Go to the previous, next section.
Here is the procedure for installing GNU CC on a Unix system. See section Installing GNU CC on VMS, for VMS systems. In this section we assume you compile in the same directory that contains the source files; see section Compilation in a Separate Directory, to find out how to compile in a separate directory on Unix systems.
You cannot install GNU C by itself on MSDOS; it will not compile under any MSDOS compiler except itself. You need to get the complete compilation package DJGPP, which includes binaries as well as sources, and includes all the necessary compilation tools and libraries.
PATH
. The cc
command in
`/usr/ucb' uses libraries which have bugs.
The build machine is the system which you are using, the host machine is the system where you want to run the resulting compiler (normally the build machine), and the target machine is the system for which you want the compiler to generate code.
If you are building a compiler to produce code for the machine it runs on (a native compiler), you normally do not need to specify any operands to `configure'; it will try to guess the type of machine you are on and use that as the build, host and target machines. So you don't need to specify a configuration when building a native compiler unless `configure' cannot figure out what your configuration is or guesses wrong.
In those cases, specify the build machine's configuration name with the `--build' option; the host and target will default to be the same as the build machine. (If you are building a cross-compiler, see section Building and Installing a Cross-Compiler.)
Here is an example:
./configure --build=sparc-sun-sunos4.1
A configuration name may be canonical or it may be more or less abbreviated.
A canonical configuration name has three parts, separated by dashes. It looks like this: `cpu-company-system'. (The three parts may themselves contain dashes; `configure' can figure out which dashes serve which purpose.) For example, `m68k-sun-sunos4.1' specifies a Sun 3.
You can also replace parts of the configuration by nicknames or aliases. For example, `sun3' stands for `m68k-sun', so `sun3-sunos4.1' is another way to specify a Sun 3. You can also use simply `sun3-sunos', since the version of SunOS is assumed by default to be version 4. `sun3-bsd' also works, since `configure' knows that the only BSD variant on a Sun 3 is SunOS.
You can specify a version number after any of the system types, and some of the CPU types. In most cases, the version is irrelevant, and will be ignored. So you might as well specify the version if you know it.
See section Configurations Supported by GNU CC, for a list of supported configuration names and notes on many of the configurations. You should check the notes in that section before proceding any further with the installation of GNU CC.
There are four additional options you can specify independently to describe variant hardware and software configurations. These are `--with-gnu-as', `--with-gnu-ld', `--with-stabs' and `--nfp'.
Using this option does not install GAS. It only modifies the output of GNU CC to work with GAS. Building and installing GAS is up to you.
Conversely, if you do not wish to use GAS and do not specify
`--with-gnu-as' when building GNU CC, it is up to you to make sure
that GAS is not installed. GNU CC searches for a program named
as
in various directories; if the program it finds is GAS, then
it runs GAS. If you are not sure where GNU CC finds the assembler it is
using, try specifying `-v' when you run it.
The systems where it makes a difference whether you use GAS are
`hppa1.0-any-any', `hppa1.1-any-any',
`i386-any-sysv', `i386-any-isc',
`i860-any-bsd', `m68k-bull-sysv', `m68k-hp-hpux',
`m68k-sony-bsd',
`m68k-altos-sysv', `m68000-hp-hpux', `m68000-att-sysv',
and `mips-any'). On any other system, `--with-gnu-as'
has no effect.
On the systems listed above (except for the HP-PA, for ISC on the 386, and for `mips-sgi-irix5.*'), if you use GAS, you should also use the GNU linker (and specify `--with-gnu-ld').
This option does not cause the GNU linker to be installed; it just
modifies the behavior of GNU CC to work with the GNU linker.
Specifically, it inhibits the installation of collect2
, a program
which otherwise serves as a front-end for the system's linker on most
configurations.
Normally, GNU CC uses the ECOFF debugging format by default; if you prefer BSD stabs, specify `--with-stabs' when you configure GNU CC.
No matter which default you choose when you configure GNU CC, the user can use the `-gcoff' and `-gstabs+' options to specify explicitly the debug format for a particular compilation.
`--with-stabs' is meaningful on the ISC system on the 386, also, if `--with-gas' is used. It selects use of stabs debugging information embedded in COFF output. This kind of debugging information supports C++ well; ordinary COFF debugging information does not.
`--with-stabs' is also meaningful on 386 systems running SVR4. It selects use of stabs debugging information embedded in ELF output. The C++ compiler currently (2.6.0) does not support the DWARF debugging information normally used on 386 SVR4 platforms; stabs provide a workable alternative. This requires gas and gdb, as the normal SVR4 tools can not generate or interpret stabs.
The `configure' script searches subdirectories of the source directory for other compilers that are to be integrated into GNU CC. The GNU compiler for C++, called G++ is in a subdirectory named `cp'. `configure' inserts rules into `Makefile' to build all of those compilers.
Here we spell out what files will be set up by configure
. Normally
you need not be concerned with these files.
The top-level config file is located in the subdirectory `config'. Its name is always `xm-something.h'; usually `xm-machine.h', but there are some exceptions.
If your system does not support symbolic links, you might want to set up `config.h' to contain a `#include' command which refers to the appropriate file.
--local-prefix
option below.
You should specify `--local-prefix' only if your site has a different convention (not `/usr/local') for where to put site-specific files.
Do not specify `/usr' as the `--local-prefix'! The
directory you use for `--local-prefix' must not contain
any of the system's standard header files. If it did contain them,
certain programs would be miscompiled (including GNU Emacs, on certain
targets), because this would override and nullify the header file
corrections made by the fixincludes
script.
Bison versions older than Sept 8, 1988 will produce incorrect output for `c-parse.c'.
Alternatively, you can do subsequent compilation using a value of the
PATH
environment variable such that the necessary GNU tools come
before the standard system tools.
`LANGUAGES=c' specifies that only the C compiler should be compiled. The makefile normally builds compilers for all the supported languages; currently, C, C++ and Objective C. However, C is the only language that is sure to work when you build with other non-GNU C compilers. In addition, building anything but C at this stage is a waste of time.
In general, you can specify the languages to build by typing the argument `LANGUAGES="list"', where list is one or more words from the list `c', `c++', and `objective-c'. If you have any additional GNU compilers as subdirectories of the GNU CC source directory, you may also specify their names in this list.
Ignore any warnings you may see about "statement not reached" in `insn-emit.c'; they are normal. Also, warnings about "unknown escape sequence" are normal in `genopinit.c' and perhaps some other files. Likewise, you should ignore warnings about "constant is so large that it is unsigned" in `insn-emit.c' and `insn-recog.c'. Any other compilation errors may represent bugs in the port to your machine or operating system, and should be investigated and reported (see section Reporting Bugs).
Some commercial compilers fail to compile GNU CC because they have bugs or limitations. For example, the Microsoft compiler is said to run out of macro space. Some Ultrix compilers run out of expression space; then you need to break up the statement where the problem happens.
make stage1
The files are moved into a subdirectory named `stage1'.
Once installation is complete, you may wish to delete these files
with rm -r stage1
.
Alternatively, you can do subsequent compilation using a value of the
PATH
environment variable such that the necessary GNU tools come
before the standard system tools.
make CC="stage1/xgcc -Bstage1/" CFLAGS="-g -O"
This is called making the stage 2 compiler.
The command shown above builds compilers for all the supported
languages. If you don't want them all, you can specify the languages to
build by typing the argument `LANGUAGES="list"'. list
should contain one or more words from the list `c', `c++',
`objective-c', and `proto'. Separate the words with spaces.
`proto' stands for the programs protoize
and
unprotoize
; they are not a separate language, but you use
LANGUAGES
to enable or disable their installation.
If you are going to build the stage 3 compiler, then you might want to build only the C language in stage 2.
Once you have built the stage 2 compiler, if you are short of disk space, you can delete the subdirectory `stage1'.
On a 68000 or 68020 system lacking floating point hardware, unless you have selected a `tm.h' file that expects by default that there is no such hardware, do this instead:
make CC="stage1/xgcc -Bstage1/" CFLAGS="-g -O -msoft-float"
make stage2 make CC="stage2/xgcc -Bstage2/" CFLAGS="-g -O"
This is called making the stage 3 compiler. Aside from the `-B'
option, the compiler options should be the same as when you made the
stage 2 compiler. But the LANGUAGES
option need not be the
same. The command shown above builds compilers for all the supported
languages; if you don't want them all, you can specify the languages to
build by typing the argument `LANGUAGES="list"', as described
above.
If you do not have to install any additional GNU tools, you may use the command
make bootstrap LANGUAGES=language-list BOOT_CFLAGS=option-list
instead of making `stage1', `stage2', and performing the two compiler builds.
On some systems, meaningful comparison of object files is impossible; they always appear "different." This is currently true on Solaris and probably on all systems that use ELF object file format. On some versions of Irix on SGI machines and OSF/1 on Alpha systems, you will not be able to compare the files without specifying `-save-temps'; see the description of individual systems above to see if you get comparison failures. You may have similar problems on other systems.
Use this command to compare the files:
make compare
This will mention any object files that differ between stage 2 and stage 3. Any difference, no matter how innocuous, indicates that the stage 2 compiler has compiled GNU CC incorrectly, and is therefore a potentially serious bug which you should investigate and report (see section Reporting Bugs).
If your system does not put time stamps in the object files, then this is a faster way to compare them (using the Bourne shell):
for file in *.o; do cmp $file stage2/$file done
If you have built the compiler with the `-mno-mips-tfile' option on MIPS machines, you will not be able to compare the files.
make objc-runtime CC="stage2/xgcc -Bstage2/" CFLAGS="-g -O"
CC
,
CFLAGS
and LANGUAGES
that you used when compiling the
files that are being installed. One reason this is necessary is that
some versions of Make have bugs and recompile files gratuitously when
you do this step. If you use the same variable values, those files will
be recompiled properly.
For example, if you have built the stage 2 compiler, you can use the following command:
make install CC="stage2/xgcc -Bstage2/" CFLAGS="-g -O" LANGUAGES="list"
This copies the files `cc1', `cpp' and `libgcc.a' to files `cc1', `cpp' and `libgcc.a' in the directory `/usr/local/lib/gcc-lib/target/version', which is where the compiler driver program looks for them. Here target is the target machine type specified when you ran `configure', and version is the version number of GNU CC. This naming scheme permits various versions and/or cross-compilers to coexist.
This also copies the driver program `xgcc' into `/usr/local/bin/gcc', so that it appears in typical execution search paths.
On some systems, this command causes recompilation of some files. This
is usually due to bugs in make
. You should either ignore this
problem, or use GNU Make.
Warning: there is a bug in alloca
in the Sun library. To
avoid this bug, be sure to install the executables of GNU CC that were
compiled by GNU CC. (That is, the executables from stage 2 or 3, not
stage 1.) They use alloca
as a built-in function and never the
one in the library.
(It is usually better to install GNU CC executables from stage 2 or 3, since they usually run faster than the ones compiled with some other compiler.)
make install-libobjc CC="stage2/xgcc -Bstage2/" CFLAGS="-g -O"
Here are the possible CPU types:
1750a, a29k, alpha, arm, cn, clipper, dsp16xx, elxsi, h8300, hppa1.0, hppa1.1, i370, i386, i486, i860, i960, m68000, m68k, m88k, mips, ns32k, powerpc, pyramid, romp, rs6000, sh, sparc, sparclite, sparc64, vax, we32k.
Here are the recognized company names. As you can see, customary abbreviations are used rather than the longer official names.
acorn, alliant, altos, apollo, att, bull, cbm, convergent, convex, crds, dec, dg, dolphin, elxsi, encore, harris, hitachi, hp, ibm, intergraph, isi, mips, motorola, ncr, next, ns, omron, plexus, sequent, sgi, sony, sun, tti, unicom.
The company name is meaningful only to disambiguate when the rest of the information supplied is insufficient. You can omit it, writing just `cpu-system', if it is not needed. For example, `vax-ultrix4.2' is equivalent to `vax-dec-ultrix4.2'.
Here is a list of system types:
386bsd, aix, acis, amigados, aos, aout, bosx, bsd, clix, ctix, cxux, dgux, dynix, ebmon, elf, esix, freebsd, hms, genix, gnu, gnu/linux, hiux, hpux, iris, irix, isc, luna, lynxos, mach, minix, msdos, mvs, netbsd, newsos, nindy, ns, osf, osfrose, ptx, riscix, riscos, rtu, sco, solaris, sunos, sym, sysv, ultrix, unicos, uniplus, unos, vms, vxworks, xenix.
You can omit the system type; then `configure' guesses the operating system from the CPU and company.
You can add a version number to the system type; this may or may not make a difference. For example, you can write `bsd4.3' or `bsd4.4' to distinguish versions of BSD. In practice, the version number is most needed for `sysv3' and `sysv4', which are often treated differently.
If you specify an impossible combination such as `i860-dg-vms', then you may get an error message from `configure', or it may ignore part of the information and do the best it can with the rest. `configure' always prints the canonical name for the alternative that it used. GNU CC does not support all possible alternatives.
Often a particular model of machine has a name. Many machine names are recognized as aliases for CPU/company combinations. Thus, the machine name `sun3', mentioned above, is an alias for `m68k-sun'. Sometimes we accept a company name as a machine name, when the name is popularly used for a particular machine. Here is a table of the known machine names:
3300, 3b1, 3bn, 7300, altos3068, altos, apollo68, att-7300, balance, convex-cn, crds, decstation-3100, decstation, delta, encore, fx2800, gmicro, hp7nn, hp8nn, hp9k2nn, hp9k3nn, hp9k7nn, hp9k8nn, iris4d, iris, isi68, m3230, magnum, merlin, miniframe, mmax, news-3600, news800, news, next, pbd, pc532, pmax, powerpc, ps2, risc-news, rtpc, sun2, sun386i, sun386, sun3, sun4, symmetry, tower-32, tower.
Remember that a machine name specifies both the cpu type and the company name. If you want to install your own homemade configuration files, you can use `local' as the company name to access them. If you use configuration `cpu-local', the configuration name without the cpu prefix is used to form the configuration file names.
Thus, if you specify `m68k-local', configuration uses files `m68k.md', `local.h', `m68k.c', `xm-local.h', `t-local', and `x-local', all in the directory `config/m68k'.
Here is a list of configurations that have special treatment or special things you must know:
Starting with GCC 2.6.1, the MIL-STD-1750A cross configuration no longer
supports the Tektronix Assembler, but instead produces output for
as1750
, an assembler/linker available under the GNU Public
License for the 1750A. Contact okellogg@salyko.cube.net for more
details on obtaining `as1750'. A similarly licensed simulator for
the 1750A is available from same address.
You should ignore a fatal error during the building of libgcc (libgcc is not yet implemented for the 1750A.)
The as1750
assembler requires the file `ms1750.inc', which is
found in the directory `config/1750a'.
GNU CC produced the same sections as the Fairchild F9450 C Compiler, namely:
NREL
SREL
KREL
IREL
The smallest addressable unit is 16 bits (BITS_PER_UNIT is 16). This means that type `char' is represented with a 16-bit word per character. The 1750A's "Load/Store Upper/Lower Byte" instructions are not used by GNU CC.
There is a problem with long argument lists to functions. The compiler aborts if the sum of space needed by all arguments exceeds 14 words. This is because the arguments are passed in registers (R0..R13) not on the stack, and there is a problem with passing further arguments (i.e. beyond those in R0..R13) via the stack.
If efficiency is less important than using long argument lists, you
can change the definition of the FUNCTION_ARG
macro in
`config/1750/1750a.h' to always return zero. If you do that,
GNU CC will pass all parameters on the stack.
GNU CC writes a `.verstamp' directive to the assembler output file unless it is built as a cross-compiler. It gets the version to use from the system header file `/usr/include/stamp.h'. If you install a new version of OSF/1, you should rebuild GCC to pick up the new version stamp.
Note that since the Alpha is a 64-bit architecture, cross-compilers from 32-bit machines will not generate code as efficient as that generated when the compiler is running on a 64-bit machine because many optimizations that depend on being able to represent a word on the target in an integral value on the host cannot be performed. Building cross-compilers on the Alpha for 32-bit machines has only been tested in a few cases and may not work properly.
make compare
may fail on old versions of OSF/1 unless you add
`-save-temps' to CFLAGS
. On these systems, the name of the
assembler input file is stored in the object file, and that makes
comparison fail if it differs between the stage1
and
stage2
compilations. The option `-save-temps' forces a
fixed name to be used for the assembler input file, instead of a
randomly chosen name in `/tmp'. Do not add `-save-temps'
unless the comparisons fail without that option. If you add
`-save-temps', you will have to manually delete the `.i' and
`.s' files after each series of compilations.
GNU CC now supports both the native (ECOFF) debugging format used by DBX and GDB and an encapsulated STABS format for use only with GDB. See the discussion of the `--with-stabs' option of `configure' above for more information on these formats and how to select them.
There is a bug in DEC's assembler that produces incorrect line numbers for ECOFF format when the `.align' directive is used. To work around this problem, GNU CC will not emit such alignment directives while writing ECOFF format debugging information even if optimization is being performed. Unfortunately, this has the very undesirable side-effect that code addresses when `-O' is specified are different depending on whether or not `-g' is also specified.
To avoid this behavior, specify `-gstabs+' and use GDB instead of DBX. DEC is now aware of this problem with the assembler and hopes to provide a fix shortly.
You may need to make a variant of the file `arm.h' for your particular configuration.
You may need to make a variant of the file `a29k.h' for your particular configuration.
The MIPS C compiler needs to be told to increase its table size
for switch statements with the `-Wf,-XNg1500' option in
order to compile `cp/parse.c'. If you use the `-O2'
optimization option, you also need to use `-Olimit 3000'.
Both of these options are automatically generated in the
`Makefile' that the shell script `configure' builds.
If you override the CC
make variable and use the MIPS
compilers, you may need to add `-Wf,-XNg1500 -Olimit 3000'.
mrs@cygnus.com
for more details.
The easiest way to handle this problem is to use `configure hpnnn' or `configure hpnnn-hpux', where nnn is the model number of the machine. Then `configure' will figure out if the machine is a 1.0 or 1.1. Use `uname -a' to find out the model number of your machine.
`-g' does not work on HP-UX, since that system uses a peculiar debugging format which GNU CC does not know about. However, `-g' will work if you also use GAS and GDB in conjunction with GCC. We highly recommend using GAS for all HP-PA configurations.
You should be using GAS-2.3 (or later) along with GDB-4.12 (or later). These can be retrieved from all the traditional GNU ftp archive sites.
Build GAS and install the resulting binary as:
/usr/local/lib/gcc-lib/configuration/gccversion/as
where configuration is the configuration name (perhaps `hpnnn-hpux') and gccversion is the GNU CC version number. Do this before starting the build process, otherwise you will get errors from the HPUX assembler while building `libgcc2.a'. The command
make install-dir
will create the necessary directory hierarchy so you can install GAS before building GCC.
To enable debugging, configure GNU CC with the `--with-gnu-as' option before building.
It has been reported that GNU CC produces invalid assembly code for 1.1 machines running HP-UX 8.02 when using the HP assembler. Typically the errors look like this:
as: bug.s @line#15 [err#1060] Argument 0 or 2 in FARG upper - lookahead = ARGW1=FR,RTNVAL=GR as: foo.s @line#28 [err#1060] Argument 0 or 2 in FARG upper - lookahead = ARGW1=FR
You can check the version of HP-UX you are running by executing the command `uname -r'. If you are indeed running HP-UX 8.02 on a PA and using the HP assembler then configure GCC with "hpnnn-hpux8.02".
In ISC version 4.1, `sed' core dumps when building `deduced.h'. Use the version of `sed' from version 4.0.
See section Installing GNU CC on the Sun, for information on installing GNU CC on Sun systems.
law@cs.utah.edu
to get binaries of GNU CC for bootstrapping.
Installing GNU CC on the 3b1 is difficult if you do not already have GNU CC running, due to bugs in the installed C compiler. However, the following procedure might work. We are unable to test it.
obstack_free
in the file
`tree.c' with _obstack_free
.
make
to get the first-stage GNU CC.
F.Pierresteguy@frcl.bull.fr
.
The Unos assembler is named casm
instead of as
. For some
strange reason linking `/bin/as' to `/bin/casm' changes the
behavior, and does not work. So, when installing GNU CC, you should
install the following script as `as' in the subdirectory where
the passes of GCC are installed:
#!/bin/sh casm $*
The default Unos library is named `libunos.a' instead of `libc.a'. To allow GNU CC to function, either change all references to `-lc' in `gcc.c' to `-lunos' or link `/lib/libc.a' to `/lib/libunos.a'.
When compiling GNU CC with the standard compiler, to overcome bugs in
the support of alloca
, do not use `-O' when making stage 2.
Then use the stage 2 compiler with `-O' to make the stage 3
compiler. This compiler will have the same characteristics as the usual
stage 2 compiler on other systems. Use it to make a stage 4 compiler
and compare that with stage 3 to verify proper compilation.
(Perhaps simply defining ALLOCA
in `x-crds' as described in
the comments there will make the above paragraph superfluous. Please
inform us of whether this works.)
Unos uses memory segmentation instead of demand paging, so you will need a lot of memory. 5 Mb is barely enough if no other tasks are running. If linking `cc1' fails, try putting the object files into a library and linking from that library.
In addition, if you wish to use gas `--with-gnu-as' you must use gas version 2.1 or later, and you must use the GNU linker version 2.1 or later. Earlier versions of gas relied upon a program which converted the gas output into the native HP/UX format, but that program has not been kept up to date. gdb does not understand that native HP/UX format, so you must use gas if you wish to use gdb.
See section Installing GNU CC on the Sun, for information on installing GNU CC on Sun systems.
It is best, however, to use an older version of GNU CC for bootstrapping if you have one.
If you do not specify a configuration name, `configure' guesses the configuration based on the current software development environment.
memcpy
, memcmp
, and memset
. If your system lacks
these, you must remove or undo the definition of
TARGET_MEM_FUNCTIONS
in `mips-bsd.h'.
The MIPS C compiler needs to be told to increase its table size
for switch statements with the `-Wf,-XNg1500' option in
order to compile `cp/parse.c'. If you use the `-O2'
optimization option, you also need to use `-Olimit 3000'.
Both of these options are automatically generated in the
`Makefile' that the shell script `configure' builds.
If you override the CC
make variable and use the MIPS
compilers, you may need to add `-Wf,-XNg1500 -Olimit 3000'.
CC
make variable and use the MIPS
compilers, you may need to add `-Wf,-XNg1500 -Olimit 3000'.
MIPS computers running RISC-OS can support four different personalities: default, BSD 4.3, System V.3, and System V.4 (older versions of RISC-OS don't support V.4). To configure GCC for these platforms use the following configurations:
rev
'
rev
.
rev
bsd'
rev
.
rev
sysv4'
rev
.
rev
sysv'
rev
.
The revision rev
mentioned above is the revision of
RISC-OS to use. You must reconfigure GCC when going from a
RISC-OS revision 4 to RISC-OS revision 5. This has the effect of
avoiding a linker
bug (see section Installation Problems, for more details).
make compare
may fail on version 5 of IRIX unless you add
`-save-temps' to CFLAGS
. On these systems, the name of the
assembler input file is stored in the object file, and that makes
comparison fail if it differs between the stage1
and
stage2
compilations. The option `-save-temps' forces a
fixed name to be used for the assembler input file, instead of a
randomly chosen name in `/tmp'. Do not add `-save-temps'
unless the comparisons fail without that option. If you do you
`-save-temps', you will have to manually delete the `.i' and
`.s' files after each series of compilations.
The MIPS C compiler needs to be told to increase its table size
for switch statements with the `-Wf,-XNg1500' option in
order to compile `cp/parse.c'. If you use the `-O2'
optimization option, you also need to use `-Olimit 3000'.
Both of these options are automatically generated in the
`Makefile' that the shell script `configure' builds.
If you override the CC
make variable and use the MIPS
compilers, you may need to add `-Wf,-XNg1500 -Olimit 3000'.
On Irix version 4.0.5F, and perhaps on some other versions as well, there is an assembler bug that reorders instructions incorrectly. To work around it, specify the target configuration `mips-sgi-irix4loser'. This configuration inhibits assembler optimization.
In a compiler configured with target `mips-sgi-irix4', you can turn off assembler optimization by using the `-noasmopt' option. This compiler option passes the option `-O0' to the assembler, to inhibit reordering.
The `-noasmopt' option can be useful for testing whether a problem is due to erroneous assembler reordering. Even if a problem does not go away with `-noasmopt', it may still be due to assembler reordering--perhaps GNU CC itself was miscompiled as a result.
To enable debugging under Irix 5, you must use GNU as 2.5 or later, and use the --with-gnu-as configure option when configuring gcc. GNU as is distributed as part of the binutils package.
alloca
and malloc
; you must get the compiled versions of these from GNU
Emacs.
hc
, the Metaware compiler, it will work, but you will get
mismatches between the stage 2 and stage 3 compilers in various files.
These errors are minor differences in some floating-point constants and
can be safely ignored; the stage 3 compiler is correct.
Only AIX is supported on the PowerPC. GNU CC does not yet support the 64-bit PowerPC instructions.
Objective C does not work on this architecture.
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 GNU CC 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".
vcc
). It produces incorrect code
in some cases (for example, when alloca
is used).
Meanwhile, compiling `cp/parse.c' with pcc does not work because of an internal table size limitation in that compiler. To avoid this problem, compile just the GNU C compiler first, and use it to recompile building all the languages that you want to run.
Don't use `-g' when compiling with the system's compiler. The system's linker seems to be unable to handle such a large program with debugging information.
The system's compiler runs out of capacity when compiling `stmt.c' in GNU CC. You can work around this by building `cpp' in GNU CC first, then use that instead of the system's preprocessor with the system's C compiler to compile `stmt.c'. Here is how:
mv /lib/cpp /lib/cpp.att cp cpp /lib/cpp.gnu echo '/lib/cpp.gnu -traditional ${1+"$@"}' > /lib/cpp chmod +x /lib/cpp
The system's compiler produces bad code for some of the GNU CC optimization files. So you must build the stage 2 compiler without optimization. Then build a stage 3 compiler with optimization. That executable should work. Here are the necessary commands:
make LANGUAGES=c CC=stage1/xgcc CFLAGS="-Bstage1/ -g" make stage2 make CC=stage2/xgcc CFLAGS="-Bstage2/ -g -O"
You may need to raise the ULIMIT setting to build a C++ compiler, as the file `cc1plus' is larger than one megabyte.
If you wish to build the object files and executables in a directory other than the one containing the source files, here is what you must do differently:
VPATH
feature. (GNU Make supports it, as do Make versions on most BSD
systems.)
make distclean
mkdir gcc-sun3 cd gcc-sun3
On systems that do not support symbolic links, this directory must be on the same file system as the source code directory.
../gcc/configure ...
This also tells configure
where to find the compiler sources;
configure
takes the directory from the file name that was used to
invoke it. But if you want to be sure, you can specify the source
directory with the `--srcdir' option, like this:
../gcc/configure --srcdir=../gcc other options
The directory you specify with `--srcdir' need not be the same
as the one that configure
is found in.
Now, you can run make
in that directory. You need not repeat the
configuration steps shown above, when ordinary source files change. You
must, however, run configure
again when the configuration files
change, if your system does not support symbolic links.
GNU CC can function as a cross-compiler for many machines, but not all.
Since GNU CC generates assembler code, you probably need a cross-assembler that GNU CC can run, in order to produce object files. If you want to link on other than the target machine, you need a cross-linker as well. You also need header files and libraries suitable for the target machine that you can install on the host machine.
To compile and run a program using a cross-compiler involves several steps:
It is most convenient to do all of these steps on the same host machine, since then you can do it all with a single invocation of GNU CC. This requires a suitable cross-assembler and cross-linker. For some targets, the GNU assembler and linker are available.
To build GNU CC as a cross-compiler, you start out by running `configure'. Use the `--target=target' to specify the target type. If `configure' was unable to correctly identify the system you are running on, also specify the `--build=build' option. For example, here is how to configure for a cross-compiler that produces code for an HP 68030 system running BSD on a system that `configure' can correctly identify:
./configure --target=m68k-hp-bsd4.3
If you have a cross-assembler and cross-linker available, you should install them now. Put them in the directory `/usr/local/target/bin'. Here is a table of the tools you should put in this directory:
The installation of GNU CC will find these programs in that directory, and copy or link them to the proper place to for the cross-compiler to find them when run later.
The easiest way to provide these files is to build the Binutils package and GAS. Configure them with the same `--host' and `--target' options that you use for configuring GNU CC, then build and install them. They install their executables automatically into the proper directory. Alas, they do not support all the targets that GNU CC supports.
If you want to install libraries to use with the cross-compiler, such as a standard C library, put them in the directory `/usr/local/target/lib'; installation of GNU CC copies all all the files in that subdirectory into the proper place for GNU CC to find them and link with them. Here's an example of copying some libraries from a target machine:
ftp target-machine lcd /usr/local/target/lib cd /lib get libc.a cd /usr/lib get libg.a get libm.a quit
The precise set of libraries you'll need, and their locations on the target machine, vary depending on its operating system.
Many targets require "start files" such as `crt0.o' and
`crtn.o' which are linked into each executable; these too should be
placed in `/usr/local/target/lib'. There may be several
alternatives for `crt0.o', for use with profiling or other
compilation options. Check your target's definition of
STARTFILE_SPEC
to find out what start files it uses.
Here's an example of copying these files from a target machine:
ftp target-machine lcd /usr/local/target/lib prompt cd /lib mget *crt*.o cd /usr/lib mget *crt*.o quit
Code compiled by GNU CC uses certain runtime support functions implicitly. Some of these functions can be compiled successfully with GNU CC itself, but a few cannot be. These problem functions are in the source file `libgcc1.c'; the library made from them is called `libgcc1.a'.
When you build a native compiler, these functions are compiled with some other compiler--the one that you use for bootstrapping GNU CC. Presumably it knows how to open code these operations, or else knows how to call the run-time emulation facilities that the machine comes with. But this approach doesn't work for building a cross-compiler. The compiler that you use for building knows about the host system, not the target system.
So, when you build a cross-compiler you have to supply a suitable library `libgcc1.a' that does the job it is expected to do.
To compile `libgcc1.c' with the cross-compiler itself does not work. The functions in this file are supposed to implement arithmetic operations that GNU CC does not know how to open code, for your target machine. If these functions are compiled with GNU CC itself, they will compile into infinite recursion.
On any given target, most of these functions are not needed. If GNU CC can open code an arithmetic operation, it will not call these functions to perform the operation. It is possible that on your target machine, none of these functions is needed. If so, you can supply an empty library as `libgcc1.a'.
Many targets need library support only for multiplication and division.
If you are linking with a library that contains functions for
multiplication and division, you can tell GNU CC to call them directly
by defining the macros MULSI3_LIBCALL
, and the like. These
macros need to be defined in the target description macro file. For
some targets, they are defined already. This may be sufficient to
avoid the need for libgcc1.a; if so, you can supply an empty library.
Some targets do not have floating point instructions; they need other functions in `libgcc1.a', which do floating arithmetic. Recent versions of GNU CC have a file which emulates floating point. With a certain amount of work, you should be able to construct a floating point emulator that can be used as `libgcc1.a'. Perhaps future versions will contain code to do this automatically and conveniently. That depends on whether someone wants to implement it.
If your target system has another C compiler, you can configure GNU CC as a native compiler on that machine, build just `libgcc1.a' with `make libgcc1.a' on that machine, and use the resulting file with the cross-compiler. To do this, execute the following on the target machine:
cd target-build-dir ./configure --host=sparc --target=sun3 make libgcc1.a
And then this on the host machine:
ftp target-machine binary cd target-build-dir get libgcc1.a quit
Another way to provide the functions you need in `libgcc1.a' is to
define the appropriate perform_...
macros for those
functions. If these definitions do not use the C arithmetic operators
that they are meant to implement, you should be able to compile them
with the cross-compiler you are building. (If these definitions already
exist for your target file, then you are all set.)
To build `libgcc1.a' using the perform macros, use
`LIBGCC1=libgcc1.a OLDCC=./xgcc' when building the compiler.
Otherwise, you should place your replacement library under the name
`libgcc1.a' in the directory in which you will build the
cross-compiler, before you run make
.
If you are cross-compiling a standalone program or a program for an embedded system, then you may not need any header files except the few that are part of GNU CC (and those of your program). However, if you intend to link your program with a standard C library such as `libc.a', then you probably need to compile with the header files that go with the library you use.
The GNU C compiler does not come with these files, because (1) they are system-specific, and (2) they belong in a C library, not in a compiler.
If the GNU C library supports your target machine, then you can get the header files from there (assuming you actually use the GNU library when you link your program).
If your target machine comes with a C compiler, it probably comes with suitable header files also. If you make these files accessible from the host machine, the cross-compiler can use them also.
Otherwise, you're on your own in finding header files to use when cross-compiling.
When you have found suitable header files, put them in `/usr/local/target/include', before building the cross compiler. Then installation will run fixincludes properly and install the corrected versions of the header files where the compiler will use them.
Provide the header files before you build the cross-compiler, because the build stage actually runs the cross-compiler to produce parts of `libgcc.a'. (These are the parts that can be compiled with GNU CC.) Some of them need suitable header files.
Here's an example showing how to copy the header files from a target machine. On the target machine, do this:
(cd /usr/include; tar cf - .) > tarfile
Then, on the host machine, do this:
ftp target-machine lcd /usr/local/target/include get tarfile quit tar xf tarfile
Now you can proceed just as for compiling a single-machine compiler through the step of building stage 1. If you have not provided some sort of `libgcc1.a', then compilation will give up at the point where it needs that file, printing a suitable error message. If you do provide `libgcc1.a', then building the compiler will automatically compile and link a test program called `cross-test'; if you get errors in the linking, it means that not all of the necessary routines in `libgcc1.a' are available.
If you are making a cross-compiler for an embedded system, and there is no `stdio.h' header for it, then the compilation of `enquire' will probably fail. The job of `enquire' is to run on the target machine and figure out by experiment the nature of its floating point representation. `enquire' records its findings in the header file `float.h'. If you can't produce this file by running `enquire' on the target machine, then you will need to come up with a suitable `float.h' in some other way (or else, avoid using it in your programs).
Do not try to build stage 2 for a cross-compiler. It doesn't work to rebuild GNU CC as a cross-compiler using the cross-compiler, because that would produce a program that runs on the target machine, not on the host. For example, if you compile a 386-to-68030 cross-compiler with itself, the result will not be right either for the 386 (because it was compiled into 68030 code) or for the 68030 (because it was configured for a 386 as the host). If you want to compile GNU CC into 68030 code, whether you compile it on a 68030 or with a cross-compiler on a 386, you must specify a 68030 as the host when you configure it.
To install the cross-compiler, use `make install', as usual.
On Solaris (version 2.1), do not use the linker or other tools in
`/usr/ucb' to build GNU CC. Use /usr/ccs/bin
.
Make sure the environment variable FLOAT_OPTION
is not set when
you compile `libgcc.a'. If this option were set to f68881
when `libgcc.a' is compiled, the resulting code would demand to be
linked with a special startup file and would not link properly without
special pains.
There is a bug in alloca
in certain versions of the Sun library.
To avoid this bug, install the binaries of GNU CC that were compiled by
GNU CC. They use alloca
as a built-in function and never the one
in the library.
Some versions of the Sun compiler crash when compiling GNU CC. The problem is a segmentation fault in cpp. This problem seems to be due to the bulk of data in the environment variables. You may be able to avoid it by using the following command to compile GNU CC with Sun CC:
make CC="TERMCAP=x OBJS=x LIBFUNCS=x STAGESTUFF=x cc"
The VMS version of GNU CC is distributed in a backup saveset containing both source code and precompiled binaries.
To install the `gcc' command so you can use the compiler easily, in the same manner as you use the VMS C compiler, you must install the VMS CLD file for GNU CC as follows:
$ assign /system /translation=concealed - disk:[gcc.] gnu_cc $ assign /system /translation=concealed - disk:[gcc.include.] gnu_cc_include
with the appropriate disk and directory names. These commands can be placed in your system startup file so they will be executed whenever the machine is rebooted. You may, if you choose, do this via the `GCC_INSTALL.COM' script in the `[GCC]' directory.
$ set command /table=sys$common:[syslib]dcltables - /output=sys$common:[syslib]dcltables gnu_cc:[000000]gcc $ install replace sys$common:[syslib]dcltables
$ library/help sys$library:helplib.hlb gcc.hlp
Now you can invoke the compiler with a command like `gcc /verbose file.c', which is equivalent to the command `gcc -v -c file.c' in Unix.
If you wish to use GNU C++ you must first install GNU CC, and then perform the following steps:
$ assign /system /translation=concealed - disk:[gcc.gxx_include.] gnu_gxx_include
with the appropriate disk and directory name. If you are going to be using libg++, this is where the libg++ install procedure will install the libg++ header files.
The GNU C++ compiler can be invoked with a command like `gcc /plus /verbose file.cc', which is equivalent to the command `g++ -v -c file.cc' in Unix.
We try to put corresponding binaries and sources on the VMS distribution tape. But sometimes the binaries will be from an older version than the sources, because we don't always have time to update them. (Use the `/version' option to determine the version number of the binaries and compare it with the source file `version.c' to tell whether this is so.) In this case, you should use the binaries you get to recompile the sources. If you must recompile, here is how:
$ @vmsconfig.com
$ assign /system /translation=concealed - disk:[bison.] gnu_bison
You may, if you choose, use the `INSTALL_BISON.COM' script in the `[BISON]' directory.
$ set command /table=sys$common:[syslib]dcltables - /output=sys$common:[syslib]dcltables - gnu_bison:[000000]bison $ install replace sys$common:[syslib]dcltables
To install the library, use the following commands:
$ library gnu_cc:[000000]gcclib/delete=(new,eprintf) $ library gnu_cc:[000000]gcclib/delete=L_* $ library libgcc2/extract=*/output=libgcc2.obj $ library gnu_cc:[000000]gcclib libgcc2.obj
The first command simply removes old modules that will be replaced with
modules from `libgcc2' under different module names. The modules
new
and eprintf
may not actually be present in your
`gcclib.olb'---if the VMS librarian complains about those modules
not being present, simply ignore the message and continue on with the
next command. The second command removes the modules that came from the
previous version of the library `libgcc2.c'.
Whenever you update the compiler on your system, you should also update the library with the above procedure.
$ assign dua0:[gcc.build_dir.]/translation=concealed, - dua1:[gcc.source_dir.]/translation=concealed gcc_build $ set default gcc_build:[000000]
where the directory `dua1:[gcc.source_dir]' contains the source code, and the directory `dua0:[gcc.build_dir]' is meant to contain all of the generated object files and executables. Once you have done this, you can proceed building GCC as described above. (Keep in mind that `gcc_build' is a rooted logical name, and thus the device names in each element of the search list must be an actual physical device name rather than another rooted logical name).
extern const
variables will not have
the read-only bit set, and the linker will generate warning messages
about mismatched psect attributes for these variables. These warning
messages are merely a nuisance, and can safely be ignored.
If you are compiling with a version of GNU CC older than 1.33, specify
`/DEFINE=("inline=")' as an option in all the compilations. This
requires editing all the gcc
commands in `make-cc1.com'.
(The older versions had problems supporting inline
.) Once you
have a working 1.33 or newer GNU CC, you can change this file back.
CC
, CFLAGS
, and
LIBS
. See comments in those files. However, you must
also have a working version of the GNU assembler (GNU as, aka GAS) as
it is used as the back-end for GNU CC to produce binary object modules
and is not included in the GNU CC sources. GAS is also needed to
compile `libgcc2' in order to build `gcclib' (see above);
`make-l2.com' expects to be able to find it operational in
`gnu_cc:[000000]gnu-as.exe'.
To use GNU CC on VMS, you need the VMS driver programs `gcc.exe', `gcc.com', and `gcc.cld'. They are distributed with the VMS binaries (`gcc-vms') rather than the GNU CC sources. GAS is also included in `gcc-vms', as is Bison.
Once you have successfully built GNU CC with VAX C, you should use the
resulting compiler to rebuild itself. Before doing this, be sure to
restore the CC
, CFLAGS
, and LIBS
definitions in
`make-cccp.com' and `make-cc1.com'. The second generation
compiler will be able to take advantage of many optimizations that must
be suppressed when building with other compilers.
Under previous versions of GNU CC, the generated code would occasionally give strange results when linked with the sharable `VAXCRTL' library. Now this should work.
Even with this version, however, GNU CC itself should not be linked with
the sharable `VAXCRTL'. The version of qsort
in
`VAXCRTL' has a bug (known to be present in VMS versions V4.6
through V5.5) which causes the compiler to fail.
The executables are generated by `make-cc1.com' and
`make-cccp.com' use the object library version of `VAXCRTL' in
order to make use of the qsort
routine in `gcclib.olb'. If
you wish to link the compiler executables with the shareable image
version of `VAXCRTL', you should edit the file `tm.h' (created
by `vmsconfig.com') to define the macro QSORT_WORKAROUND
.
QSORT_WORKAROUND
is always defined when GNU CC is compiled with
VAX C, to avoid a problem in case `gcclib.olb' is not yet
available.
collect2
Many target systems do not have support in the assembler and linker for
"constructors"---initialization functions to be called before the
official "start" of main
. On such systems, GNU CC uses a
utility called collect2
to arrange to call these functions at
start time.
The program collect2
works by linking the program once and
looking through the linker output file for symbols with particular names
indicating they are constructor functions. If it finds any, it
creates a new temporary `.c' file containing a table of them,
compiles it, and links the program a second time including that file.
The actual calls to the constructors are carried out by a subroutine
called __main
, which is called (automatically) at the beginning
of the body of main
(provided main
was compiled with GNU
CC). Calling __main
is necessary, even when compiling C code, to
allow linking C and C++ object code together. (If you use
`-nostdlib', you get an unresolved reference to __main
,
since it's defined in the standard GCC library. Include `-lgcc' at
the end of your compiler command line to resolve this reference.)
The program collect2
is installed as ld
in the directory
where the passes of the compiler are installed. When collect2
needs to find the real ld
, it tries the following file
names:
PATH
.
REAL_LD_FILE_NAME
configuration macro,
if specified.
collect2
will not execute itself recursively.
PATH
.
"The compiler's search directories" means all the directories where
gcc
searches for passes of the compiler. This includes
directories that you specify with `-B'.
Cross-compilers search a little differently:
PATH
.
REAL_LD_FILE_NAME
configuration macro,
if specified.
PATH
.
collect2
explicitly avoids running ld
using the file name
under which collect2
itself was invoked. In fact, it remembers
up a list of such names--in case one copy of collect2
finds
another copy (or version) of collect2
installed as ld
in a
second place in the search path.
collect2
searches for the utilities nm
and strip
using the same algorithm as above for ld
.
GCC_INCLUDE_DIR
means the same thing for native and cross. It is
where GNU CC stores its private include files, and also where GNU CC
stores the fixed include files. A cross compiled GNU CC runs
fixincludes
on the header files in `$(tooldir)/include'.
(If the cross compilation header files need to be fixed, they must be
installed before GNU CC is built. If the cross compilation header files
are already suitable for ANSI C and GNU CC, nothing special need be
done).
GPLUS_INCLUDE_DIR
means the same thing for native and cross. It
is where g++
looks first for header files. libg++
installs only target independent header files in that directory.
LOCAL_INCLUDE_DIR
is used only for a native compiler. It is
normally `/usr/local/include'. GNU CC searches this directory so
that users can install header files in `/usr/local/include'.
CROSS_INCLUDE_DIR
is used only for a cross compiler. GNU CC
doesn't install anything there.
TOOL_INCLUDE_DIR
is used for both native and cross compilers. It
is the place for other packages to install header files that GNU CC will
use. For a cross-compiler, this is the equivalent of
`/usr/include'. When you build a cross-compiler,
fixincludes
processes any header files in this directory.
Go to the previous, next section.