virtual class cannot
     be retrofitted on top of existing class hierarchies.  So you would
     be required to write interface classes as subtypes of the abstract
     virtual class.

   There is one more detail about signatures.  A signature declaration
can contain member function *definitions* as well as member function
declarations.  A signature member function with a full definition is
called a *default implementation*; classes need not contain that
particular interface in order to conform.  For example, a class `C' can
conform to the signature

     signature T
     {
       int f (int);
       int f0 () { return f (0); };
     };

whether or not `C' implements the member function `int f0 ()'.  If you
define `C::f0', that definition takes precedence; otherwise, the
default implementation `S::f0' applies.


File: gcc.info,  Node: Gcov,  Next: Trouble,  Prev: C++ Extensions,  Up: Top

`gcov': a Test Coverage Program
*******************************

   `gcov' is a tool you can use in conjunction with GNU CC to test code
coverage in your programs.

   This chapter describes version 1.5 of `gcov'.

* Menu:

* Gcov Intro::                  Introduction to gcov.
* Invoking Gcov::               How to use gcov.
* Gcov and Optimization::       Using gcov with GCC optimization.
* Gcov Data Files::             The files used by gcov.


File: gcc.info,  Node: Gcov Intro,  Next: Invoking Gcov,  Up: Gcov

Introduction to `gcov'
======================

   `gcov' is a test coverage program.  Use it in concert with GNU CC to
analyze your programs to help create more efficient, faster running
code.  You can use `gcov' as a profiling tool to help discover where
your optimization efforts will best affect your code.  You can also use
`gcov' along with the other profiling tool, `gprof', to assess which
parts of your code use the greatest amount of computing time.

   Profiling tools help you analyze your code's performance.  Using a
profiler such as `gcov' or `gprof', you can find out some basic
performance statistics, such as:

   * how often each line of code executes

   * what lines of code are actually executed

   * how much computing time each section of code uses

   Once you know these things about how your code works when compiled,
you can look at each module to see which modules should be optimized.
`gcov' helps you determine where to work on optimization.

   Software developers also use coverage testing in concert with
testsuites, to make sure software is actually good enough for a release.
Testsuites can verify that a program works as expected; a coverage
program tests to see how much of the program is exercised by the
testsuite.  Developers can then determine what kinds of test cases need
to be added to the testsuites to create both better testing and a better
final product.

   You should compile your code without optimization if you plan to use
`gcov' because the optimization, by combining some lines of code into
one function, may not give you as much information as you need to look
for `hot spots' where the code is using a great deal of computer time.
Likewise, because `gcov' accumulates statistics by line (at the lowest
resolution), it works best with a programming style that places only
one statement on each line.  If you use complicated macros that expand
to loops or to other control structures, the statistics are less
helpful--they only report on the line where the macro call appears.  If
your complex macros behave like functions, you can replace them with
inline functions to solve this problem.

   `gcov' creates a logfile called `SOURCEFILE.gcov' which indicates
how many times each line of a source file `SOURCEFILE.c' has executed.
You can use these logfiles along with `gprof' to aid in fine-tuning the
performance of your programs.  `gprof' gives timing information you can
use along with the information you get from `gcov'.

   `gcov' works only on code compiled with GNU CC.  It is not
compatible with any other profiling or test coverage mechanism.


File: gcc.info,  Node: Invoking Gcov,  Next: Gcov and Optimization,  Prev: Gcov Intro,  Up: Gcov

Invoking gcov
=============

     gcov [-b] [-v] [-n] [-l] [-f] [-o directory] SOURCEFILE

`-b'
     Write branch frequencies to the output file, and write branch
     summary info to the standard output.  This option allows you to
     see how often each branch in your program was taken.

`-v'
     Display the `gcov' version number (on the standard error stream).

`-n'
     Do not create the `gcov' output file.

`-l'
     Create long file names for included source files.  For example, if
     the header file `x.h' contains code, and was included in the file
     `a.c', then running `gcov' on the file `a.c' will produce an
     output file called `a.c.x.h.gcov' instead of `x.h.gcov'.  This can
     be useful if `x.h' is included in multiple source files.

`-f'
     Output summaries for each function in addition to the file level
     summary.

`-o'
     The directory where the object files live.  Gcov will search for
     `.bb', `.bbg', and `.da' files in this directory.

   When using `gcov', you must first compile your program with two
special GNU CC options: `-fprofile-arcs -ftest-coverage'.  This tells
the compiler to generate additional information needed by gcov
(basically a flow graph of the program) and also includes additional
code in the object files for generating the extra profiling information
needed by gcov.  These additional files are placed in the directory
where the source code is located.

   Running the program will cause profile output to be generated.  For
each source file compiled with -fprofile-arcs, an accompanying `.da'
file will be placed in the source directory.

   Running `gcov' with your program's source file names as arguments
will now produce a listing of the code along with frequency of execution
for each line.  For example, if your program is called `tmp.c', this is
what you see when you use the basic `gcov' facility:

     $ gcc -fprofile-arcs -ftest-coverage tmp.c
     $ a.out
     $ gcov tmp.c
      87.50% of 8 source lines executed in file tmp.c
     Creating tmp.c.gcov.

   The file `tmp.c.gcov' contains output from `gcov'.  Here is a sample:

                     main()
                     {
                1      int i, total;
     
                1      total = 0;
     
               11      for (i = 0; i < 10; i++)
               10        total += i;
     
                1      if (total != 45)
           ######        printf ("Failure\n");
                       else
                1        printf ("Success\n");
                1    }

   When you use the `-b' option, your output looks like this:

     $ gcov -b tmp.c
      87.50% of 8 source lines executed in file tmp.c
      80.00% of 5 branches executed in file tmp.c
      80.00% of 5 branches taken at least once in file tmp.c
      50.00% of 2 calls executed in file tmp.c
     Creating tmp.c.gcov.

   Here is a sample of a resulting `tmp.c.gcov' file:

                     main()
                     {
                1      int i, total;
     
                1      total = 0;
     
               11      for (i = 0; i < 10; i++)
     branch 0 taken = 91%
     branch 1 taken = 100%
     branch 2 taken = 100%
               10        total += i;
     
                1      if (total != 45)
     branch 0 taken = 100%
           ######        printf ("Failure\n");
     call 0 never executed
     branch 1 never executed
                       else
                1        printf ("Success\n");
     call 0 returns = 100%
                1    }

   For each basic block, a line is printed after the last line of the
basic block describing the branch or call that ends the basic block.
There can be multiple branches and calls listed for a single source
line if there are multiple basic blocks that end on that line.  In this
case, the branches and calls are each given a number.  There is no
simple way to map these branches and calls back to source constructs.
In general, though, the lowest numbered branch or call will correspond
to the leftmost construct on the source line.

   For a branch, if it was executed at least once, then a percentage
indicating the number of times the branch was taken divided by the
number of times the branch was executed will be printed.  Otherwise, the
message "never executed" is printed.

   For a call, if it was executed at least once, then a percentage
indicating the number of times the call returned divided by the number
of times the call was executed will be printed.  This will usually be
100%, but may be less for functions call `exit' or `longjmp', and thus
may not return everytime they are called.

   The execution counts are cumulative.  If the example program were
executed again without removing the `.da' file, the count for the
number of times each line in the source was executed would be added to
the results of the previous run(s).  This is potentially useful in
several ways.  For example, it could be used to accumulate data over a
number of program runs as part of a test verification suite, or to
provide more accurate long-term information over a large number of
program runs.

   The data in the `.da' files is saved immediately before the program
exits.  For each source file compiled with -fprofile-arcs, the profiling
code first attempts to read in an existing `.da' file; if the file
doesn't match the executable (differing number of basic block counts) it
will ignore the contents of the file.  It then adds in the new execution
counts and finally writes the data to the file.


File: gcc.info,  Node: Gcov and Optimization,  Next: Gcov Data Files,  Prev: Invoking Gcov,  Up: Gcov

Using `gcov' with GCC Optimization
==================================

   If you plan to use `gcov' to help optimize your code, you must first
compile your program with two special GNU CC options: `-fprofile-arcs
-ftest-coverage'.  Aside from that, you can use any other GNU CC
options; but if you want to prove that every single line in your
program was executed, you should not compile with optimization at the
same time.  On some machines the optimizer can eliminate some simple
code lines by combining them with other lines.  For example, code like
this:

     if (a != b)
       c = 1;
     else
       c = 0;

can be compiled into one instruction on some machines.  In this case,
there is no way for `gcov' to calculate separate execution counts for
each line because there isn't separate code for each line.  Hence the
`gcov' output looks like this if you compiled the program with
optimization:

           100  if (a != b)
           100    c = 1;
           100  else
           100    c = 0;

   The output shows that this block of code, combined by optimization,
executed 100 times.  In one sense this result is correct, because there
was only one instruction representing all four of these lines.  However,
the output does not indicate how many times the result was 0 and how
many times the result was 1.


File: gcc.info,  Node: Gcov Data Files,  Prev: Gcov and Optimization,  Up: Gcov

Brief description of `gcov' data files
======================================

   `gcov' uses three files for doing profiling.  The names of these
files are derived from the original *source* file by substituting the
file suffix with either `.bb', `.bbg', or `.da'.  All of these files
are placed in the same directory as the source file, and contain data
stored in a platform-independent method.

   The `.bb' and `.bbg' files are generated when the source file is
compiled with the GNU CC `-ftest-coverage' option.  The `.bb' file
contains a list of source files (including headers), functions within
those files, and line numbers corresponding to each basic block in the
source file.

   The `.bb' file format consists of several lists of 4-byte integers
which correspond to the line numbers of each basic block in the file.
Each list is terminated by a line number of 0.  A line number of -1 is
used to designate that the source file name (padded to a 4-byte
boundary and followed by another -1) follows.  In addition, a line
number of -2 is used to designate that the name of a function (also
padded to a 4-byte boundary and followed by a -2) follows.

   The `.bbg' file is used to reconstruct the program flow graph for
the source file.  It contains a list of the program flow arcs (possible
branches taken from one basic block to another) for each function which,
in combination with the `.bb' file, enables gcov to reconstruct the
program flow.

   In the `.bbg' file, the format is:
             number of basic blocks for function #0 (4-byte number)
             total number of arcs for function #0 (4-byte number)
             count of arcs in basic block #0 (4-byte number)
             destination basic block of arc #0 (4-byte number)
             flag bits (4-byte number)
             destination basic block of arc #1 (4-byte number)
             flag bits (4-byte number)
             ...
             destination basic block of arc #N (4-byte number)
             flag bits (4-byte number)
             count of arcs in basic block #1 (4-byte number)
             destination basic block of arc #0 (4-byte number)
             flag bits (4-byte number)
             ...

   A -1 (stored as a 4-byte number) is used to separate each function's
list of basic blocks, and to verify that the file has been read
correctly.

   The `.da' file is generated when a program containing object files
built with the GNU CC `-fprofile-arcs' option is executed.  A separate
`.da' file is created for each source file compiled with this option,
and the name of the `.da' file is stored as an absolute pathname in the
resulting object file.  This path name is derived from the source file
name by substituting a `.da' suffix.

   The format of the `.da' file is fairly simple.  The first 8-byte
number is the number of counts in the file, followed by the counts
(stored as 8-byte numbers).  Each count corresponds to the number of
times each arc in the program is executed.  The counts are cumulative;
each time the program is executed, it attemps to combine the existing
`.da' files with the new counts for this invocation of the program.  It
ignores the contents of any `.da' files whose number of arcs doesn't
correspond to the current program, and merely overwrites them instead.

   All three of these files use the functions in `gcov-io.h' to store
integers; the functions in this header provide a machine-independent
mechanism for storing and retrieving data from a stream.


File: gcc.info,  Node: Trouble,  Next: Bugs,  Prev: Gcov,  Up: Top

Known Causes of Trouble with GNU CC
***********************************

   This section describes known problems that affect users of GNU CC.
Most of these are not GNU CC bugs per se--if they were, we would fix
them.  But the result for a user may be like the result of a bug.

   Some of these problems are due to bugs in other software, some are
missing features that are too much work to add, and some are places
where people's opinions differ as to what is best.

* Menu:

* Actual Bugs::               Bugs we will fix later.
* Installation Problems::     Problems that manifest when you install GNU CC.
* Cross-Compiler Problems::   Common problems of cross compiling with GNU CC.
* Interoperation::      Problems using GNU CC with other compilers,
                           and with certain linkers, assemblers and debuggers.
* External Bugs::       Problems compiling certain programs.
* Incompatibilities::   GNU CC is incompatible with traditional C.
* Fixed Headers::       GNU C uses corrected versions of system header files.
                           This is necessary, but doesn't always work smoothly.
* Standard Libraries::  GNU C uses the system C library, which might not be
                           compliant with the ISO/ANSI C standard.
* Disappointments::     Regrettable things we can't change, but not quite bugs.
* C++ Misunderstandings::     Common misunderstandings with GNU C++.
* Protoize Caveats::    Things to watch out for when using `protoize'.
* Non-bugs::            Things we think are right, but some others disagree.
* Warnings and Errors:: Which problems in your code get warnings,
                         and which get errors.


File: gcc.info,  Node: Actual Bugs,  Next: Installation Problems,  Up: Trouble

Actual Bugs We Haven't Fixed Yet
================================

   * The `fixincludes' script interacts badly with automounters; if the
     directory of system header files is automounted, it tends to be
     unmounted while `fixincludes' is running.  This would seem to be a
     bug in the automounter.  We don't know any good way to work around
     it.

   * The `fixproto' script will sometimes add prototypes for the
     `sigsetjmp' and `siglongjmp' functions that reference the
     `jmp_buf' type before that type is defined.  To work around this,
     edit the offending file and place the typedef in front of the
     prototypes.

   * There are several obscure case of mis-using struct, union, and
     enum tags that are not detected as errors by the compiler.

   * When `-pedantic-errors' is specified, GNU C will incorrectly give
     an error message when a function name is specified in an expression
     involving the comma operator.

   * Loop unrolling doesn't work properly for certain C++ programs.
     This is a bug in the C++ front end.  It sometimes emits incorrect
     debug info, and the loop unrolling code is unable to recover from
     this error.


File: gcc.info,  Node: Installation Problems,  Next: Cross-Compiler Problems,  Prev: Actual Bugs,  Up: Trouble

Installation Problems
=====================

   This is a list of problems (and some apparent problems which don't
really mean anything is wrong) that show up during installation of GNU
CC.

   * On certain systems, defining certain environment variables such as
     `CC' can interfere with the functioning of `make'.

   * If you encounter seemingly strange errors when trying to build the
     compiler in a directory other than the source directory, it could
     be because you have previously configured the compiler in the
     source directory.  Make sure you have done all the necessary
     preparations.  *Note Other Dir::.

   * If you build GNU CC on a BSD system using a directory stored in a
     System V file system, problems may occur in running `fixincludes'
     if the System V file system doesn't support symbolic links.  These
     problems result in a failure to fix the declaration of `size_t' in
     `sys/types.h'.  If you find that `size_t' is a signed type and
     that type mismatches occur, this could be the cause.

     The solution is not to use such a directory for building GNU CC.

   * In previous versions of GNU CC, the `gcc' driver program looked for
     `as' and `ld' in various places; for example, in files beginning
     with `/usr/local/lib/gcc-'.  GNU CC version 2 looks for them in
     the directory `/usr/local/lib/gcc-lib/TARGET/VERSION'.

     Thus, to use a version of `as' or `ld' that is not the system
     default, for example `gas' or GNU `ld', you must put them in that
     directory (or make links to them from that directory).

   * Some commands executed when making the compiler may fail (return a
     non-zero status) and be ignored by `make'.  These failures, which
     are often due to files that were not found, are expected, and can
     safely be ignored.

   * It is normal to have warnings in compiling certain files about
     unreachable code and about enumeration type clashes.  These files'
     names begin with `insn-'.  Also, `real.c' may get some warnings
     that you can ignore.

   * Sometimes `make' recompiles parts of the compiler when installing
     the compiler.  In one case, this was traced down to a bug in
     `make'.  Either ignore the problem or switch to GNU Make.

   * If you have installed a program known as purify, you may find that
     it causes errors while linking `enquire', which is part of building
     GNU CC.  The fix is to get rid of the file `real-ld' which purify
     installs--so that GNU CC won't try to use it.

   * On GNU/Linux SLS 1.01, there is a problem with `libc.a': it does
     not contain the obstack functions.  However, GNU CC assumes that
     the obstack functions are in `libc.a' when it is the GNU C
     library.  To work around this problem, change the
     `__GNU_LIBRARY__' conditional around line 31 to `#if 1'.

   * On some 386 systems, building the compiler never finishes because
     `enquire' hangs due to a hardware problem in the motherboard--it
     reports floating point exceptions to the kernel incorrectly.  You
     can install GNU CC except for `float.h' by patching out the
     command to run `enquire'.  You may also be able to fix the problem
     for real by getting a replacement motherboard.  This problem was
     observed in Revision E of the Micronics motherboard, and is fixed
     in Revision F.  It has also been observed in the MYLEX MXA-33
     motherboard.

     If you encounter this problem, you may also want to consider
     removing the FPU from the socket during the compilation.
     Alternatively, if you are running SCO Unix, you can reboot and
     force the FPU to be ignored.  To do this, type `hd(40)unix auto
     ignorefpu'.

   * On some 386 systems, GNU CC crashes trying to compile `enquire.c'.
     This happens on machines that don't have a 387 FPU chip.  On 386
     machines, the system kernel is supposed to emulate the 387 when you
     don't have one.  The crash is due to a bug in the emulator.

     One of these systems is the Unix from Interactive Systems: 386/ix.
     On this system, an alternate emulator is provided, and it does
     work.  To use it, execute this command as super-user:

          ln /etc/emulator.rel1 /etc/emulator

     and then reboot the system.  (The default emulator file remains
     present under the name `emulator.dflt'.)

     Try using `/etc/emulator.att', if you have such a problem on the
     SCO system.

     Another system which has this problem is Esix.  We don't know
     whether it has an alternate emulator that works.

     On NetBSD 0.8, a similar problem manifests itself as these error
     messages:

          enquire.c: In function `fprop':
          enquire.c:2328: floating overflow

   * On SCO systems, when compiling GNU CC with the system's compiler,
     do not use `-O'.  Some versions of the system's compiler miscompile
     GNU CC with `-O'.

   * Sometimes on a Sun 4 you may observe a crash in the program
     `genflags' or `genoutput' while building GNU CC.  This is said to
     be due to a bug in `sh'.  You can probably get around it by running
     `genflags' or `genoutput' manually and then retrying the `make'.

   * On Solaris 2, executables of GNU CC version 2.0.2 are commonly
     available, but they have a bug that shows up when compiling current
     versions of GNU CC: undefined symbol errors occur during assembly
     if you use `-g'.

     The solution is to compile the current version of GNU CC without
     `-g'.  That makes a working compiler which you can use to recompile
     with `-g'.

   * Solaris 2 comes with a number of optional OS packages.  Some of
     these packages are needed to use GNU CC fully.  If you did not
     install all optional packages when installing Solaris, you will
     need to verify that the packages that GNU CC needs are installed.

     To check whether an optional package is installed, use the
     `pkginfo' command.  To add an optional package, use the `pkgadd'
     command.  For further details, see the Solaris documentation.

     For Solaris 2.0 and 2.1, GNU CC needs six packages: `SUNWarc',
     `SUNWbtool', `SUNWesu', `SUNWhea', `SUNWlibm', and `SUNWtoo'.

     For Solaris 2.2, GNU CC needs an additional seventh package:
     `SUNWsprot'.

   * On Solaris 2, trying to use the linker and other tools in
     `/usr/ucb' to install GNU CC has been observed to cause trouble.
     For example, the linker may hang indefinitely.  The fix is to
     remove `/usr/ucb' from your `PATH'.

   * If you use the 1.31 version of the MIPS assembler (such as was
     shipped with Ultrix 3.1), you will need to use the
     -fno-delayed-branch switch when optimizing floating point code.
     Otherwise, the assembler will complain when the GCC compiler fills
     a branch delay slot with a floating point instruction, such as
     `add.d'.

   * If on a MIPS system you get an error message saying "does not have
     gp sections for all it's [sic] sectons [sic]", don't worry about
     it.  This happens whenever you use GAS with the MIPS linker, but
     there is not really anything wrong, and it is okay to use the
     output file.  You can stop such warnings by installing the GNU
     linker.

     It would be nice to extend GAS to produce the gp tables, but they
     are optional, and there should not be a warning about their
     absence.

   * In Ultrix 4.0 on the MIPS machine, `stdio.h' does not work with GNU
     CC at all unless it has been fixed with `fixincludes'.  This causes
     problems in building GNU CC.  Once GNU CC is installed, the
     problems go away.

     To work around this problem, when making the stage 1 compiler,
     specify this option to Make:

          GCC_FOR_TARGET="./xgcc -B./ -I./include"

     When making stage 2 and stage 3, specify this option:

          CFLAGS="-g -I./include"

   * Users have reported some problems with version 2.0 of the MIPS
     compiler tools that were shipped with Ultrix 4.1.  Version 2.10
     which came with Ultrix 4.2 seems to work fine.

     Users have also reported some problems with version 2.20 of the
     MIPS compiler tools that were shipped with RISC/os 4.x.  The
     earlier version 2.11 seems to work fine.

   * Some versions of the MIPS linker will issue an assertion failure
     when linking code that uses `alloca' against shared libraries on
     RISC-OS 5.0, and DEC's OSF/1 systems.  This is a bug in the
     linker, that is supposed to be fixed in future revisions.  To
     protect against this, GNU CC passes `-non_shared' to the linker
     unless you pass an explicit `-shared' or `-call_shared' switch.

   * On System V release 3, you may get this error message while
     linking:

          ld fatal: failed to write symbol name SOMETHING
           in strings table for file WHATEVER

     This probably indicates that the disk is full or your ULIMIT won't
     allow the file to be as large as it needs to be.

     This problem can also result because the kernel parameter `MAXUMEM'
     is too small.  If so, you must regenerate the kernel and make the
     value much larger.  The default value is reported to be 1024; a
     value of 32768 is said to work.  Smaller values may also work.

   * On System V, if you get an error like this,

          /usr/local/lib/bison.simple: In function `yyparse':
          /usr/local/lib/bison.simple:625: virtual memory exhausted

     that too indicates a problem with disk space, ULIMIT, or `MAXUMEM'.

   * Current GNU CC versions probably do not work on version 2 of the
     NeXT operating system.

   * On NeXTStep 3.0, the Objective C compiler does not work, due,
     apparently, to a kernel bug that it happens to trigger.  This
     problem does not happen on 3.1.

   * On the Tower models 4N0 and 6N0, by default a process is not
     allowed to have more than one megabyte of memory.  GNU CC cannot
     compile itself (or many other programs) with `-O' in that much
     memory.

     To solve this problem, reconfigure the kernel adding the following
     line to the configuration file:

          MAXUMEM = 4096

   * On HP 9000 series 300 or 400 running HP-UX release 8.0, there is a
     bug in the assembler that must be fixed before GNU CC can be
     built.  This bug manifests itself during the first stage of
     compilation, while building `libgcc2.a':

          _floatdisf
          cc1: warning: `-g' option not supported on this version of GCC
          cc1: warning: `-g1' option not supported on this version of GCC
          ./xgcc: Internal compiler error: program as got fatal signal 11

     A patched version of the assembler is available by anonymous ftp
     from `altdorf.ai.mit.edu' as the file
     `archive/cph/hpux-8.0-assembler'.  If you have HP software support,
     the patch can also be obtained directly from HP, as described in
     the following note:

          This is the patched assembler, to patch SR#1653-010439, where
          the assembler aborts on floating point constants.

          The bug is not really in the assembler, but in the shared
          library version of the function "cvtnum(3c)".  The bug on
          "cvtnum(3c)" is SR#4701-078451.  Anyway, the attached
          assembler uses the archive library version of "cvtnum(3c)"
          and thus does not exhibit the bug.

     This patch is also known as PHCO_4484.

   * On HP-UX version 8.05, but not on 8.07 or more recent versions,
     the `fixproto' shell script triggers a bug in the system shell.
     If you encounter this problem, upgrade your operating system or
     use BASH (the GNU shell) to run `fixproto'.

   * Some versions of the Pyramid C compiler are reported to be unable
     to compile GNU CC.  You must use an older version of GNU CC for
     bootstrapping.  One indication of this problem is if you get a
     crash when GNU CC compiles the function `muldi3' in file
     `libgcc2.c'.

     You may be able to succeed by getting GNU CC version 1, installing
     it, and using it to compile GNU CC version 2.  The bug in the
     Pyramid C compiler does not seem to affect GNU CC version 1.

   * There may be similar problems on System V Release 3.1 on 386
     systems.

   * On the Intel Paragon (an i860 machine), if you are using operating
     system version 1.0, you will get warnings or errors about
     redefinition of `va_arg' when you build GNU CC.

     If this happens, then you need to link most programs with the
     library `iclib.a'.  You must also modify `stdio.h' as follows:
     before the lines

          #if     defined(__i860__) && !defined(_VA_LIST)
          #include <va_list.h>

     insert the line

          #if __PGC__

     and after the lines

          extern int  vprintf(const char *, va_list );
          extern int  vsprintf(char *, const char *, va_list );
          #endif

     insert the line

          #endif /* __PGC__ */

     These problems don't exist in operating system version 1.1.

   * On the Altos 3068, programs compiled with GNU CC won't work unless
     you fix a kernel bug.  This happens using system versions V.2.2
     1.0gT1 and V.2.2 1.0e and perhaps later versions as well.  See the
     file `README.ALTOS'.

   * You will get several sorts of compilation and linking errors on the
     we32k if you don't follow the special instructions.  *Note
     Configurations::.

   * A bug in the HP-UX 8.05 (and earlier) shell will cause the fixproto
     program to report an error of the form:

          ./fixproto: sh internal 1K buffer overflow

     To fix this, change the first line of the fixproto script to look
     like:

          #!/bin/ksh


File: gcc.info,  Node: Cross-Compiler Problems,  Next: Interoperation,  Prev: Installation Problems,  Up: Trouble

Cross-Compiler Problems
=======================

   You may run into problems with cross compilation on certain machines,
for several reasons.

   * Cross compilation can run into trouble for certain machines because
     some target machines' assemblers require floating point numbers to
     be written as *integer* constants in certain contexts.

     The compiler writes these integer constants by examining the
     floating point value as an integer and printing that integer,
     because this is simple to write and independent of the details of
     the floating point representation.  But this does not work if the
     compiler is running on a different machine with an incompatible
     floating point format, or even a different byte-ordering.

     In addition, correct constant folding of floating point values
     requires representing them in the target machine's format.  (The C
     standard does not quite require this, but in practice it is the
     only way to win.)

     It is now possible to overcome these problems by defining macros
     such as `REAL_VALUE_TYPE'.  But doing so is a substantial amount of
     work for each target machine.  *Note Cross-compilation::.

   * At present, the program `mips-tfile' which adds debug support to
     object files on MIPS systems does not work in a cross compile
     environment.


File: gcc.info,  Node: Interoperation,  Next: External Bugs,  Prev: Cross-Compiler Problems,  Up: Trouble

Interoperation
==============

   This section lists various difficulties encountered in using GNU C or
GNU C++ together with other compilers or with the assemblers, linkers,
libraries and debuggers on certain systems.

   * Objective C does not work on the RS/6000.

   * GNU C++ does not do name mangling in the same way as other C++
     compilers.  This means that object files compiled with one compiler
     cannot be used with another.

     This effect is intentional, to protect you from more subtle
     problems.  Compilers differ as to many internal details of C++
     implementation, including: how class instances are laid out, how
     multiple inheritance is implemented, and how virtual function
     calls are handled.  If the name encoding were made the same, your
     programs would link against libraries provided from other
     compilers--but the programs would then crash when run.
     Incompatible libraries are then detected at link time, rather than
     at run time.

   * Older GDB versions sometimes fail to read the output of GNU CC
     version 2.  If you have trouble, get GDB version 4.4 or later.

   * DBX rejects some files produced by GNU CC, though it accepts
     similar constructs in output from PCC.  Until someone can supply a
     coherent description of what is valid DBX input and what is not,
     there is nothing I can do about these problems.  You are on your
     own.

   * The GNU assembler (GAS) does not support PIC.  To generate PIC
     code, you must use some other assembler, such as `/bin/as'.

   * On some BSD systems, including some versions of Ultrix, use of
     profiling causes static variable destructors (currently used only
     in C++) not to be run.

   * Use of `-I/usr/include' may cause trouble.

     Many systems come with header files that won't work with GNU CC
     unless corrected by `fixincludes'.  The corrected header files go
     in a new directory; GNU CC searches this directory before
     `/usr/include'.  If you use `-I/usr/include', this tells GNU CC to
     search `/usr/include' earlier on, before the corrected headers.
     The result is that you get the uncorrected header files.

     Instead, you should use these options (when compiling C programs):

          -I/usr/local/lib/gcc-lib/TARGET/VERSION/include -I/usr/include

     For C++ programs, GNU CC also uses a special directory that
     defines C++ interfaces to standard C subroutines.  This directory
     is meant to be searched *before* other standard include
     directories, so that it takes precedence.  If you are compiling
     C++ programs and specifying include directories explicitly, use
     this option first, then the two options above:

          -I/usr/local/lib/g++-include

   * On some SGI systems, when you use `-lgl_s' as an option, it gets
     translated magically to `-lgl_s -lX11_s -lc_s'.  Naturally, this
     does not happen when you use GNU CC.  You must specify all three
     options explicitly.

   * On a Sparc, GNU CC aligns all values of type `double' on an 8-byte
     boundary, and it expects every `double' to be so aligned.  The Sun
     compiler usually gives `double' values 8-byte alignment, with one
     exception: function arguments of type `double' may not be aligned.

     As a result, if a function compiled with Sun CC takes the address
     of an argument of type `double' and passes this pointer of type
     `double *' to a function compiled with GNU CC, dereferencing the
     pointer may cause a fatal signal.

     One way to solve this problem is to compile your entire program
     with GNU CC.  Another solution is to modify the function that is
     compiled with Sun CC to copy the argument into a local variable;
     local variables are always properly aligned.  A third solution is
     to modify the function that uses the pointer to dereference it via
     the following function `access_double' instead of directly with
     `*':

          inline double
          access_double (double *unaligned_ptr)
          {
            union d2i { double d; int i[2]; };
          
            union d2i *p = (union d2i *) unaligned_ptr;
            union d2i u;
          
            u.i[0] = p->i[0];
            u.i[1] = p->i[1];
          
            return u.d;
          }

     Storing into the pointer can be done likewise with the same union.

   * On Solaris, the `malloc' function in the `libmalloc.a' library may
     allocate memory that is only 4 byte aligned.  Since GNU CC on the
     Sparc assumes that doubles are 8 byte aligned, this may result in a
     fatal signal if doubles are stored in memory allocated by the
     `libmalloc.a' library.

     The solution is to not use the `libmalloc.a' library.  Use instead
     `malloc' and related functions from `libc.a'; they do not have
     this problem.

   * Sun forgot to include a static version of `libdl.a' with some
     versions of SunOS (mainly 4.1).  This results in undefined symbols
     when linking static binaries (that is, if you use `-static').  If
     you see undefined symbols `_dlclose', `_dlsym' or `_dlopen' when
     linking, compile and link against the file `mit/util/misc/dlsym.c'
     from the MIT version of X windows.

   * The 128-bit long double format that the Sparc port supports
     currently works by using the architecturally defined quad-word
     floating point instructions.  Since there is no hardware that
     supports these instructions they must be emulated by the operating
     system.  Long doubles do not work in Sun OS versions 4.0.3 and
     earlier, because the kernel emulator uses an obsolete and
     incompatible format.  Long doubles do not work in Sun OS version
     4.1.1 due to a problem in a Sun library.  Long doubles do work on
     Sun OS versions 4.1.2 and higher, but GNU CC does not enable them
     by default.  Long doubles appear to work in Sun OS 5.x (Solaris
     2.x).

   * On HP-UX version 9.01 on the HP PA, the HP compiler `cc' does not
     compile GNU CC correctly.  We do not yet know why.  However, GNU CC
     compiled on earlier HP-UX versions works properly on HP-UX 9.01
     and can compile itself properly on 9.01.

   * On the HP PA machine, ADB sometimes fails to work on functions
     compiled with GNU CC.  Specifically, it fails to work on functions
     that use `alloca' or variable-size arrays.  This is because GNU CC
     doesn't generate HP-UX unwind descriptors for such functions.  It
     may even be impossible to generate them.

   * Debugging (`-g') is not supported on the HP PA machine, unless you
     use the preliminary GNU tools (*note Installation::.).

   * Taking the address of a label may generate errors from the HP-UX
     PA assembler.  GAS for the PA does not have this problem.

   * Using floating point parameters for indirect calls to static
     functions will not work when using the HP assembler.  There simply
     is no way for GCC to specify what registers hold arguments for
     static functions when using the HP assembler.  GAS for the PA does
     not have this problem.

   * In extremely rare cases involving some very large functions you may
     receive errors from the HP linker complaining about an out of
     bounds unconditional branch offset.  This used to occur more often
     in previous versions of GNU CC, but is now exceptionally rare.  If
     you should run into it, you can work around by making your
     function smaller.

   * GNU CC compiled code sometimes emits warnings from the HP-UX
     assembler of the form:

          (warning) Use of GR3 when
            frame >= 8192 may cause conflict.

     These warnings are harmless and can be safely ignored.

   * The current version of the assembler (`/bin/as') for the RS/6000
     has certain problems that prevent the `-g' option in GCC from
     working.  Note that `Makefile.in' uses `-g' by default when
     compiling `libgcc2.c'.

     IBM has produced a fixed version of the assembler.  The upgraded
     assembler unfortunately was not included in any of the AIX 3.2
     update PTF releases (3.2.2, 3.2.3, or 3.2.3e).  Users of AIX 3.1
     should request PTF U403044 from IBM and users of AIX 3.2 should
     request PTF U416277.  See the file `README.RS6000' for more
     details on these updates.

     You can test for the presense of a fixed assembler by using the
     command

          as -u < /dev/null

     If the command exits normally, the assembler fix already is
     installed.  If the assembler complains that "-u" is an unknown
     flag, you need to order the fix.

   * On the IBM RS/6000, compiling code of the form

          extern int foo;
          
          ... foo ...
          
          static int foo;

     will cause the linker to report an undefined symbol `foo'.
     Although this behavior differs from most other systems, it is not a
     bug because redefining an `extern' variable as `static' is
     undefined in ANSI C.

   * AIX on the RS/6000 provides support (NLS) for environments outside
     of the United States.  Compilers and assemblers use NLS to support
     locale-specific representations of various objects including
     floating-point numbers ("." vs "," for separating decimal
     fractions).  There have been problems reported where the library
     linked with GCC does not produce the same floating-point formats
     that the assembler accepts.  If you have this problem, set the
     LANG environment variable to "C" or "En_US".

   * Even if you specify `-fdollars-in-identifiers', you cannot
     successfully use `$' in identifiers on the RS/6000 due to a
     restriction in the IBM assembler.  GAS supports these identifiers.

   * On the RS/6000, XLC version 1.3.0.0 will miscompile `jump.c'.  XLC
     version 1.3.0.1 or later fixes this problem.  You can obtain
     XLC-1.3.0.2 by requesting PTF 421749 from IBM.

   * There is an assembler bug in versions of DG/UX prior to 5.4.2.01
     that occurs when the `fldcr' instruction is used.  GNU CC uses
     `fldcr' on the 88100 to serialize volatile memory references.  Use
     the option `-mno-serialize-volatile' if your version of the
     assembler has this bug.

   * On VMS, GAS versions 1.38.1 and earlier may cause spurious warning
     messages from the linker.  These warning messages complain of
     mismatched psect attributes.  You can ignore them.  *Note VMS
     Install::.

   * On NewsOS version 3, if you include both of the files `stddef.h'
     and `sys/types.h', you get an error because there are two typedefs
     of `size_t'.  You should change `sys/types.h' by adding these
     lines around the definition of `size_t':

          #ifndef _SIZE_T
          #define _SIZE_T
          ACTUAL TYPEDEF HERE
          #endif

   * On the Alliant, the system's own convention for returning
     structures and unions is unusual, and is not compatible with GNU
     CC no matter what options are used.

   * On the IBM RT PC, the MetaWare HighC compiler (hc) uses a different
     convention for structure and union returning.  Use the option
     `-mhc-struct-return' to tell GNU CC to use a convention compatible
     with it.

   * On Ultrix, the Fortran compiler expects registers 2 through 5 to
     be saved by function calls.  However, the C compiler uses
     conventions compatible with BSD Unix: registers 2 through 5 may be
     clobbered by function calls.

     GNU CC uses the same convention as the Ultrix C compiler.  You can
     use these options to produce code compatible with the Fortran
     compiler:

          -fcall-saved-r2 -fcall-saved-r3 -fcall-saved-r4 -fcall-saved-r5

   * On the WE32k, you may find that programs compiled with GNU CC do
     not work with the standard shared C library.  You may need to link
     with the ordinary C compiler.  If you do so, you must specify the
     following options:

          -L/usr/local/lib/gcc-lib/we32k-att-sysv/2.8.1 -lgcc -lc_s

     The first specifies where to find the library `libgcc.a' specified
     with the `-lgcc' option.

     GNU CC does linking by invoking `ld', just as `cc' does, and there
     is no reason why it *should* matter which compilation program you
     use to invoke `ld'.  If someone tracks this problem down, it can
     probably be fixed easily.

   * On the Alpha, you may get assembler errors about invalid syntax as
     a result of floating point constants.  This is due to a bug in the
     C library functions `ecvt', `fcvt' and `gcvt'.  Given valid
     floating point numbers, they sometimes print `NaN'.

   * On Irix 4.0.5F (and perhaps in some other versions), an assembler
     bug sometimes reorders instructions incorrectly when optimization
     is turned on.  If you think this may be happening to you, try
     using the GNU assembler; GAS version 2.1 supports ECOFF on Irix.

     Or use the `-noasmopt' option when you compile GNU CC with itself,
     and then again when you compile your program.  (This is a temporary
     kludge to turn off assembler optimization on Irix.)  If this
     proves to be what you need, edit the assembler spec in the file
     `specs' so that it unconditionally passes `-O0' to the assembler,
     and never passes `-O2' or `-O3'.