Internet Info 1997 December/Internet_Info_CD-ROM_Walnut_Creek_December_1997.iso/faqs/gnu/gpp/help/FAQ_for_g++_and_libg++,_plain_text_version__[Revised_01_Oct_1997]

Internet Message Format | 1997-10-02 | 73KB

Path: senator-bedfellow.mit.edu!bloom-beacon.mit.edu!nntprelay.mathworks.com!newsfeed.internetmci.com!199.0.154.56!ais.net!uunet!in5.uu.net!aesop.synopsys.com!news.synopsys.com!jbuck
From: [e-mail] (Joe Buck)
Newsgroups: gnu.g++.help,comp.lang.c++,news.answers,comp.answers
Subject: FAQ for g++ and libg++, plain text version  [Revised 01 Oct 1997]
Supersedes: <[e-mail]>
Followup-To: poster
Date: 1 Oct 1997 13:00:08 GMT
Organization: Synopsys, Inc.
Lines: 1682
Approved: [e-mail]
Expires: 1 Nov 1997 00:00:00 GMT
Message-ID: <[e-mail]>
NNTP-Posting-Host: atrus.synopsys.com
Originator: jbuck@atrus
Xref: senator-bedfellow.mit.edu gnu.g++.help:18645 comp.lang.c++:284086 news.answers:113565 comp.answers:28319

Archive-name: g++-FAQ/plain
Last-modified: 01 Oct 1997
Frequency: bimonthly

[ this is the plain text version, the parent is the texinfo version ]


FAQ for g++ and libg++, by Joe Buck ([e-mail])
********************************************************

   This is a list of frequently asked questions (FAQ) for g++ users;
thanks to all those who sent suggestions for improvements.  Thanks to
Marcus Speh for doing the index.  A hypertext version is available on
the World Wide Web at `http://www.cygnus.com/misc/g++FAQ_toc.html'.

   This document has just been reorganized a bit.  There is some new
information about upcoming g++ releases and egcs; more needs to be done
but that will need to wait for next time.  A diff would look
misleadingly large, since I blew away and rebuilt the texinfo menus.

   Please send updates and corrections to the FAQ to
`[e-mail]'.  Please do *not* use me as a resource to get your
questions answered; that's what `gnu.g++.help' is for and I don't have
the time to support the net's use of g++.

   Many FAQs, including this one, are available on the archive site
"rtfm.mit.edu"; see
`ftp://rtfm.mit.edu/pub/usenet/news.answers'.  This FAQ may be found in
the subdirectory g++-FAQ.

   This FAQ is intended to supplement, not replace, Marshall Cline's
excellent FAQ for the C++ language and for the newsgroup
`comp.lang.c++'.  Especially if g++ is the first C++ compiler you've
ever used, the question "How do I do <X> with g++?" is probably really
"How do I do <X> in C++?".  You can find this FAQ at
`ftp://rtfm.mit.edu/pub/usenet/comp.lang.c++', or in HTML form at
`http://www.cerfnet.com/~mpcline/On-Line-C++-FAQs/'.

The basics: what is g++?
************************

   g++ is the traditional nickname of GNU C++, a freely redistributable
C++ compiler produced by the Free Software Foundation plus dozens of
skilled volunteers.  I say "traditional nickname" because the GNU
compiler suite, gcc, bundles together compilers for C, Objective-C, and
C++ in one package.

   While the source code to gcc/g++ can be downloaded for free, it is
not public domain, but is protected by the GNU Public License, or GPL
(*note legalities::.).

What is the latest version of gcc, g++, and libg++?
===================================================

   The current version of gcc/g++ is 2.7.2.3, released August 20, 1997.
Although that looks very recent, the only change is a minor patch to
resolve a problem with Linux and the GNU C library; users not interested
in that functionality have no reason to upgrade.

   The current version of libg++ is 2.7.2, released July 4, 1996.  The
last release of gcc/g++ with improvements to the C++ front end was
2.7.2, released Nov. 25, 1995, nearly two years ago.

   I would strongly recommend that anyone using a g++ version earlier
than 2.7.2 should upgrade if at all possible (*note version 2.7.x::.).

   For some non-Unix platforms, the latest port of gcc may be an earlier
version (2.6.3, say).  You'll need to use a version of libg++ that has
the same first two digits as the compiler version, e.g. use libg++
2.6.x (for the latest x you can find) with gcc version 2.6.3.

   The latest "1.x" version of gcc is 1.42, and the latest "1.x"
version of g++ is 1.42.0.  While gcc 1.42 is quite usable for C
programs, I recommend against using g++ 1.x except in special
circumstances (and I can't think of any such circumstances).

How do I get a copy of g++ for Unix?
====================================

   First, you may already have it if you have gcc for your platform;
g++ and gcc are combined now (as of gcc version 2.0).

   You can get g++ from a friend who has a copy, by anonymous FTP or
UUCP, or by ordering a tape or CD-ROM from the Free Software Foundation.

   The Free Software Foundation is a nonprofit organization that
distributes software and manuals to raise funds for more GNU
development.  Getting your copy from the FSF contributes directly to
paying staff to develop GNU software.  CD-ROMs cost $400 if an
organization is buying, or $100 if an individual is buying.  Tapes cost
around $200 depending on media type.  I recommend asking for version 2,
not version 1, of g++.

   For more information about ordering from the FSF, contact
[e-mail], phone [phone] or anonymous ftp file
`ftp://prep.ai.mit.edu/pub/gnu/GNUinfo/ORDERS' (you can also use one of
the sites listed below if you can't get into "prep").

   Here is a list of anonymous FTP archive sites for GNU software.  If
no directory is given, look in `/pub/gnu'.

     ASIA: ftp.cs.titech.ac.jp, tron.um.u-tokyo.ac.jp:/pub/GNU/prep
     cair-archive.kaist.ac.kr, ftp.nectec.or.th:/pub/mirrors/gnu

     AUSTRALIA: archie.au:/gnu (archie.oz or archie.oz.au for ACSnet)

     AFRICA: ftp.sun.ac.za

     MIDDLE-EAST: ftp.technion.ac.il:/pub/unsupported/gnu

     EUROPE: irisa.irisa.fr, ftp.univ-lyon1.fr,
     ftp.mcc.ac.uk, unix.hensa.ac.uk:/mirrors/uunet/systems/gnu,
     src.doc.ic.ac.uk:/gnu, ftp.ieunet.ie, ftp.eunet.ch,
     nic.switch.ch:/mirror/gnu, ftp.informatik.rwth-aachen.de,
     ftp.informatik.tu-muenchen.de, ftp.win.tue.nl, ftp.nl.net,
     ftp.etsimo.uniovi.es, ftp.funet.fi, ftp.denet.dk,
     ftp.stacken.kth.se, isy.liu.se, ftp.luth.se:/pub/unix/gnu,
     ftp.sunet.se, archive.eu.net

     SOUTH AMERICA: ftp.inf.utfsm.cl, ftp.unicamp.br

     WESTERN CANADA: ftp.cs.ubc.ca:/mirror2/gnu

     USA: wuarchive.wustl.edu:/systems/gnu, labrea.stanford.edu,
     ftp.digex.net, ftp.kpc.com:/pub/mirror/gnu, f.ms.uky.edu:/pub3/gnu,
     jaguar.utah.edu:/gnustuff, ftp.hawaii.edu:/mirrors/gnu,
     uiarchive.cso.uiuc.edu, ftp.cs.columbia.edu:/archives/gnu/prep,
     gatekeeper.dec.com:/pub/GNU, ftp.uu.net:/systems/gnu

   The "official site" is prep.ai.mit.edu, but your transfer will
probably go faster if you use one of the above machines.

   Most GNU utilities are compressed with "gzip", the GNU compression
utility.  All GNU archive sites should have a copy of this program,
which you will need to uncompress the distributions.

   Don't forget to retrieve libg++ as well!

Getting gcc/g++ for the HP Precision Architecture
=================================================

   If you use the HP Precision Architecture (HP-9000/7xx and
HP-9000/8xx) and you want to use debugging, you'll need to use the GNU
assembler, GAS (version 2.3 or later).  If you build from source, you
must tell the configure program that you are using GAS or you won't get
debugging support.  A non-standard debug format is used, since until
recently HP considered their debug format a trade secret.  Thanks to
the work of lots of good folks both inside and outside HP, the company
has seen the error of its ways and has now released the required
information.  The team at the University of Utah that did the gcc port
now has code that understands the native HP format.

   There are binaries for GNU tools in `ftp://jaguar.cs.utah.edu/dist/',
but these are older versions.

   Jeff Law has left the University of Utah, so the Utah prebuilt
binaries may be discontinued.

Getting gcc/g++ binaries for Solaris 2.x
========================================

   "Sun took the C compiler out of Solaris 2.x.  Am I stuck?"

   You'll need to get prebuilt binaries from someone.

   It used to be that you could get GCC binaries from prep.ai.mit.edu;
these are no longer there.

   The WWW site `http://smc.vnet.net/solaris_2.5.html' contains various
GNU and freeware programs for Solaris2.5 running on the sparc. These are
packaged to enable easy installation using the Solaris "pkgadd" utility.
These include GNU emacs, gcc, gdb, perl, and others.  These versions
are more recent than the binaries at "prep" (gcc 2.7.2 and libg++ 2.7.1
are there).

How do I get a copy of g++ for (some other platform)?
=====================================================

   As of gcc-2.7.x, there is Windows NT support in gcc.  Some special
utilities are required.  See the INSTALL file from the distribution.
If you're interested in GNU tools on Windows NT, see
`http://www.cygnus.com/misc/gnu-win32/' on the WWW, or the anonymous
FTP directory `ftp://ftp.cygnus.com/pub/gnu-win32/'.

   The standard gcc/g++ distribution includes VMS support for the Vax.
Since the FSF people don't use VMS, it's likely to be somewhat less
solid than the Unix version.  Precompiled copies of g++ and libg++ in
VMS-installable form for the Vax are available by FTP from
`ftp://mango.rsmas.miami.edu/pub/VMS-gcc/'.

   Klaus Kaempf ([e-mail]) has done a port to OpenVMS for the
Alpha; this is not yet a part of the official gcc/g++.  The port
includes g++ and all libraries from the libg++ distribution.  See
`http://www.progis.de' for more details.

   There are two different versions of gcc/g++ for MS-DOS: EMX and
DJGPP.  EMX also works for OS/2 and is described later.  DJGPP is DJ
Delorie's port.  It can be found on many FTP archive sites; try
`ftp://ftp.coast.net/SimTel/vendors/djgpp/' or, for a complete list, see
`http://www.delorie.com/djgpp/getting.html'.

   The latest version of DJGPP is 2.00.  See
`http://www.delorie.com/djgpp/v2/' for information on this version.

   FSF sells floppies with DJGPP on them; see above for ordering
software from the FSF.

   DJGPP has its own newsgroup: `comp.os.msdos.djgpp'.

   Development and porting efforts for GNU tools, including gcc/g++, for
the Amiga are maintained by an initiative named ADE (Amiga Developers
Environment.  More information about ADE is available at
`http://www.ninemoons.com/'.

   For more information on Amiga ports of gcc/g++, retrieve the file
`ftp://prep.ai.mit.edu/pub/gnu/MicrosPorts/Amiga'.

   A port of gcc to the Atari ST can be found at
`ftp://atari.archive.umich.edu/atari/Gnustuff/Tos' along with many
other GNU programs.  This version is usually the same as the latest FSF
release.  See the "Software FAQ" for the Usenet group
`comp.sys.atari.st' for more information.

   EMX is a port of gcc to OS/2; it can also be used on MS-DOS.  In
addition to the compiler port, the EMX port's C library attempts to
provide a Unix-like environment.  For more information ask around on
`comp.os.os2.programmer.porting'.  Version 0.9c, based on gcc-2.7.2.1,
was released in November 1996.  It is available by FTP and the WWW
from, among other places

     `http://www.os2ss.com/unix/emx09c/'
     `ftp://ftp.cdrom.com/pub/os2/emx09c/' (US)
     `ftp://ftp.leo.org/pub/comp/os/os2/leo/devtools/emx+gcc/' (Germany)

   Eberhard Mattes did the EMX port.  His address is
[e-mail].  Read the FAQ file included
with the distribution before harrassing the author.

   I'm looking for more information on gcc/g++ support on the Apple
Macintosh.  Until recently, this FAQ did not provide such information,
but FSF is no longer boycotting Apple as the League for Programming
Freedom boycott has been dropped.

   Versions 1.37.1 and 2.3.3 of gcc were ported by Stan Shebs and are
available at
`ftp://ftp.cygnus.com/pub/mac'

   They are both interfaced to MPW.  Stan is working on a version using
the current (post-2.7) sources, contact him directly ([e-mail])
for more information.

But I can only find g++-1.42!
=============================

   "I keep hearing people talking about g++ 2.7.2 (or some other number
starting with 2), but the latest version I can find is g++ 1.42.  Where
is it?"

   As of gcc 2.0, C, C++, and Objective-C as well are all combined into
a single distribution called gcc.  If you get gcc you already have g++.
The standard installation procedure for any gcc version 2 compiler will
install the C++ compiler as well.

   One could argue that we shouldn't even refer to "g++-2.x.y" but it's
a convention.  It means "the C++ compiler included with gcc-2.x.y."

Installation Issues and Problems
********************************

I can't build g++ 1.x.y with gcc-2.x.y!
=======================================

   "I obtained gcc-2.x.y and g++ 1.x.y and I'm trying to build it, but
I'm having major problems.  What's going on?"

   If you wish to build g++-1.42, you must obtain gcc-1.42 first.  The
installation instructions for g++ version 1 leave a lot to be desired,
unfortunately, and I would recommend that, unless you have a special
reason for needing the 1.x compiler, that C++ users use the latest
g++-2.x version, as it is the version that is being actively maintained.

   There is no template support in g++-1.x, and it is generally much
further away from the ANSI draft standard than g++-2.x is.

OK, I've obtained gcc; what else do I need?
===========================================

   First off, you'll want libg++ as you can do almost nothing without it
(unless you replace it with some other class library).

   Second, depending on your platform, you may need "GAS", the GNU
assembler, or the GNU linker (see next question).

   Finally, while it is not required, you'll almost certainly want the
GNU debugger, gdb.  The latest version is 4.16, released April 22, 1996.
Other debuggers (like dbx, for example) will normally not be able to
understand at least some of the debug information produced by g++.

Should I use the GNU linker, or should I use "collect"?
=======================================================

   First off, for novices: special measures must be taken with C++ to
arrange for the calling of constructors for global or static objects
before the execution of your program, and for the calling of
destructors at the end.  (Exception: System VR3 and System VR4 linkers,
Linux/ELF, and some other systems support user-defined segments; g++ on
these systems requires neither the GNU linker nor collect.  So if you
have such a system, the answer is that you don't need either one,
though using GNU ld does have some advantages over the native linker in
some cases).

   If you have experience with AT&T's "cfront", this function is
performed there by programs named "patch" or "munch".  With GNU C++, it
is performed either by the GNU linker or by a program known as
"collect".  The collect program is part of the gcc-2.x distribution;
you can obtain the GNU linker separately as part of the "binutils"
package.  The latest version of binutils is 2.7, released July 10,
1996; 2.6 is in common use and works well.

   (To be technical, it's "collect2"; there were originally several
alternative versions of collect, and this is the one that survived).

   There are advantages and disadvantages to either choice.

   Advantages of the GNU linker:

   It's faster than using collect - collect basically runs the standard
Unix linker on your program twice, inserting some extra code after the
first pass to call the constructors.  This is a sizable time penalty
for large programs.  The GNU linker does not require this extra pass.

   GNU ld reports undefined symbols using their true names, not the
mangled names (but as of 2.7.0 so does collect).

   If there are undefined symbols, GNU ld reports which object file(s)
refer to the undefined symbol(s).  On some OSes (e.g. SunOS, Solaris)
the native linker does not do this, so you have to track down who's
referring to the missing symbols yourself.

   As of binutils version 2.2, on systems that use the so-called "a.out"
debug format (e.g. Suns running SunOS 4.x), the GNU linker compresses
the debug symbol table considerably.  The 2.7 version adds some symbol
table compression for ELF and Solaris targets.

   Advantages of collect:

   If your native linker supports shared libraries, you can use shared
libraries with collect.  This used to be a strong reason *not* to use
the GNU linker, but recent versions of GNU ld support linking with
shared libraries on many platforms, and creating shared libraries on a
few (such as Intel x86 systems that use ELF object format as well as
SunOS and Solaris).

   *Note shared libraries::

   The GNU linker has not been ported to as many platforms as g++ has,
so you may be forced to use collect.

   If you use collect, you don't need to get something extra and figure
out how to install it; the standard gcc installation procedure will do
it for you.

   I used to say at this point that I don't see a clear win for either
linking alternative, but with all the improvements in the GNU linker I
think that it is now the better choice.  Take your pick.

   If you run Linux, the only available linker is the GNU linker.

Should I use the GNU assembler, or my vendor's assembler?
=========================================================

   This depends on your platform and your decision about the GNU
linker.  For most platforms, you'll need to use GAS if you use the GNU
linker.  For some platforms, you have no choice; check the gcc
installation notes to see whether you must use GAS.  But you can
usually use the vendor's assembler if you don't use the GNU linker.

   The GNU assembler assembles faster than many native assemblers;
however, on many platforms it cannot support the local debugging format.

   It used to be that the GNU assembler couldn't handle
position-independent code on SunOS.  This is no longer true if you have
version 2.6 or newer.

   On HPUX or IRIX, you must use GAS (and configure gcc with the
`--with-gnu-as' option) to debug your programs.  GAS is strongly
recommended particularly on the HP platform because of limitations in
the HP assembler.

   The GNU assembler has been merged with the binutils distribution, so
the GNU assembler and linker are now together in this package (as of
binutils version 2.5.1).

   On Linux the assembler is the GNU assembler.

How do I build shared libraries with g++?
=========================================

   For gcc-2.7.0 and later, building C++ shared libraries should work
fine on supported platforms (HPUX 9+, IRIX 5+, DEC UNIX (formerly
OSF/1), SGI/IRIX, AIX, SunOS 4, Linux/ELF and all targets using
SVR4-style ELF shared libraries).  There are two separate issues:
building libg++ as a shared library, and making your own shared
libraries.  For libg++ it is simply a matter of giving the
`--enable-shared' option to the configure program.  When compiling your
own code for shared libraries you generally must use the `-fPIC' flag
to get position-independent code.

   If your shared library contains global or static objects with
constructors, then make sure to use `gcc -shared', not `ld', to create
the shared library.  This will make sure that any processor-specific
magic needed to execute the constructors is included.

   In theory, constructors for objects in your shared library should be
called when the library is opened (by dlopen or equivalent).  This does
not work on some platforms (e.g. SunOS4; it does work on Solaris and
ELF systems such as Linux): on the broken platforms, the constructors
are not called correctly.

   David Nilsen has suggested the following workaround:

   The thing to realize is that if you link your dynamic module with the
`-shared' flag, the collect program nicely groups all the static
ctors/dtors for you into a list and sets up a function that will call
them (Note: this means that this trick won't work if you use the GNU
linker without collect (*note use GNU linker?::.).

   The magic is knowing these function names.  Currently, they're
called:

     _GLOBAL__DI   <-- calls all module constructors
     _GLOBAL__DD   <-- calls all module destructors

   [ possibly the leading underscore will differ between platforms:
jbuck ]

   Therefore, if you make a wrapper around dlopen that looks up the
symbol `_GLOBAL__DI' (or `__GLOBAL__DI' on SunOS4 machines), and calls
it, you'll simulate getting the constructors called.

   You also need to set up the destructors to be called as well, so you
need to put a wrapper around dlclose, which will call the `_GLOBAL__DD'
function in the module when/if it's unloaded.

   Lastly, to get things 100% correct, you need to set up the
destructors to also be called if the module is not unloaded, but the
main program exits.  I do this by registering a single function with
`atexit()' that calls all the destructors left in dynamically loaded
modules.

   Check the file `README.SHLIB' from the libg++ distribution for more
about making and using shared libraries.

   A patch is needed to build shared versions of version 2.7.2 of libg++
and libstdc++ on the HP-PA architecture.  You can find the patch at
`ftp://ftp.cygnus.com/pub/g++/libg++-2.7.2-hppa-gcc-fix'.

How do I use the new repository code?
=====================================

   Because there is some disagreement about the details of the template
repository mechanism, you'll need to obtain a patch from Cygnus Support
to enable the 2.7.2 repository code.  You can obtain the patch by
anonymous FTP: `ftp://ftp.cygnus.com/pub/g++/gcc-2.7.2-repo.gz'.

   There are patches for 2.7.0 and 2.7.1 in the same directory, though
if you're going to rebuild the compiler you should use the latest one.

   If you're running NetBSD or BSDI, the Cygnus repo patch is not quite
correct.  Tim Liddelow has made an alternate version available at
`ftp://ftp.cst.com.au/pub/gcc-2.7.2-repo-bsd.gz'.

   After you've applied the patch, the `-frepo' flag will enable the
repository mechanism.  The flag works much like the existing
`-fno-implicit-templates' flag, except that auxiliary files, with an
`.rpo' extension, are built that specify what template expansions are
needed.  At link time, the (patched) collect program detects missing
templates and recompiles some of the object files so that the required
templates are expanded.

   Note that the mechanism differs from that of cfront in that template
definitions still must be visible at the point where they are to be
expanded.  No assumption is made that `foo.C' contains template
definitions corresponding to template declarations in `foo.h'.

   Jason Merrill writes: "To perform closure on a set of objects, just
try to link them together.  It will fail, but as a side effect all
needed instances will be generated in the objects."

Known bugs and problems with the repo patch
===========================================

   "The `-frepo' won't expand templated friend functions!"

   This is a known bug; currently you'll have to explicitly instantiate
friend functions when using `-frepo' due to this bug (in 2.7.0 through
2.7.2 at least).

   With earlier versions of the repo patch, there was a bug that happens
when you have given a quoted command line switch, something like

     -D'MESSAGE="hello there"'

   The repo code tries to recompile files using the same flags you
originally specified, but doesn't quote arguments that need quoting,
resulting in failures in some cases.  This is no longer a problem with
the 2.7.2 patch.

Should I use the GNU C library?
===============================

   At this point in time, no (unless you are running Linux or the GNU
Hurd system).  The GNU C library is still very young, and libg++ still
conflicts with it in some places.  Use your native C library unless you
know a lot about the gory details of libg++ and gnu-libc.  This will
probably change in the future.

Global constructors aren't being called
=======================================

   "I've installed gcc and it almost works, but constructors and
destructors for global objects and objects at file scope aren't being
called.  What did I do wrong?"

   It appears that you are running on a platform that requires you to
install either "collect2" or the GNU linker, and you have done neither.
For more information, see the section discussing the GNU linker (*note
use GNU linker?::.).

   On Solaris 2.x, you shouldn't need a collect program and GNU ld
doesn't run.  If your global constructors aren't being called, you may
need to install a patch, available from Sun, to fix your linker.  The
number of the "jumbo patch" that applies is 101409-03.  Thanks to
Russell Street ([e-mail]) for this info.

   It appears that on IRIX, the collect2 program is not being installed
by default during the installation process, though it is required; you
can install it manually by executing

     make install-collect2

   from the gcc source directory after installing the compiler.  (I'm
not certain for which versions of gcc this problem occurs, and whether
it is still present).

Strange assembler errors when linking C++ programs
==================================================

   "I've installed gcc and it seemed to go OK, but when I attempt to
link any C++ program, I'm getting strange errors from the assembler!
How can that be?"

   The messages in question might look something like

     as: "/usr/tmp/cca14605.s", line 8: error: statement syntax
     as: "/usr/tmp/cca14605.s", line 14: error: statement syntax

   (on a Sun, different on other platforms).  The important thing is
that the errors come out at the link step, *not* when a C++ file is
being compiled.

   Here's what's going on: the collect2 program uses the Unix "nm"
program to obtain a list of symbols for the global constructors and
destructors, and it builds a little assembly language module that will
permit them all to be called.  If you're seeing this symptom, you have
an old version of GNU nm somewhere on your path.  This old version
prints out symbol names in a format that the collect2 program does not
expect, so bad assembly code is generated.

   The solution is either to remove the old version of GNU nm from your
path (and that of everyone else who uses g++), or to install a newer
version (it is part of the GNU "binutils" package).  Recent versions of
GNU nm do not have this problem.

Other problems building libg++
==============================

   "I am having trouble building libg++.  Help!"

   On some platforms (for example, Ultrix), you may see errors
complaining about being unable to open dummy.o.  On other platforms
(for example, SunOS), you may see problems having to do with the type
of size_t.  The fix for these problems is to make libg++ by saying
"make CC=gcc".  According to Per Bothner, it should no longer be
necessary to specify "CC=gcc" for libg++-2.3.1 or later.

   "I built and installed libg++, but g++ can't find it.  Help!"

   The string given to `configure' that identifies your system must be
the same when you install libg++ as it was when you installed gcc.
Also, if you used the `--prefix' option to install gcc somewhere other
than `/usr/local', you must use the same value for `--prefix' when
installing libg++, or else g++ will not be able to find libg++.

   The toplevel Makefile in the libg++ 2.6.2 distribution is broken,
which along with a bug in g++ 2.6.3 causes problems linking programs
that use the libstdc++ complex classes.  A patch for this is available
from `ftp://ftp.cygnus.com//pub/g++/libg++-2.6.2-fix.gz'.

But I'm *still* having problems with `size_t'!
==============================================

   "I did all that, and I'm *still* having problems with disagreeing
definitions of size_t, SIZE_TYPE, and the type of functions like
`strlen'."

   The problem may be that you have an old version of `_G_config.h'
lying around.  As of libg++ version 2.4, `_G_config.h', since it is
platform-specific, is inserted into a different directory; most include
files are in `$prefix/lib/g++-include', but this file now lives in
`$prefix/$arch/include'.  If, after upgrading your libg++, you find that
there is an old copy of `_G_config.h' left around, remove it, otherwise
g++ will find the old one first.

Do I need to rebuild libg++ to go with my new g++?
==================================================

   "After I upgraded g++ to the latest version, I'm seeing undefined
symbols."

   or

   "If I upgrade to a new version of g++, do I need to reinstall
libg++?"

   As a rule, the first two digits of your g++ and libg++ should be the
same.  Normally when you do an upgrade in the "minor version number"
(2.5.7 to 2.5.8, say) there isn't a need to rebuild libg++, but there
have been a couple of exceptions in the past.

I want several versions of g++ and libg++ to co-exist.
======================================================

   I recommend against using the `-V' flag to make multiple versions of
gcc/g++ co-exist, unless they are different minor releases that can use
the same compiled version of libg++.  The reason is that all these
versions will try to use the same libg++ version, which usually will
not work.

   Instead, use the `--prefix' flag when configuring gcc.  Use a
different value of `--prefix' for each gcc version.  Use the same value
of `--prefix' when configuring libg++.  You can then have any number of
co-existing gcc/libg++ pairs.  Symbolic links can be used so that users
don't need to put all these different directories on their paths.

   One possible system to use is to set `--prefix' to
`/usr/local/gcc-2.x.y' for version 2.x.y of gcc, and to link whichever
version of gcc you wish to be the default into `/usr/local/bin/gcc' and
`/usr/local/bin/g++'.

Trouble installing g++ and libg++ on Linux
==========================================

   "I've downloaded the latest g++ and libg++ and I'm trying to install
them on Linux, and I'm having lots of problems."

   FSF releases of libg++ won't install on Linux unchanged, since Linux
uses are part of the libio library from libg++ for its standard C
library, only this is changed in a way that it clashes with libg++.
This means that you'll need a patched version of libg++ for it to work.

   If you want to upgrade to a new gcc/libg++ combination, the easiest
thing to do is to grab the prebuilt versions of gcc and libg++ for Linux
from `ftp://tsx-11.mit.edu/pub/linux/packages/GCC'.  Follow the
directions carefully.  If you want to build from source, you'll need a
patch for libg++; the Linux developers have named the patched libg++
version libg++-2.7.1.3 and there is a patch file in the above-named
directory.

   See `http://sunsite.unc.edu/LDP/HOWTO/GCC-HOWTO.html', the Linux GCC
HOWTO, for more on gcc/g++ and Linux.

   Linux is in the process of switching over to the GNU C library,
version 2, which will become Linux libc version 6.  Once this process is
complete, there's a good chance that the installation process on Linux
will be smoother, but only experts should try making this new library
work at this point.

Problems with g++ on Linux Slackware 3.0
========================================

   "When I try to compile the traditional Hello, world program on Linux,
the compiler can't find `iostream.h'.  What's the deal?"

   You probably have the Slackware 3.0 release.  There's an error in the
setup.  It's easy to fix, though; log in as root, and make a symbolic
link:

     ln -s /usr/lib/g++-include /usr/include/g++

The Evolution of g++
********************

   This chapter discusses the evolution of g++ and describes what can
be expected in the future.

What's new in version 2.7.x of gcc/g++
======================================

   The current version of gcc/g++ is 2.7.2.2, released February 10,
1997.  The only change between 2.7.2.1 and 2.7.2.2 is that support was
added for using the GNU C library, version 2, on Linux; users not
interested in that functionality have no reason to upgrade.  The
previous version of gcc/g++ is 2.7.2.1, released August 14, 1996.  The
current version of libg++ is 2.7.2, released July 4, 1996.

   Note that gcc 2.7.2.1 just consists of several small patches to
gcc-2.7.2.  The release is mainly intended to fix platform-specific
bugs and does not affect the C++ "front end" of the compiler (the part
that parses your C++ code).

   The 2.7.x releases represent a great deal of work on the part of the
g++ maintainers to fix outstanding bugs and move the compiler closer to
the current ANSI/ISO standards committee's working paper, including
supporting many of the new features that have been added to the
language.  I recommend that everyone read the NEWS file contained in the
distribution (and that system administrators make the file available to
their users).  I've borrowed liberally from this file here.

   If any features seem unfamiliar, you will probably want to look at
the recently-released public review copy of the C++ Working Paper.  A
new draft, dated 2 December 1996, has been released for public comment.
You can find it on the web at `http://www.cygnus.com/misc/wp/' or
`http://www.maths.warwick.ac.uk/c++/pub/wp/html/cd2/'.  See
`http://www.setech.com/x3.html' or
`http://www.maths.warwick.ac.uk/c++/pub/' to download the document in
PostScript, PDF (Adobe Acrobat), HTML, or ASCII form.

   Here are the main points:

   * As described above, the scope of variables declared in the
     initialization part of a for statement has been changed; such
     variables are now visible only in the loop body.  Use
     `-fno-for-scope' to get the old behavior.  You'll need this flag
     to build groff version 1.09, Ptolemy, and many other free software
     packages.

   * Code that does not use #pragma interface/implementation will most
     likely shrink dramatically, as g++ now only emits the vtable for a
     class in the translation unit where its first non-inline,
     non-abstract virtual function is defined.

   * Support for automatic template instantiation has *not* been enabled
     in the official distribution, due to a disagreement over design
     philosophies.  But you can get a patch from Cygnus to turn it on;
     retrieve the patch from
     `ftp://ftp.cygnus.com/pub/g++/gcc-2.7.2-repo.gz' to patch
     gcc-2.7.2 (there are also patches for earlier gcc versions).

   * *Note exceptions::

   * Support for Run-Time Type Identification has been added with
     `-frtti'.  This support is still in alpha; one major restriction
     is that any file compiled with `-frtti' must include `<typeinfo>'
     (*not* `typeinfo.h' as the NEWS file says).  Also, all C++ code
     you link with (including libg++) has to be built with `-frtti', so
     it's still tricky to use.

   * Synthesis of compiler-generated constructors, destructors and
     assignment operators is now deferred until the functions are used.

   * The parsing of expressions such as `a ? b : c = 1' has changed from
     `(a ? b : c) = 1' to `a ? b : (c = 1)'.  This is a new C/C++
     incompatibility brought to you by the ANSI/ISO standards committee.

   * The operator keywords and, and_eq, bitand, bitor, compl, not,
     not_eq, or, or_eq, xor and xor_eq are now supported.  Use `-ansi'
     or `-foperator-names' to enable them.

   * The `explicit' keyword is now supported.  `explicit' is used to
     mark constructors and type conversion operators that should not be
     used implicitly.

   * Handling of user-defined type conversion has been improved.

   * Explicit instantiation of template methods is now supported.  Also,
     `inline template class foo<int>;' can be used to emit only the
     vtable for a template class.

   * With -fcheck-new, g++ will check the return value of all calls to
     operator new, and not attempt to modify a returned null pointer.

   * collect2 now demangles linker output, and c++filt has become part
     of the gcc distribution.

   * Improvements to template instantiation: only members actually used
     are instantiated.  (Actually this is not quite true: some inline
     templates that are not successfully inlined may be expanded even
     though they are not needed).

The GNU Standard C++ Library
============================

   The GNU Standard C++ Library (also called the "GNU ANSI C++ Library"
in places in the code) is not libg++, though it is included in the
libg++ distribution.  Rather, it contains classes and functions
required by the ANSI/ISO standard.  The copyright conditions are the
same as those for for the iostreams classes; the LGPL is not used
(*note legalities::.).

   This library, libstdc++, is in the libg++ distribution in versions
2.6.2 and later.  It requires at least gcc 2.6.3 to build the
libg++-2.6.2 version; use at least gcc 2.7.0 to build the libg++ 2.7.0
version.  It contains a hacked-up version of HP's implementation of the
Standard Template Library (*note Standard Template Library::.).  I've
successfully used this Standard Template Library version to build a
number of the demos you'll see on various web pages.

   As of version 2.7.0, the streams classes are now in libstdc++
instead of libg++, and libiostream is being phased out (don't use it).
The g++ program searches this library.

   The maintainers of libg++ have de-emphasized work on the older
libg++ classes in favor of enhancing libstdc++ to cover the full
language, so while libg++ will always be available, enhancements to it
should not be expected.

What can we expect in future gcc releases?
==========================================

   A great deal of work has gone into enhancements to the C++ front
end, as well as to other aspects of the compiler.

   The next major release(s) of gcc/g++ can be expected to have the
following features:

   * A completely new template implementation, much closer to the draft
     standard.  Limitations in 2.7.2.x concerning inlining template
     functions will be eliminated.  Static template data members,
     template class member functions, partial specification, and
     default template arguments will be supported.  An instantiation
     method resembling that used in Borland C++ (instantiating
     functions possibly in multiple .o files and using weak symbols to
     link correctly) will be provided, in addition to other options.
     The SGI version of STL will be shipped with libstdc++ and will
     compile unchanged.

   * Exception handling has been re-worked; exceptions will work
     together with optimization.  Actually, there are two separate
     implementations: one based on setjmp/longjmp and designed to be
     highly portable, and one designed to be more efficient but
     requiring more processor-specific support (getting exceptions
     right has proven to be extremely difficult and has been the chief
     obstacle to getting a new release out).

   * RTTI has been re-done to work correctly and is on by default.

   * Overloading has been re-worked to conform to the latest draft of
     the standard.

   Features that are still missing include namespaces and templates as
template arguments.

What's this I hear about egcs?
==============================

   The egcs effort is a new effort to merge several threads of gcc
development and to provide a faster development process.  For more
information see `http://www.cygnus.com/egcs/'.

OK, when can I get this stuff?
==============================

   The FSF has a policy of never announcing release dates in advance.
I'm sure this is frustrating to a lot of people, since it's taken so
long, and this frustration was one of the reasons the egcs effort was
created.  An egcs release should be expected to occur in the very near
future.   [ More on this next time ].

User Problems
*************

Linker complains about missing virtual table
============================================

   "I'm getting a message complaining about an undefined virtual table.
Is this a compiler bug?"

   (On platforms that run neither collect nor the GNU linker, like
Solaris, you may see an odd undefined symbol like "_vt.3foo", where foo
is a class name).

   This is probably because you are missing a definition for the first
(non-inline) virtual function of the class.  Since gcc-2.7.0, g++ uses
a trick borrowed from cfront: the .o file containing the definition for
the first non-inline virtual function for the class will also contain
the virtual function table.

gcc-2.7.0 breaks declarations in "for" statements!
==================================================

   gcc-2.7.0 implements the new ANSI/ISO rule on the scope of variables
declared in for loops.

     for (int i = 1; i <= 10; i++) {
             // do something here
     }
     foo(i);

   In the above example, most existing C++ compilers would pass the
value 11 to the function `foo'.  In gcc 2.7 and in the ANSI/ISO working
paper, the scope of `i' is only the for loop body, so this is an error.
So that old code can be compiled, the new gcc has a flag
`-fno-for-scope' that causes the old rule to be used.

   As of 2.7.1, the compiler attempts to issue warnings about code that
has different meanings under the two sets of rules, but the code is not
perfect: the intent was that code that has valid, but different,
meanings under the ARM rules and the working paper rules would give
warnings but have the new behavior, and this doesn't seem to happen.

   The `-ffor-scope' flag under 2.7.1 and 2.7.2 gives the 2.7.0
behavior.

g++ seems to want a const constructor.  What's that?
====================================================

   gcc-2.7.1 introduced a bug that causes the compiler to ask for a
const constructor (there's no such thing in C++) in certain situations
where a const object appears in a template class.  Most cases have been
fixed in gcc-2.7.2, but unfortunately not all.  Still, if you're running
gcc-2.7.1 and have this problem, upgrade to 2.7.2; it is a vast
improvement.

   The default constructor for the template `pair' in ObjectSpace's
implementation of STL triggers the bug in one place, for gcc 2.7.2.  If
you're using ObjectSpace<STL> and having this problem, simply change
the default constructor from

     os_pair () : first (T1 ()), second (T2 ()) {}

   to just

     os_pair () {}

   Once this is done, ObjectSpace<STL> works fairly well.

How to silence "unused parameter" warnings
==========================================

   "When I use `-Wall' (or `-Wunused'), g++ warns about unused
parameters.  But the parameters have to be there, for use in derived
class functions.  How do I get g++ to stop complaining?"

   The answer is to simply omit the names of the unused parameters when
defining the function.  This makes clear, both to g++ and to readers of
your code, that the parameter is unused.  For example:

     int Foo::bar(int arg) { return 0; }

   will give a warning for the unused parameter `arg'.  To suppress the
warning write

     int Foo::bar(int) { return 0; }

g++ objects to a declaration in a case statement
================================================

   "The compiler objects to my declaring a variable in one of the
branches of a case statement.  Earlier versions used to accept this
code.  Why?"

   The draft standard does not allow a goto or a jump to a case label to
skip over an initialization of a variable or a class object.  For
example:

     switch ( i ) {
       case 1:
         Object obj(0);
             ...
         break;
       case 2:
         ...
         break;
     }

   The reason is that `obj' is also in scope in the rest of the switch
statement.

   As of version 2.7.0, the compiler will object that the jump to the
second case level crosses the initialization of `obj'.  Older compiler
versions would object only if class Object has a destructor.  In either
case, the solution is to add a set of curly braces around the case
branch:

       case 1:
         {
            Object obj(0);
             ...
            break;
         }

Where can I find a demangler?
=============================

   A g++-compatible demangler named `c++filt' can be found in the
`binutils' distribution.  This distribution (which also contains the
GNU linker) can be found at any GNU archive site.

   As of version 2.7.0, `c++filt' is included with gcc and is installed
automatically.  Even better, it is used by the `collect' linker, so you
don't see mangled symbols anymore (except on platforms that use neither
collect nor the GNU linker, like Solaris).

Linker reports undefined symbols for static data members
========================================================

   "g++ reports undefined symbols for all my static data members when I
link, even though the program works correctly for compiler XYZ.  What's
going on?"

   The problem is almost certainly that you don't give definitions for
your static data members.  If you have

     class Foo {
         ...
         void method();
         static int bar;
     };

   you have only declared that there is an int named Foo::bar and a
member function named Foo::method that is defined somewhere.  You still
need to define *both* method() and bar in some source file.  According
to the draft ANSI standard, you must supply an initializer, such as

     int Foo::bar = 0;

in one (and only one) source file.

What does "Internal compiler error" mean?
=========================================

   It means that the compiler has detected a bug in itself.
Unfortunately, g++ still has many bugs, though it is a lot better than
it used to be.  If you see this message, please send in a complete bug
report (see next section).

I think I have found a bug in g++.
==================================

   "I think I have found a bug in g++, but I'm not sure.  How do I know,
and who should I tell?"

   First, see the excellent section on bugs and bug reports in the gcc
manual (which is included in the gcc distribution).  As a short summary
of that section: if the compiler gets a fatal signal, for any input,
it's a bug (newer versions of g++ will ask you to send in a bug report
when they detect an error in themselves).  Same thing for producing
invalid assembly code.

   When you report a bug, make sure to describe your platform (the type
of computer, and the version of the operating system it is running) and
the version of the compiler that you are running.  See the output of the
command `g++ -v' if you aren't sure.  Also provide enough code so that
the g++ maintainers can duplicate your bug.  Remember that the
maintainers won't have your header files; one possibility is to send
the output of the preprocessor (use `g++ -E' to get this).  This is
what a "complete bug report" means.

   I will add some extra notes that are C++-specific, since the notes
from the gcc documentation are generally C-specific.

   First, mail your bug report to "[e-mail]".  You may
also post to `gnu.g++.bug', but it's better to use mail, particularly
if you have any doubt as to whether your news software generates
correct reply addresses.  Don't mail C++ bugs to
[e-mail].

   *News:* as I write this (late February 1996) the gateway connecting
the bug-g++ mailing list and the `gnu.g++.bug' newsgroup is
(temporarily?) broken.  Please mail, do not post bug reports.

   If your bug involves libg++ rather than the compiler, mail to
[e-mail].  If you're not sure, choose one, and if you
guessed wrong, the maintainers will forward it to the other list.

   Second, if your program does one thing, and you think it should do
something else, it is best to consult a good reference if in doubt.
The standard reference is the draft working paper from the ANSI/ISO C++
standardization committee, which you can get on the net.  For
PostScript and PDF (Adobe Acrobat) versions, see the archive at
`ftp://research.att.com/dist/stdc++/WP'.  For HTML and ASCII versions,
see `ftp://ftp.cygnus.com/pub/g++'.  On the World Wide Web, see
`http://www.cygnus.com/misc/wp/'.

   An older standard reference is "The Annotated C++ Reference Manual",
by Ellis and Stroustrup (copyright 1990, ISBN #0-201-51459-1).  This is
what they're talking about on the net when they refer to "the ARM".
But you should know that changes have been made to the language since
then.

   The ANSI/ISO C++ standards committee have adopted some changes to the
C++ language since the publication of the original ARM, and newer
versions of g++ (2.5.x and later) support some of these changes, notably
the mutable keyword (added in 2.5.0), the bool type (added in 2.6.0),
and changes in the scope of variables defined in for statements (added
in 2.7.0).  You can obtain an addendum to the ARM explaining many of
these changes by FTP from
`ftp://ftp.std.com/AW/stroustrup2e/new_iso.ps'.

   Note that the behavior of (any version of) AT&T's "cfront" compiler
is NOT the standard for the language.

Porting programs from other compilers to g++
============================================

   "I have a program that runs on <some other C++ compiler>, and I want
to get it running under g++.  Is there anything I should watch out for?"

   Note that g++ supports many of the newer keywords that have recently
been added to the language.  Your other C++ compiler may not support
them, so you may need to rename variables and members that conflict
with these keywords.

   There are two other reasons why a program that worked under one
compiler might fail under another: your program may depend on the order
of evaluation of side effects in an expression, or it may depend on the
lifetime of a temporary (you may be assuming that a temporary object
"lives" longer than the standard guarantees).  As an example of the
first:

     void func(int,int);

     int i = 3;
     func(i++,i++);

   Novice programmers think that the increments will be evaluated in
strict left-to-right order.  Neither C nor C++ guarantees this; the
second increment might happen first, for example.  func might get 3,4,
or it might get 4,3.

   The second problem often happens with classes like the libg++ String
class.  Let's say I have

     String func1();
     void func2(const char*);

   and I say

     func2(func1());

   because I know that class String has an "operator const char*".  So
what really happens is

     func2(func1().convert());

   where I'm pretending I have a convert() method that is the same as
the cast.  This is unsafe in g++ versions before 2.6.0, because the
temporary String object may be deleted after its last use (the call to
the conversion function), leaving the pointer pointing to garbage, so by
the time func2 is called, it gets an invalid argument.

   Both the cfront and the old g++ behaviors are legal according to the
ARM, but the powers that be have decided that compiler writers were
given too much freedom here.

   The ANSI C++ committee has now come to a resolution of the lifetime
of temporaries problem: they specify that temporaries should be deleted
at end-of-statement (and at a couple of other points).  This means that
g++ versions before 2.6.0 now delete temporaries too early, and cfront
deletes temporaries too late.  As of version 2.6.0, g++ does things
according to the new standard.

   For now, the safe way to write such code is to give the temporary a
name, which forces it to live until the end of the scope of the name.
For example:

     String& tmp = func1();
     func2(tmp);

   Finally, like all compilers (but especially C++ compilers, it seems),
g++ has bugs, and you may have tweaked one.  If so, please file a bug
report (after checking the above issues).

Why does g++ mangle names differently from other C++ compilers?
===============================================================

   See the answer to the next question.

Why can't g++ code link with code from other C++ compilers?
===========================================================

   "Why can't I link g++-compiled programs against libraries compiled by
some other C++ compiler?"

   Some people think that, if only the FSF and Cygnus Support folks
would stop being stubborn and mangle names the same way that, say,
cfront does, then any g++-compiled program would link successfully
against any cfront-compiled library and vice versa.  Name mangling is
the least of the problems.  Compilers differ as to how objects are laid
out, how multiple inheritance is implemented, how virtual function
calls are handled, and so on, so if the name mangling were made the
same, your programs would link against libraries provided from other
compilers but then crash when run.  For this reason, the ARM
*encourages* compiler writers to make their name mangling different
from that of other compilers for the same platform.  Incompatible
libraries are then detected at link time, rather than at run time.

What documentation exists for g++ 2.x?
======================================

   Relatively little.  While the gcc manual that comes with the
distribution has some coverage of the C++ part of the compiler, it
focuses mainly on the C compiler (though the information on the "back
end" pertains to C++ as well).  Still, there is useful information on
the command line options and the #pragma interface and #pragma
implementation directives in the manual, and there is a useful section
on template instantiation in the 2.6 version.  There is a Unix-style
manual entry, "g++.1", in the gcc-2.x distribution; the information
here is a subset of what is in the manual.

   You can buy a nicely printed and bound copy of this manual from the
FSF; see above for ordering information.

   A draft of a document describing the g++ internals appears in the gcc
distribution (called g++int.texi); it is incomplete but gives lots of
information.

   For class libraries, there are several resources available:

   * The libg++ distribution has a manual `libg++/libg++.texi'
     describing the old libg++ classes, and another manual
     `libio/iostream.texi' describing the iostreams implementation.

   * While there is no libg++-specific document describing the STL
     implementation, SGI's web site, at
     `http://www.sgi.com/Technology/STL/', is an excellent resource.

Problems with the template implementation
=========================================

   g++ does not implement a separate pass to instantiate template
functions and classes at this point; for this reason, it will not work,
for the most part, to declare your template functions in one file and
define them in another.  The compiler will need to see the entire
definition of the function, and will generate a static copy of the
function in each file in which it is used.

   (The experimental template repository code (*note repository::.) that
can be added to 2.7.0 or later does implement a separate pass, but there
is still no searching of files that the compiler never saw).

   For version 2.6.0, however, a new switch `-fno-implicit-templates'
was added; with this switch, templates are expanded only under user
control.  I recommend that all g++ users that use templates read the
section "Template Instantiation" in the gcc manual (version 2.6.x and
newer).  g++ now supports explicit template expansion using the syntax
from the latest C++ working paper:

     template class A<int>;
     template ostream& operator << (ostream&, const A<int>&);

   As of version 2.6.3, there are still a few limitations in the
template implementation besides the above (thanks to Jason Merrill for
this info): These are still present in version 2.7.2, but a new
implementation of templates planned for version 2.8 will eliminate them.

  1. Static data member templates are not supported.  You can work
     around this by explicitly declaring the static variable for each
     template specialization:

          template <class T> struct A {
            static T t;
          };

          template <class T> T A<T>::t = 0; // gets bogus error
          int A<int>::t = 0;                // OK (workaround)

     (still a limitation in 2.7.2)

  2. Template member names are not available when defining member
     function templates.

          template <class T> struct A {
            typedef T foo;
            void f (foo);
            void g (foo arg) { ... }; // this works
          };

          template <class T> void A<T>::f (foo) { } // gets bogus error

  3. Templates are instantiated using the parser.  This results in two
     problems:

     a) Class templates are instantiated in some situations where such
     instantiation should not occur.

          template <class T> class A { };
          A<int> *aip = 0; // should not instantiate A<int> (but does)

     b) Function templates cannot be inlined at the site of their
     instantiation.

          template <class T> inline T min (T a, T b) { return a < b ? a : b; }

          void f () {
            int i = min (1, 0);           // not inlined
          }

          void g () {
            int j = min (1, 0);           // inlined
          }

     A workaround that works in version 2.6.1 and later is to specify

          extern template int min (int, int);

     before `f()'; this will force it to be instantiated (though not
     emitted).

  4. Member function templates are always instantiated when their
     containing class is.  This is wrong.

I get undefined symbols when using templates
============================================

   (Thanks to Jason Merrill for this section).

   g++ does not automatically instantiate templates defined in other
files.  Because of this, code written for cfront will often produce
undefined symbol errors when compiled with g++.  You need to tell g++
which template instances you want, by explicitly instantiating them in
the file where they are defined.  For instance, given the files

   `templates.h':
     template <class T>
     class A {
     public:
       void f ();
       T t;
     };

     template <class T> void g (T a);

   `templates.cc':
     #include "templates.h"

     template <class T>
     void A<T>::f () { }

     template <class T>
     void g (T a) { }

   main.cc:
     #include "templates.h"

     main ()
     {
       A<int> a;
       a.f ();
       g (a);
     }

   compiling everything with `g++ main.cc templates.cc' will result in
undefined symbol errors for `A<int>::f ()' and `g (A<int>)'.  To fix
these errors, add the lines

     template class A<int>;
     template void g (A<int>);

   to the bottom of `templates.cc' and recompile.

I get multiply defined symbols using templates
==============================================

   You may be running into a bug that was introduced in version 2.6.1
(and is still present in 2.6.3) that generated external linkage for
templates even when neither `-fexternal-templates' nor
`-fno-implicit-templates' is specified.  There is a patch for this
problem at
`ftp://ftp.cygnus.com/pub/g++/gcc-2.6.3-template-fix'.

   I recommend either applying the patch or using
`-fno-implicit-templates' together with explicit template instantiation
as described in previous sections.

   This bug is fixed in 2.7.0.

Does g++ support the Standard Template Library?
===============================================

   The Standard Template Library (STL) uses many of the extensions that
the ANSI/ISO committee has made to templates, and g++ doesn't support
some of these yet.  So if you grab HP's free implementation of STL it
isn't going to work.  However, starting with libg++-2.6.2 libg++
contains a hacked version of STL, based on work by Carsten Bormann,
which permits g++ to compile at least the containers (thanks to Per
Bothner for this text).

   Actually, as of libg++ version 2.7.2 most of this works quite well,
most of the time; I've succeeded in making significant use of it.
Almost all of the ObjectSpace examples (a set of over 200 simple
examples of STL usage) now work.

   When version 2.8.0 is out (with its complete redesign of the template
implementation) a much more complete implementation of the STL (based
on a newer free implementation from SGI) will be included.  In the
meantime, a group at the Moscow Center for Sparc Technology has a port
of the SGI STL implementation that mostly works with gcc-2.7.2.  See
`http://www.ipmce.su/people/fbp/stl/stlport.html'.

   In addition, there are several commercial suppliers of STL
implementations; ObjectSpace's version supports gcc-2.7.x.

   Mumit Khan has produced an "STL newbie guide" with lots of
information on using STL with gcc.  See

   `http://www.xraylith.wisc.edu/~khan/software/stl/STL.newbie.html'

I'm having problems mixing STL and the standard string class
============================================================

   This is due to a bug in g++ version 2.7.2 and 2.7.2.1; the compiler
is confused by the operator declarations.  There is an easy workaround,
however; just make sure that the `<string>' header is included before
any STL headers.  That is, just say

     #include <string>

   before any other `#include' directives.

   Unfortunately, this doesn't solve all problems; you may still have
difficulty with the relational operators !=, <=, >, and >=, thanks to a
conflict with the very general definition of these operators in
function.h.  One trick that sometimes works is to try to use == and <
in your code instead of the other operators.  Another is to use a
derived class of <string>.  The only completely satisfactory solution,
I'm afraid, is to wait for the new release.

Problems and limitations with exceptions
========================================

   Recent g++ versions provide limited support for exceptions.  You must
provide the `-fhandle-exceptions' flag to enable exception handling.
As of version 2.7.2, exceptions may not work properly (and you may get
odd error messages when compiling) if you turn on optimization (the
`-O' flag).

   You must give the `-frtti' switch to enable catching of derived
exception objects with handlers for the base exception class; if
`-frtti' is not given, only exact type matching works.

   For exception handling to work with 2.7.0 your CPU must be a SPARC,
RS6000/PowerPC, 386/486/Pentium, or ARM.  Release 2.7.1 added support
for the Alpha, and "m68k is rumored to work on some platforms" and "VAX
may also work" (according to Mike Stump).  *It still doesn't work on
HP-PA or MIPS platforms.*

Does g++ support namespaces?
============================

   As of version 2.7.2, g++ recognizes the keywords `namespace' and
`using', and there is some rudimentary code present, but almost nothing
connected with namespaces works yet.  It appears that this will still
be true when 2.8.0 is released.

What are the differences between g++ and the ARM specification of C++?
======================================================================

   As of version 2.7.0, g++ has exception support on most but not all
platforms (no support on MIPS-based platforms yet), but it doesn't work
right if optimizaton is enabled, which means the exception
implementation is still not really ready for production use.

   Some features that the ANSI/ISO standardization committee has voted
in that don't appear in the ARM are supported, notably the `mutable'
keyword, in version 2.5.x.  2.6.x adds support for the built-in boolean
type `bool', with constants `true' and `false'.  The beginnings of
run-time type identification are present, so there are more reserved
words: `typeid', `static_cast', `reinterpret_cast', `const_cast', and
`dynamic_cast'.

   As with any beta-test compiler, there are bugs.  You can help improve
the compiler by submitting detailed bug reports.

   One of the weakest areas of g++ other than templates is the
resolution of overloaded functions and operators in complex cases.  The
usual symptom is that in a case where the ARM says that it is ambiguous
which function should be chosen, g++ chooses one (often the first one
declared).  This is usually not a problem when porting C++ code from
other compilers to g++, but shows up as errors when code developed under
g++ is ported to other compilers.  (I believe this is no longer a
significant problem in 2.7.0).

   [A full bug list would be very long indeed, so I won't put one here.
I may add a list of frequently-reported bugs and "non-bugs" like the
static class members issue mentioned above].

Will g++ compile InterViews?  The NIH class library?  Rogue Wave?
=================================================================

   The NIH class library uses a non-portable, compiler-dependent hack
to initialize itself, which makes life difficult for g++ users.  It
will not work without modification, and I don't know what modifications
are required or whether anyone has done them successfully.

   In short, it's not going to happen any time soon (previous FAQs
referred to patches that a new NIHCL release would hopefully contain,
but this hasn't happened).

   *Note:* I thought I saw an item indicating that someone *had*
patched NIHCL to work with g++.  Any pointers?

   I think that as of version 2.5.6, the standard g++ will compile the
standard 3.1 InterViews completely successfully.  Note that you'll need
the `-fno-for-scope' flag if you use gcc-2.7.0; with 2.7.2 you may be
able to omit this flag but you'll get warnings.

   According to Jason Merrill, gcc-2.7.0 and newer works with Rogue
Wave's `tools.h++' class library, but you may want to grab
`ftp://ftp.cygnus.com/pub/g++/Tools.h++-6.1-patch'.  Again, you'll need
the `-fno-for-scope' flag since Rogue Wave hasn't fixed their code to
comply with the new standard yet.

Debugging on SVR4 systems
=========================

   "How do I get debugging to work on my System V Release 4 system?"

   Most systems based on System V Release 4 (except Solaris) encode
symbolic debugging information in a format known as `DWARF'.

   Although the GNU C compiler already knows how to write out symbolic
debugging information in the DWARF format, the GNU C++ compiler does
not yet have this feature yet.  However, work is in progress for DWARF
2 debug support for gcc and g++ and will be available in a future
release (probably 2.8.0).

   In the meantime, you *can* get g++ debugging under SVR4 systems by
configuring gcc with the `--with-stabs' option.  This causes gcc to use
an alternate debugging format, one more like that used under SunOS4.
You won't need to do anything special to GDB; it will always understand
the "stabs" format.

debugging problems on Solaris
=============================

   "I'm on Solaris, and gdb says it doesn't know about some of my local
symbols.  Help!"

   This problem was introduced in gcc 2.7.2; debug symbols for locals
that aren't declared at the beginning of a block come out in the wrong
order, and gdb can't find such symbols.

   This problem is fixed in gcc-2.7.2.1.

X11 conflicts with libg++ in definition of String
=================================================

   "X11 and Motif define String, and this conflicts with the String
class in libg++.  How can I use both together?"

   One possible method is the following:

     #define String XString
     #include <X11/Intrinsic.h>
     /* include other X11 and Motif headers */
     #undef String

   and remember to use the correct `String' or `XString' when you
declare things later.

Why can't I assign one stream to another?
=========================================

   [ Thanks to Per Bothner and Jerry Schwarz for this section. ]

   Assigning one stream to another seems like a reasonable thing to do,
but it's a bad idea.  Usually, this comes up because people want to
assign to `cout'.  This is poor style, especially for libraries, and is
contrary to good object-oriented design.  (Libraries that write directly
to `cout' are less flexible, modular, and object-oriented).

   The iostream classes do not allow assigning to arbitrary streams,
because this can violate typing:

     ifstream foo ("foo");
     istrstream str(...);
     foo = str;
     foo->close ();  /* Oops! Not defined for istrstream! */

   The original cfront implementation of iostreams by Jerry Schwarz
allows you to assign to `cin', `cout', `cerr', and `clog', but this is
not part of the draft standard for iostreams and generally isn't
considered a good idea, so standard-conforming code shouldn't use this
technique.

   The GNU implementation of iostream did not support assigning to
`cin', `cout', `cerr', and `clog' for quite a while, but it now does,
for backward compatibility with cfront iostream (versions 2.6.1 and
later of libg++).

   The ANSI/ISO C++ Working Paper does provide ways of changing the
streambuf associated with a stream.  Assignment isn't allowed; there is
an explicit named member that must be used.

   However, it is not wise to do this, and the results are confusing.
For example: `fstream::rdbuf' is supposed to return the *original*
filebuf, not the one you assigned. (This is not yet implemented in GNU
iostream.)  This must be so because `fstream::rdbuf' is defined to
return a `filebuf *'.

What are the rules for shipping code built with g++ and libg++?
***************************************************************

   "Is it is possible to distribute programs for profit that are created
with g++ and use the g++ libraries?"

   I am not a lawyer, and this is not legal advice.  In any case, I have
little interest in telling people how to violate the spirit of the GNU
licenses without violating the letter.  This section tells you how to
comply with the intention of the GNU licenses as best I understand them.

   The FSF has no objection to your making money.  Its only interest is
that source code to their programs, and libraries, and to modified
versions of their programs and libraries, is always available.

   The short answer is that you do not need to release the source to
your program, but you can't just ship a stripped executable either,
unless you use only the subset of libg++ that includes the iostreams
classes (see discussion below) or the new libstdc++ library (available
in libg++ 2.6.2 and later).

   Compiling your code with a GNU compiler does not affect its
copyright; it is still yours.  However, in order to ship code that
links in a GNU library such as libg++ there are certain rules you must
follow.  The rules are described in the file COPYING.LIB that
accompanies gcc distributions; it is also included in the libg++
distribution.  See that file for the exact rules.  The agreement is
called the Library GNU Public License or LGPL.  It is much "looser"
than the GNU Public License, or GPL, that covers must GNU programs.

   Here's the deal: let's say that you use some version of libg++,
completely unchanged, in your software, and you want to ship only a
binary form of your code.  You can do this, but there are several
special requirements.  If you want to use libg++ but ship only object
code for your code, you have to ship source for libg++ (or ensure
somehow that your customer already has the source for the exact version
you are using), and ship your application in linkable form.  You cannot
forbid your customer from reverse-engineering or extending your program
by exploiting its linkable form.

   Furthermore, if you modify libg++ itself, you must provide source
for your modifications (making a derived class does not count as
modifying the library - that is "a work that uses the library").

   For certain portions of libg++ that implement required parts of the
C++ language (such as iostreams and other standard classes), the FSF has
loosened the copyright requirement still more by adding the "special
exception" clause, which reads as follows:

     As a special exception, if you link this library with files
     compiled with GCC to produce an executable, this does not cause
     the resulting executable to be covered by the GNU General Public
     License.  This exception does not however invalidate any other
     reasons why the executable file might be covered by the GNU
     General Public License.

   If your only use of libg++ uses code with this exception, you may
ship stripped executables or license your executables under different
conditions without fear of violating an FSF copyright.  It is the intent
of FSF and Cygnus that, as the other classes required by the ANSI/ISO
draft standard are developed, these will also be placed under this
"special exception" license.  The code in the new libstdc++ library,
intended to implement standard classes as defined by ANSI/ISO, is also
licensed this way.

   To avoid coming under the influence of the LGPL, you can link with
`-liostream' rather than `-lg++' (for version 2.6.x and earlier), or
`-lstdc++' now that it is available.  In version 2.7.0 all the standard
classes are in `-lstdc++'; you can do the link step with `c++' instead
of `g++' to search only the `-lstdc++' library and avoid the LGPL'ed
code in `-lg++'.

   If you wish to discuss legal issues connected with GNU software on
the net, please use `gnu.misc.discuss', not the technical newsgroups.

--
-- Joe Buck    http://www.synopsys.com/pubs/research/people/jbuck.html
 If you thought flashing ads on Web sites were annoying, wait till the Web
 and your operating system are "seamlessly integrated" and watch the
 pulsating promotional crud spill out over your desktop.  -- Scott Rosenburg