path: root/gcc/f/g77install.texi

@c Copyright (C) 1995-1997 Free Software Foundation, Inc.
@c This is part of the G77 manual.
@c For copying conditions, see the file g77.texi.

@c The text of this file appears in the file INSTALL
@c in the G77 distribution, as well as in the G77 manual.

@c 1998-07-13

@set version-g77 0.5.24
@set version-gcc 2.8.1
@set version-autoconf 2.12
@set version-bison 1.25
@set version-gperf 2.5
@set version-gzip 1.2.4
@set version-make 3.76.1
@set version-makeinfo 1.68
@set version-patch 2.5
@set version-sed 2.05
@set version-tar 1.12
@set version-texinfo 3.11

@ifclear INSTALLONLY
@node Installation
@chapter Installing GNU Fortran
@end ifclear
@cindex installing GNU Fortran

The following information describes how to install @code{g77}.

Note that, for @code{egcs} users,
much of this information is obsolete,
and is superceded by the
@code{egcs} installation procedures.
Such information is explicitly flagged as such.

The information in this file generally pertains to dealing
with @emph{source} distributions of @code{g77} and @code{gcc}.
It is possible that some of this information will be applicable
to some @emph{binary} distributions of these products---however,
since these distributions are not made by the maintainers of
@code{g77}, responsibility for binary distributions rests with
whoever built and first distributed them.

Nevertheless, efforts to make @code{g77} easier to both build
and install from source and package up as a binary distribution
are ongoing.

@menu
* Prerequisites::          Make sure your system is ready for @code{g77}.
* Problems Installing::    Known trouble areas.
* Settings::               Changing @code{g77} internals before building.
* Quick Start::            The easier procedure for non-experts.
* Complete Installation::  For experts, or those who want to be: the details.
* Distributing Binaries::  If you plan on distributing your @code{g77}.
@end menu

@node Prerequisites
@section Prerequisites
@cindex prerequisites

@emph{Version info:}
For @code{egcs} users, the following information is
superceded by the @code{egcs} installation instructions.

The procedures described to unpack, configure, build, and
install @code{g77} assume your system has certain programs
already installed.

The following prerequisites should be met by your
system before you follow the @code{g77} installation instructions:

@table @asis
@item @code{gzip} and @code{tar}
To unpack the @code{gcc} and @code{g77} distributions,
you'll need the @code{gunzip} utility in the @code{gzip}
distribution.
Most UNIX systems already have @code{gzip} installed.
If yours doesn't, you can get it from the FSF.

Note that you'll need @code{tar} and other utilities
as well, but all UNIX systems have these.
There are GNU versions of all these available---in fact,
a complete GNU UNIX system can be put together on
most systems, if desired.

The version of GNU @code{gzip} used to package this release
is @value{version-gzip}.
(The version of GNU @code{tar} used to package this release
is @value{version-tar}.)

@item @file{gcc-@value{version-gcc}.tar.gz}
You need to have this, or some other applicable, version
of @code{gcc} on your system.
The version should be an exact copy of a distribution
from the FSF.
Its size is approximately 8.4MB.

If you've already unpacked @file{gcc-@value{version-gcc}.tar.gz} into a
directory (named @file{gcc-@value{version-gcc}}) called the @dfn{source tree}
for @code{gcc}, you can delete the distribution
itself, but you'll need to remember to skip any instructions to unpack
this distribution.

Without an applicable @code{gcc} source tree, you cannot
build @code{g77}.
You can obtain an FSF distribution of @code{gcc} from the FSF.

@item @file{g77-@value{version-g77}.tar.gz}
You probably have already unpacked this package,
or you are reading an advance copy of these installation instructions,
which are contained in this distribution.
The size of this package is approximately 1.4MB.

You can obtain an FSF distribution of @code{g77} from the FSF,
the same way you obtained @code{gcc}.

@item Enough disk space
The amount of disk space needed to unpack, build, install,
and use @code{g77} depends on the type of system you're
using, how you build @code{g77}, and how much of it you
install (primarily, which languages you install).

The sizes shown below assume all languages distributed
in @code{gcc-@value{version-gcc}}, plus @code{g77}, will be built
and installed.
These sizes are indicative of GNU/Linux systems on
Intel x86 running COFF and on Digital Alpha (AXP) systems
running ELF.
These should be fairly representative of 32-bit and 64-bit
systems, respectively.

Note that all sizes are approximate and subject to change without
notice!
They are based on preliminary releases of g77 made shortly
before the public beta release.

@itemize ---
@item
@code{gcc} and @code{g77} distributions occupy 10MB
packed, 40MB unpacked.
These consist of the source code and documentation,
plus some derived files (mostly documentation), for
@code{gcc} and @code{g77}.
Any deviations from these numbers for different
kinds of systems are likely to be very minor.

@item
A ``bootstrap'' build requires an additional 91MB
for a total of 132MB on an ix86, and an additional
136MB for a total of 177MB on an Alpha.

@item
Removing @file{gcc/stage1} after the build recovers
13MB for a total of 119MB on an ix86, and recovers
21MB for a total of 155MB on an Alpha.

After doing this, the integrity of the build can
still be verified via @samp{make compare}, and the
@code{gcc} compiler modified and used to build itself for
testing fairly quickly, using the copy of the compiler
kept in @code{gcc/stage2}.

@item
Removing @file{gcc/stage2} after the build further
recovers 39MB for a total of 80MB, and recovers
57MB for a total of 98MB on an Alpha.

After doing this, the compiler can still be installed,
especially if GNU @code{make} is used to avoid
gratuitous rebuilds (or, the installation can be done
by hand).

@item
Installing @code{gcc} and @code{g77} copies
23MB onto the @samp{--prefix} disk for a total of 103MB
on an ix86, and copies 31MB onto the @samp{--prefix}
disk for a total of 130MB on an Alpha.
@end itemize

After installation, if no further modifications and
builds of @code{gcc} or @code{g77} are planned, the
source and build directory may be removed, leaving
the total impact on a system's disk storage as
that of the amount copied during installation.

Systems with the appropriate version of @code{gcc}
installed don't require the complete
bootstrap build.
Doing a ``straight build'' requires about as much
space as does a bootstrap build followed by removing
both the @file{gcc/stage1} and @file{gcc/stage2}
directories.

Installing @code{gcc} and @code{g77} over existing
versions might require less @emph{new} disk space,
but note that, unlike many products, @code{gcc}
installs itself in a way that avoids overwriting other
installed versions of itself, so that other versions may
easily be invoked (via @samp{gcc -V @var{version}}).

So, the amount of space saved as a result of having
an existing version of @code{gcc} and @code{g77}
already installed is not much---typically only the
command drivers (@code{gcc}, @code{g77}, @code{g++},
and so on, which are small) and the documentation
is overwritten by the new installation.
The rest of the new installation is done without
replacing existing installed versions (assuming
they have different version numbers).

@item @code{make}
Your system must have @code{make}, and you will probably save
yourself a lot of trouble if it is GNU @code{make} (sometimes
referred to as @code{gmake}).
In particular, you probably need GNU @code{make}
to build outside the source directory
(with @code{configure}'s @samp{--srcdir} option.)

The version of GNU @code{make} used to develop this release
is @value{version-make}.

@item @code{cc}
Your system must have a working C compiler.
If it doesn't, you might be able to obtain
a prebuilt binary of some version of @code{gcc}
from the network or on CD-ROM,
perhaps from the FSF@.
The best source of information about binaries
is probably a system-specific Usenet news group,
initially via its FAQ.

@xref{Installation,,Installing GNU CC,gcc,Using and Porting GNU CC},
for more information on prerequisites for installing @code{gcc}.

@item @code{sed}
All UNIX systems have @code{sed}, but some have a broken
version that cannot handle configuring, building, or
installing @code{gcc} or @code{g77}.

The version of GNU @code{sed} used to develop this release
is @value{version-sed}.
(Note that GNU @code{sed} version 3.0 was withdrawn by the
FSF---if you happen to have this version installed, replace
it with version @value{version-sed} immediately.
See a GNU distribution site for further explanation.)

@item @code{root} access or equivalent
To perform the complete installation procedures on a system,
you need to have @code{root} access to that system, or
equivalent access to the @samp{--prefix} directory tree
specified on the @code{configure} command line.

Portions of the procedure (such as configuring and building
@code{g77}) can be performed by any user with enough disk
space and virtual memory.

However, these instructions are oriented towards less-experienced
users who want to install @code{g77} on their own personal
systems.

System administrators with more experience will want to
determine for themselves how they want to modify the
procedures described below to suit the needs of their
installation.

@item @code{autoconf}
The version of GNU @code{autoconf} used to develop this release
is @value{version-autoconf}.

@code{autoconf} is not needed in the typical case of
installing @code{gcc} and @code{g77}.
@xref{Missing tools?}, for information on when it
might be needed and how to work around not having it.

@item @code{bison}
The version of GNU @code{bison} used to develop this release
is @value{version-bison}.

@code{bison} is not needed in the typical case of
installing @code{gcc} and @code{g77}.
@xref{Missing tools?}, for information on when it
might be needed and how to work around not having it.

@item @code{gperf}
The version of GNU @code{gperf} used to develop this release
is @value{version-gperf}.

@code{gperf} is not needed in the typical case of
installing @code{gcc} and @code{g77}.
@xref{Missing tools?}, for information on when it
might be needed and how to work around not having it.

@item @code{makeinfo}
The version of GNU @code{makeinfo} used to develop this release
is @value{version-makeinfo}.

@code{makeinfo} is part of the GNU @code{texinfo} package;
@code{makeinfo} version @value{version-makeinfo}
is distributed as part of
GNU @code{texinfo} version @value{version-texinfo}.

@code{makeinfo} is not needed in the typical case of
installing @code{gcc} and @code{g77}.
@xref{Missing tools?}, for information on when it
might be needed and how to work around not having it.

An up-to-date version of GNU @code{makeinfo} is still convenient
when obtaining a new version of a GNU distribution such as
@code{gcc} or @code{g77},
as it allows you to obtain the @file{.diff.gz} file
instead of the entire @file{.tar.gz} distribution
(assuming you have installed @code{patch}).

@item @code{patch}
The version of GNU @code{patch} used to develop this release
is @value{version-patch}.

Beginning with @code{g77} version 0.5.23, it is no longer
necessary to patch the @code{gcc} back end to build @code{g77}.

An up-to-date version of GNU @code{patch} is still convenient
when obtaining a new version of a GNU distribution such as
@code{gcc} or @code{g77},
as it allows you to obtain the @file{.diff.gz} file
instead of the entire @file{.tar.gz} distribution
(assuming you have installed the tools needed
to rebuild derived files, such as @code{makeinfo}).
@end table

@node Problems Installing
@section Problems Installing
@cindex problems installing
@cindex installation problems

This is a list of problems (and some apparent problems which don't
really mean anything is wrong) that show up when configuring,
building, installing, or porting GNU Fortran.

@xref{Installation Problems,,,gcc,Using and Porting GNU CC},
for more information on installation problems that can afflict
either @code{gcc} or @code{g77}.

@menu
* General Problems::         Problems afflicting most or all systems.
* System-specific Problems:: Problems afflicting particular systems.
* Cross-compiler Problems::  Problems afflicting cross-compilation setups.
@end menu

@node General Problems
@subsection General Problems

These problems can occur on most or all systems.

@menu
* GNU C Required::             Why even ANSI C is not enough.
* Patching GNU CC::            Why @code{gcc} needn't be patched.
* Building GNU CC Necessary::  Why you can't build @emph{just} Fortran.
* Missing strtoul or bsearch:: When linking @samp{f771} fails.
* Cleanup Kills Stage Directories::  For @code{g77} developers.
* LANGUAGES Macro Ignored::    Sometimes @samp{LANGUAGES} is ignored.
@end menu

@node GNU C Required
@subsubsection GNU C Required
@cindex GNU C required
@cindex requirements, GNU C

Compiling @code{g77} requires GNU C, not just ANSI C.
Fixing this wouldn't
be very hard (just tedious), but the code using GNU extensions to
the C language is expected to be rewritten for 0.6 anyway,
so there are no plans for an interim fix.

This requirement does not mean you must already have @code{gcc}
installed to build @code{g77}.
As long as you have a working C compiler, you can use a
bootstrap build to automate the process of first building
@code{gcc} using the working C compiler you have, then building
@code{g77} and rebuilding @code{gcc} using that just-built @code{gcc},
and so on.

@node Patching GNU CC
@subsubsection Patching GNU CC
@cindex patch files
@cindex GBE

@code{g77} no longer requires application of a patch file
to the @code{gcc} compiler tree.
In fact, no such patch file is distributed with @code{g77}.
This is as of version 0.5.23
and @code{egcs} version 1.0.

@node Building GNU CC Necessary
@subsubsection Building GNU CC Necessary
@cindex gcc, building
@cindex building gcc

It should be possible to build the runtime without building @code{cc1}
and other non-Fortran items, but, for now, an easy way to do that
is not yet established.

@node Missing strtoul or bsearch
@subsubsection Missing strtoul or bsearch
@cindex bsearch
@cindex _bsearch
@cindex strtoul
@cindex _strtoul
@cindex undefined reference (_bsearch)
@cindex undefined reference (_strtoul)
@cindex f771, linking error for
@cindex linking error for f771
@cindex ld error for f771
@cindex ld can't find _bsearch
@cindex ld can't find _strtoul
@cindex SunOS4

@emph{Version info:}
The following information does not apply to the
@code{egcs} version of @code{g77}.

On SunOS4 systems, linking the @code{f771} program used to
produce an error message concerning an undefined symbol named
@samp{_strtoul}, because the @samp{strtoul} library function
is not provided on that system.

Other systems have, in the past, been reported to not provide
their own @samp{strtoul} or @samp{bsearch} function.

Some versions @code{g77} tried to default to providing bare-bones
versions of @code{bsearch} and @code{strtoul} automatically,
but every attempt at this has failed for at least one kind of system.

To limit the failures to those few systems actually missing the
required routines, the bare-bones versions are still provided,
in @file{gcc/f/proj.c},
if the appropriate macros are defined.
These are @code{NEED_BSEARCH} for @samp{bsearch} and
@code{NEED_STRTOUL} for @samp{NEED_STRTOUL}.

Therefore, if you are sure your system is missing
@code{bsearch} or @code{strtoul} in its library,
define the relevant macro(s) before building @code{g77}.
This can be done by editing @file{gcc/f/proj.c} and inserting
either or both of the following @samp{#define} statements
before the comment shown:

@smallexample
/* Insert #define statements here.  */

#define NEED_BSEARCH
#define NEED_STRTOUL
@end smallexample

Then, continue configuring and building @code{g77} as usual.

Or, you can define these on the @code{make} command line.
To build with the bundled @code{cc} on SunOS4, for example, try:
@smallexample
make bootstrap BOOT_CFLAGS='-O2 -g -DNEED_STRTOUL'
@end smallexample

If you then encounter problems compiling @file{gcc/f/proj.c},
it might be due to a discrepancy between how @samp{bsearch}
or @samp{strtoul} are defined by that file and how they're
declared by your system's header files.

In that case, you'll have to use some basic knowledge of C
to work around the problem, perhaps by editing @file{gcc/f/proj.c}
somewhat.

@node Cleanup Kills Stage Directories
@subsubsection Cleanup Kills Stage Directories
@cindex stage directories
@cindex make clean

It'd be helpful if @code{g77}'s @file{Makefile.in} or @file{Make-lang.in}
would create the various @file{stage@var{n}} directories and their
subdirectories, so developers and expert installers wouldn't have to
reconfigure after cleaning up.

That help has arrived as of version 0.5.23 of @code{g77}
and version 1.1 of @code{egcs}.
Configuration itself no longer creates any particular directories
that are unique to @code{g77}.
The build procedures in @file{Make-lang.in} take care of
that, on demand.

@node LANGUAGES Macro Ignored
@subsubsection LANGUAGES Macro Ignored
@cindex @samp{LANGUAGES} macro ignored
@cindex ignoring @samp{LANGUAGES} macro

Prior to version 0.5.23 of @code{g77}
and version 1.1 of @code{egcs},
@code{g77} would sometimes ignore
the absence of @samp{f77} and @samp{F77} in the
@samp{LANGUAGES} macro definition used for the
@code{make} command being processed.

As of @code{g77} version 0.5.23
and @code{egcs} version 1.1,
@code{g77} now obeys this macro
in all relevant situations.

However, in versions of @code{gcc} through 2.8.1,
non-@code{g77} portions of @code{gcc},
such as @code{g++},
are known to go ahead and perform various
language-specific activities when their
respective language strings do not appear
in the @samp{LANGUAGES} macro in effect
during that invocation of @code{make}.

It is expected that these remaining problems will
be fixed in a future version of @code{gcc}.

@node System-specific Problems
@subsection System-specific Problems

@cindex AIX
A linker bug on some versions of AIX 4.1 might prevent building
when @code{g77} is built within @code{gcc}.
It might also occur when building within @code{egcs}.
@xref{LINKFAIL}.

@node Cross-compiler Problems
@subsection Cross-compiler Problems
@cindex cross-compiler, problems

@code{g77} has been in alpha testing since September of
1992, and in public beta testing since February of 1995.
Alpha testing was done by a small number of people worldwide on a fairly
wide variety of machines, involving self-compilation in most or
all cases.
Beta testing has been done primarily via self-compilation,
but in more and more cases, cross-compilation (and ``criss-cross
compilation'', where a version of a compiler is built on one machine
to run on a second and generate code that runs on a third) has
been tried and has succeeded, to varying extents.

Generally, @code{g77} can be ported to any configuration to which
@code{gcc}, @code{f2c}, and @code{libf2c} can be ported and made
to work together, aside from the known problems described in this
manual.
If you want to port @code{g77} to a particular configuration,
you should first make sure @code{gcc} and @code{libf2c} can be
ported to that configuration before focusing on @code{g77}, because
@code{g77} is so dependent on them.

Even for cases where @code{gcc} and @code{libf2c} work,
you might run into problems with cross-compilation on certain machines,
for several reasons.

@itemize @bullet
@item
There is one known bug
(a design bug to be fixed in 0.6) that prevents configuration of
@code{g77} as a cross-compiler in some cases,
though there are assumptions made during
configuration that probably make doing non-self-hosting builds
a hassle, requiring manual intervention.

@item
@code{gcc} might still have some trouble being configured
for certain combinations of machines.
For example, it might not know how to handle floating-point
constants.

@item
Improvements to the way @code{libg2c} is built could make
building @code{g77} as a cross-compiler easier---for example,
passing and using @samp{$(LD)} and @samp{$(AR)} in the appropriate
ways.
(This is improved in the @code{egcs} version of @code{g77},
especially as of version 1.1.)

@item
There are still some challenges putting together the right
run-time libraries (needed by @code{libg2c}) for a target
system, depending on the systems involved in the configuration.
(This is a general problem with cross-compilation, and with
@code{gcc} in particular.)
@end itemize

@node Settings
@section Changing Settings Before Building

Here are some internal @code{g77} settings that can be changed
by editing source files in @file{gcc/f/} before building.

This information, and perhaps even these settings, represent
stop-gap solutions to problems people doing various ports
of @code{g77} have encountered.
As such, none of the following information is expected to
be pertinent in future versions of @code{g77}.

@menu
* Larger File Unit Numbers::     Raising @samp{MXUNIT}.
* Always Flush Output::          Synchronizing write errors.
* Maximum Stackable Size::       Large arrays forced off the stack.
* Floating-point Bit Patterns::  Possible programs building @code{g77}
                                   as a cross-compiler.
* Large Initialization::         Large arrays with @code{DATA}
                                   initialization.
* Alpha Problems Fixed::         Problems with 64-bit systems like
                                   Alphas now fixed?
@end menu

@node Larger File Unit Numbers
@subsection Larger File Unit Numbers
@cindex MXUNIT
@cindex unit numbers
@cindex maximum unit number
@cindex illegal unit number
@cindex increasing maximum unit number

As distributed, whether as part of @code{f2c} or @code{g77},
@code{libf2c} accepts file unit numbers only in the range
0 through 99.
For example, a statement such as @samp{WRITE (UNIT=100)} causes
a run-time crash in @code{libf2c}, because the unit number,
100, is out of range.

If you know that Fortran programs at your installation require
the use of unit numbers higher than 99, you can change the
value of the @samp{MXUNIT} macro, which represents the maximum unit
number, to an appropriately higher value.

To do this, edit the file @file{f/runtime/libI77/fio.h} in your
@code{g77} source tree, changing the following line:

@example
#define MXUNIT 100
@end example

Change the line so that the value of @samp{MXUNIT} is defined to be
at least one @emph{greater} than the maximum unit number used by
the Fortran programs on your system.

(For example, a program that does @samp{WRITE (UNIT=255)} would require
@samp{MXUNIT} set to at least 256 to avoid crashing.)

Then build or rebuild @code{g77} as appropriate.

@emph{Note:} Changing this macro has @emph{no} effect on other limits
your system might place on the number of files open at the same time.
That is, the macro might allow a program to do @samp{WRITE (UNIT=100)},
but the library and operating system underlying @code{libf2c} might
disallow it if many other files have already been opened (via @code{OPEN} or
implicitly via @code{READ}, @code{WRITE}, and so on).
Information on how to increase these other limits should be found
in your system's documentation.

@node Always Flush Output
@subsection Always Flush Output
@cindex ALWAYS_FLUSH
@cindex synchronous write errors
@cindex disk full
@cindex flushing output
@cindex fflush()
@cindex I/O, flushing
@cindex output, flushing
@cindex writes, flushing
@cindex NFS
@cindex network file system

Some Fortran programs require output
(writes) to be flushed to the operating system (under UNIX,
via the @code{fflush()} library call) so that errors,
such as disk full, are immediately flagged via the relevant
@code{ERR=} and @code{IOSTAT=} mechanism, instead of such
errors being flagged later as subsequent writes occur, forcing
the previously written data to disk, or when the file is
closed.

Essentially, the difference can be viewed as synchronous error
reporting (immediate flagging of errors during writes) versus
asynchronous, or, more precisely, buffered error reporting
(detection of errors might be delayed).

@code{libg2c} supports flagging write errors immediately when
it is built with the @samp{ALWAYS_FLUSH} macro defined.
This results in a @code{libg2c} that runs slower, sometimes
quite a bit slower, under certain circumstances---for example,
accessing files via the networked file system NFS---but the
effect can be more reliable, robust file I/O.

If you know that Fortran programs requiring this level of precision
of error reporting are to be compiled using the
version of @code{g77} you are building, you might wish to
modify the @code{g77} source tree so that the version of
@code{libg2c} is built with the @samp{ALWAYS_FLUSH} macro
defined, enabling this behavior.

To do this, find this line in @file{f/runtime/f2c.h} in
your @code{g77} source tree:

@example
/* #define ALWAYS_FLUSH */
@end example

Remove the leading @samp{/*@w{ }},
so the line begins with @samp{#define},
and the trailing @samp{@w{ }*/}.

Then build or rebuild @code{g77} as appropriate.

@node Maximum Stackable Size
@subsection Maximum Stackable Size
@vindex FFECOM_sizeMAXSTACKITEM
@cindex code, stack variables
@cindex maximum stackable size
@cindex stack allocation
@cindex segmentation violation
@code{g77}, on most machines, puts many variables and arrays on the stack
where possible, and can be configured (by changing
@samp{FFECOM_sizeMAXSTACKITEM} in @file{gcc/f/com.c}) to force
smaller-sized entities into static storage (saving
on stack space) or permit larger-sized entities to be put on the
stack (which can improve run-time performance, as it presents
more opportunities for the GBE to optimize the generated code).

@emph{Note:} Putting more variables and arrays on the stack
might cause problems due to system-dependent limits on stack size.
Also, the value of @samp{FFECOM_sizeMAXSTACKITEM} has no
effect on automatic variables and arrays.
@xref{But-bugs}, for more information.

@node Floating-point Bit Patterns
@subsection Floating-point Bit Patterns

@cindex cross-compiler, building
@cindex floating-point bit patterns
@cindex bit patterns
The @code{g77} build will crash if an attempt is made to build
it as a cross-compiler
for a target when @code{g77} cannot reliably determine the bit pattern of
floating-point constants for the target.
Planned improvements for version 0.6 of @code{g77}
will give it the capabilities it needs to not have to crash the build
but rather generate correct code for the target.
(Currently, @code{g77}
would generate bad code under such circumstances if it didn't crash
during the build, e.g. when compiling a source file that does
something like @samp{EQUIVALENCE (I,R)} and @samp{DATA R/9.43578/}.)

@node Large Initialization
@subsection Initialization of Large Aggregate Areas

@cindex speed, compiler
@cindex slow compiler
@cindex memory utilization
@cindex large initialization
@cindex aggregate initialization
A warning message is issued when @code{g77} sees code that provides
initial values (e.g. via @code{DATA}) to an aggregate area (@code{COMMON}
or @code{EQUIVALENCE}, or even a large enough array or @code{CHARACTER}
variable)
that is large enough to increase @code{g77}'s compile time by roughly
a factor of 10.

This size currently is quite small, since @code{g77}
currently has a known bug requiring too much memory
and time to handle such cases.
In @file{gcc/f/data.c}, the macro
@samp{FFEDATA_sizeTOO_BIG_INIT_} is defined
to the minimum size for the warning to appear.
The size is specified in storage units,
which can be bytes, words, or whatever, on a case-by-case basis.

After changing this macro definition, you must
(of course) rebuild and reinstall @code{g77} for
the change to take effect.

Note that, as of version 0.5.18, improvements have
reduced the scope of the problem for @emph{sparse}
initialization of large arrays, especially those
with large, contiguous uninitialized areas.
However, the warning is issued at a point prior to
when @code{g77} knows whether the initialization is sparse,
and delaying the warning could mean it is produced
too late to be helpful.

Therefore, the macro definition should not be adjusted to
reflect sparse cases.
Instead, adjust it to generate the warning when densely
initialized arrays begin to cause responses noticeably slower
than linear performance would suggest.

@node Alpha Problems Fixed
@subsection Alpha Problems Fixed

@cindex Alpha, support
@cindex 64-bit systems
@code{g77} used to warn when it was used to compile Fortran code
for a target configuration that is not basically a 32-bit
machine (such as an Alpha, which is a 64-bit machine, especially
if it has a 64-bit operating system running on it).
That was because @code{g77} was known to not work
properly on such configurations.

As of version 0.5.20, @code{g77} is believed to work well
enough on such systems.
So, the warning is no longer needed or provided.

However, support for 64-bit systems, especially in
areas such as cross-compilation and handling of
intrinsics, is still incomplete.
The symptoms
are believed to be compile-time diagnostics rather
than the generation of bad code.
It is hoped that version 0.6 will completely support 64-bit
systems.

@node Quick Start
@section Quick Start
@cindex quick start

@emph{Version info:}
For @code{egcs} users, the following information is
superceded by the @code{egcs} installation instructions.

This procedure configures, builds, and installs @code{g77}
``out of the box'' and works on most UNIX systems.
Each command is identified by a unique number,
used in the explanatory text that follows.
For the most part, the output of each command is not shown,
though indications of the types of responses are given in a
few cases.

To perform this procedure, the installer must be logged
in as user @code{root}.
Much of it can be done while not logged in as @code{root},
and users experienced with UNIX administration should be
able to modify the procedure properly to do so.

Following traditional UNIX conventions, it is assumed that
the source trees for @code{g77} and @code{gcc} will be
placed in @file{/usr/src}.
It also is assumed that the source distributions themselves
already reside in @file{/usr/FSF}, a naming convention
used by the author of @code{g77} on his own system:

@example
/usr/FSF/gcc-@value{version-gcc}.tar.gz
/usr/FSF/g77-@value{version-g77}.tar.gz
@end example

@c (You can use @file{gcc-2.7.2.1.tar.gz} instead, or
@c the equivalent of it obtained by applying the
@c patch distributed as @file{gcc-2.7.2-2.7.2.1.diff.gz}
@c to version 2.7.2 of @code{gcc},
@c if you remember to make the appropriate adjustments in the
@c instructions below.)

@c @cindex SunOS4
@c Users of the following systems should not blindly follow
@c these quick-start instructions, because of problems their
@c systems have coping with straightforward installation of
@c @code{g77}:
@c 
@c @itemize @bullet
@c @item
@c SunOS4
@c @end itemize
@c 
@c Instead, see @ref{Complete Installation}, for detailed information
@c on how to configure, build, and install @code{g77} for your
@c particular system.
@c Also, see @ref{Trouble,,Known Causes of Trouble with GNU Fortran},
@c for information on bugs and other problems known to afflict the
@c installation process, and how to report newly discovered ones.
@c 
@c If your system is @emph{not} on the above list, and @emph{is}
@c a UNIX system or one of its variants, you should be able to
@c follow the instructions below.

If you vary @emph{any} of the steps below, you might run into
trouble, including possibly breaking existing programs for
other users of your system.
Before doing so, it is wise to review the explanations of some
of the steps.
These explanations follow this list of steps.

@example
sh[ 1]# @kbd{cd /usr/src}
@set source-dir 1
sh[ 2]# @kbd{gunzip -c < /usr/FSF/gcc-@value{version-gcc}.tar.gz | tar xf -}
[Might say "Broken pipe"...that is normal on some systems.]
@set unpack-gcc 2
sh[ 3]# @kbd{gunzip -c < /usr/FSF/g77-@value{version-g77}.tar.gz | tar xf -}
["Broken pipe" again possible.]
@set unpack-g77 3
sh[ 4]# @kbd{ln -s gcc-@value{version-gcc} gcc}
@set link-gcc 4
sh[ 5]# @kbd{ln -s g77-@value{version-g77} g77}
@set link-g77 5
sh[ 6]# @kbd{mv -i g77/* gcc}
[No questions should be asked by mv here; or, you made a mistake.]
@set merge-g77 6
sh[ 7]# @kbd{cd gcc}
sh[ 8]# @kbd{./configure --prefix=/usr}
[Do not do the above if gcc is not installed in /usr/bin.
You might need a different @kbd{--prefix=@dots{}}, as
described below.]
@set configure-gcc 8
sh[ 9]# @kbd{make bootstrap}
[This takes a long time, and is where most problems occur.]
@set build-gcc 9
sh[10]# @kbd{make compare}
[This verifies that the compiler is `sane'.
If any files are printed, you have likely found a g77 bug.]
@set compare-gcc 10
sh[11]# @kbd{rm -fr stage1}
@set rm-stage1 11
sh[12]# @kbd{make -k install}
[The actual installation.]
@set install-g77 12
sh[13]# @kbd{g77 -v}
[Verify that g77 is installed, obtain version info.]
@set show-version 13
sh[14]#
@set end-procedure 14
@end example

@xref{Updating Documentation,,Updating Your Info Directory}, for
information on how to update your system's top-level @code{info}
directory to contain a reference to this manual, so that
users of @code{g77} can easily find documentation instead
of having to ask you for it.

Elaborations of many of the above steps follows:

@table @asis
@item Step @value{source-dir}: @kbd{cd /usr/src}
You can build @code{g77} pretty much anyplace.
By convention, this manual assumes @file{/usr/src}.
It might be helpful if other users on your system
knew where to look for the source code for the
installed version of @code{g77} and @code{gcc} in any case.

@c @item Step @value{unpack-gcc}: @kbd{gunzip -d @dots{}}
@c Here, you might wish to use @file{gcc-2.7.2.1.tar.gz}
@c instead, or apply @file{gcc-2.7.2-2.7.2.1.diff.gz} to achieve
@c similar results.

@item Step @value{unpack-g77}: @kbd{gunzip -d < /usr/FSF/g77-@value{version-g77}.tar.gz | tar xf -}
It is not always necessary to obtain the latest version of
@code{g77} as a complete @file{.tar.gz} file if you have
a complete, earlier distribution of @code{g77}.
If appropriate, you can unpack that earlier
version of @code{g77}, and then apply the appropriate patches
to achieve the same result---a source tree containing version
@value{version-g77} of @code{g77}.

@item Step @value{link-gcc}: @kbd{ln -s gcc-@value{version-gcc} gcc}
@item Step @value{link-g77}: @kbd{ln -s g77-@value{version-g77} g77}
These commands mainly help reduce typing,
and help reduce visual clutter in examples
in this manual showing what to type to install @code{g77}.

@c Of course, if appropriate, @kbd{ln -s gcc-2.7.2.1 gcc} or
@c similar.

@xref{Unpacking}, for information on
using distributions of @code{g77} made by organizations
other than the FSF.

@item Step @value{merge-g77}: @kbd{mv -i g77/* gcc}
After doing this, you can, if you like, type
@samp{rm g77} and @samp{rmdir g77-@value{version-g77}} to remove
the empty directory and the symbol link to it.
But, it might be helpful to leave them around as
quick reminders of which version(s) of @code{g77} are
installed on your system.

@xref{Unpacking}, for information
on the contents of the @file{g77} directory (as merged
into the @file{gcc} directory).

@item Step @value{configure-gcc}: @kbd{./configure --prefix=/usr}
This is where you specify that
the @file{g77} and @file{gcc} executables are to be
installed in @file{/usr/bin/},
the @code{g77} and @code{gcc} documentation is
to be installed in @file{/usr/info/} and @file{/usr/man/},
and so on.

You should ensure that any existing installation of the @file{gcc}
executable is in @file{/usr/bin/}.

However, if that existing version of @code{gcc} is not @value{version-gcc},
or if you simply wish to avoid risking overwriting it with a
newly built copy of the same version,
you can specify @samp{--prefix=/usr/local}
(which is the default)
or some other path,
and invoke the newly installed version
directly from that path's @file{bin} directory.

@xref{Where to Install,,Where in the World Does Fortran (and GNU CC) Go?},
for more information on determining where to install @code{g77}.
@xref{Configuring gcc}, for more information on the
configuration process triggered by invoking the @file{./configure}
script.

@item Step @value{build-gcc}: @kbd{make bootstrap}
@xref{Installation,,Installing GNU CC,
gcc,Using and Porting GNU CC}, for information
on the kinds of diagnostics you should expect during
this procedure.

@xref{Building gcc}, for complete @code{g77}-specific
information on this step.

@item Step @value{compare-gcc}: @kbd{make compare}
@xref{Bug Lists,,Where to Port Bugs}, for information
on where to report that you observed files
having different contents during this
phase.

@xref{Bug Reporting,,How to Report Bugs}, for
information on @emph{how} to report bugs like this.

@item Step @value{rm-stage1}: @kbd{rm -fr stage1}
You don't need to do this, but it frees up disk space.

@item Step @value{install-g77}: @kbd{make -k install}
If this doesn't seem to work, try:

@example
make -k install install-libf77
@end example

Or, make sure you're using GNU @code{make}.

@xref{Installation of Binaries}, for more information.

@xref{Updating Documentation,,Updating Your Info Directory},
for information on entering this manual into your
system's list of texinfo manuals.

@item Step @value{show-version}: @kbd{g77 -v}
If this command prints approximately 25 lines of output,
including the GNU Fortran Front End version number (which
should be the same as the version number for the version
of @code{g77} you just built and installed) and the
version numbers for the three parts of the @code{libf2c}
library (@code{libF77}, @code{libI77}, @code{libU77}), and
those version numbers are all in agreement, then there is
a high likelihood that the installation has been successfully
completed.

You might consider doing further testing.
For example, log in as a non-privileged user, then create
a small Fortran program, such as:

@example
      PROGRAM SMTEST
      DO 10 I=1, 10
         PRINT *, 'Hello World #', I
10    CONTINUE
      END
@end example

Compile, link, and run the above program, and, assuming you named
the source file @file{smtest.f}, the session should look like this:

@example
sh# @kbd{g77 -o smtest smtest.f}
sh# @kbd{./smtest}
 Hello World # 1
 Hello World # 2
 Hello World # 3
 Hello World # 4
 Hello World # 5
 Hello World # 6
 Hello World # 7
 Hello World # 8
 Hello World # 9
 Hello World # 10
sh#
@end example

If invoking @code{g77} doesn't seem to work,
the problem might be that you've installed it in
a location that is not in your shell's search path.
For example, if you specified @samp{--prefix=/gnu},
and @file{/gnu/bin} is not in your @code{PATH}
environment variable,
you must explicitly specify the location of the compiler
via @kbd{/gnu/bin/g77 -o smtest smtest.f}.

After proper installation, you don't
need to keep your gcc and g77 source and build directories
around anymore.
Removing them can free up a lot of disk space.
@end table

@node Complete Installation
@section Complete Installation

@emph{Version info:}
For @code{egcs} users, the following information is
mostly superceded by the @code{egcs} installation instructions.

Here is the complete @code{g77}-specific information on how
to configure, build, and install @code{g77}.

@menu
* Unpacking::
* Merging Distributions::
* Where to Install::
* Configuring gcc::
* Building gcc::
* Pre-installation Checks::
* Installation of Binaries::
* Updating Documentation::
* Missing tools?::
@end menu

@node Unpacking
@subsection Unpacking
@cindex unpacking distributions
@cindex distributions, unpacking
@cindex code, source
@cindex source code
@cindex source tree
@cindex packages

The @code{gcc} source distribution is a stand-alone distribution.
It is designed to be unpacked (producing the @code{gcc}
source tree) and built as is, assuming certain
prerequisites are met (including the availability of compatible
UNIX programs such as @code{make}, @code{cc}, and so on).

However, before building @code{gcc}, you will want to unpack
and merge the @code{g77} distribution in with it, so that you
build a Fortran-capable version of @code{gcc}, which includes
the @code{g77} command, the necessary run-time libraries,
and this manual.

Unlike @code{gcc}, the @code{g77} source distribution
is @emph{not} a stand-alone distribution.
It is designed to be unpacked and, afterwards, immediately merged
into an applicable @code{gcc} source tree.
That is, the @code{g77} distribution @emph{augments} a
@code{gcc} distribution---without @code{gcc}, generally
only the documentation is immediately usable.

A sequence of commands typically used to unpack @code{gcc}
and @code{g77} is:

@example
sh# @kbd{cd /usr/src}
sh# @kbd{gunzip -c /usr/FSF/gcc-@value{version-gcc}.tar.gz | tar xf -}
sh# @kbd{gunzip -c /usr/FSF/g77-@value{version-g77}.tar.gz | tar xf -}
sh# @kbd{ln -s gcc-@value{version-gcc} gcc}
sh# @kbd{ln -s g77-@value{version-g77} g77}
sh# @kbd{mv -i g77/* gcc}
@end example

@emph{Notes:} The commands beginning with @samp{gunzip@dots{}} might
print @samp{Broken pipe@dots{}} as they complete.
That is nothing to worry about, unless you actually
@emph{hear} a pipe breaking.
The @code{ln} commands are helpful in reducing typing
and clutter in installation examples in this manual.
Hereafter, the top level of @code{gcc} source tree is referred to
as @file{gcc}, and the top level of just the @code{g77}
source tree (prior to issuing the @code{mv} command, above)
is referred to as @file{g77}.

There are three top-level names in a @code{g77} distribution:

@example
g77/COPYING.g77
g77/README.g77
g77/f
@end example

All three entries should be moved (or copied) into a @code{gcc}
source tree (typically named after its version number and
as it appears in the FSF distributions---e.g. @file{gcc-@value{version-gcc}}).

@file{g77/f} is the subdirectory containing all of the
code, documentation, and other information that is specific
to @code{g77}.
The other two files exist to provide information on @code{g77}
to someone encountering a @code{gcc} source tree with @code{g77}
already present, who has not yet read these installation
instructions and thus needs help understanding that the
source tree they are looking at does not come from a single
FSF distribution.
They also help people encountering an unmerged @code{g77} source
tree for the first time.

@cindex modifying @code{g77}
@cindex code, modifying
@cindex Pentium optimizations
@cindex optimizations, Pentium
@emph{Note:} Please use @strong{only} @code{gcc} and @code{g77}
source trees as distributed by the FSF.
Use of modified versions is likely to result in problems that appear to be
in the @code{g77} code but, in fact, are not.
Do not use such modified versions
unless you understand all the differences between them and the versions
the FSF distributes---in which case you should be able to modify the
@code{g77} (or @code{gcc}) source trees appropriately so @code{g77}
and @code{gcc} can coexist as they do in the stock FSF distributions.

@node Merging Distributions
@subsection Merging Distributions
@cindex merging distributions
@cindex @code{gcc} versions supported by @code{g77}
@cindex versions of @code{gcc}
@cindex support for @code{gcc} versions

After merging the @code{g77} source tree into the @code{gcc} source tree,
you have put together a complete @code{g77} source tree.

@cindex gcc version numbering
@cindex version numbering
@cindex g77 version number
@cindex GNU version numbering
As of version 0.5.23, @code{g77} no longer modifies
the version number of @code{gcc},
nor does it patch @code{gcc} itself.

@code{g77} still depends on being merged with an
appropriate version of @code{gcc}.
For version @value{version-g77} of @code{g77},
the specific version of @code{gcc} supported is @value{version-gcc}.

However, other versions of @code{gcc} might be suitable
``hosts'' for this version of @code{g77}.

GNU version numbers make it easy to figure out whether a
particular version of a distribution is newer or older than
some other version of that distribution.
The format is,
generally, @var{major}.@var{minor}.@var{patch}, with
each field being a decimal number.
(You can safely ignore
leading zeros; for example, 1.5.3 is the same as 1.5.03.)
The @var{major} field only increases with time.
The other two fields are reset to 0 when the field to
their left is incremented; otherwise, they, too, only
increase with time.
So, version 2.6.2 is newer than version 2.5.8, and
version 3.0 is newer than both.
(Trailing @samp{.0} fields often are omitted in
announcements and in names for distributions and
the directories they create.)

If your version of @code{gcc} is older than the oldest version
supported by @code{g77}
(as casually determined by listing the contents of @file{gcc/f/INSTALL/},
which contains these installation instructions in plain-text format),
you should obtain a newer, supported version of @code{gcc}.
(You could instead obtain an older version of @code{g77},
or try and get your @code{g77} to work with the old
@code{gcc}, but neither approach is recommended, and
you shouldn't bother reporting any bugs you find if you
take either approach, because they're probably already
fixed in the newer versions you're not using.)

If your version of @code{gcc} is newer than the newest version
supported by @code{g77}, it is possible that your @code{g77}
will work with it anyway.
If the version number for @code{gcc} differs only in the
@var{patch} field, you might as well try that version of @code{gcc}.
Since it has the same @var{major} and @var{minor} fields,
the resulting combination is likely to work.

So, for example, if a particular version of @code{g77} has support for
@code{gcc} versions 2.8.0 and 2.8.1,
it is likely that @file{gcc-2.8.2} would work well with @code{g77}.

However, @file{gcc-2.9.0} would almost certainly
not work with that version of @code{g77}
without appropriate modifications,
so a new version of @code{g77} would be needed (and you should
wait for it rather than bothering the maintainers---@pxref{Changes,,
User-Visible Changes}).

@cindex distributions, why separate
@cindex separate distributions
@cindex why separate distributions
This complexity is the result of @code{gcc} and @code{g77} being
separate distributions.
By keeping them separate, each product is able to be independently
improved and distributed to its user base more frequently.

However, the GBE interface defined by @code{gcc} typically
undergoes some incompatible changes at least every time the
@var{minor} field of the version number is incremented,
and such changes require corresponding changes to
the @code{g77} front end (FFE).

@c @pindex config-lang.in
@c @emph{Note:} @code{g77}'s configuration file @file{gcc/f/config-lang.in}
@c sometimes ensures that the source code for the version of @code{gcc}
@c being configured has at least one indication of being an appropriate
@c version as required specifically by @code{g77}.
@c This configuration-time
@c checking should catch failures to use the proper version of @code{gcc} and,
@c if so caught, should abort the configuration with an explanation.
@c @emph{Please} do not try to disable this check,
@c otherwise @code{g77} might well appear to build
@c and install correctly, and even appear to compile correctly,
@c but could easily produce broken code.

@node Where to Install
@subsection Where in the World Does Fortran (and GNU CC) Go?
@cindex language f77 not recognized
@cindex gcc will not compile Fortran programs

Before configuring, you should make sure you know
where you want the @code{g77} and @code{gcc}
binaries to be installed after they're built,
because this information is given to the configuration
tool and used during the build itself.

A @code{g77} installation normally includes installation of
a Fortran-aware version of @code{gcc}, so that the @code{gcc}
command recognizes Fortran source files and knows how to compile
them.

For this to work, the version of @code{gcc} that you will be building
as part of @code{g77} @strong{must} be installed as the ``active''
version of @code{gcc} on the system.

Sometimes people make the mistake of installing @code{gcc} as
@file{/usr/local/bin/gcc},
leaving an older, non-Fortran-aware version in @file{/usr/bin/gcc}.
(Or, the opposite happens.)
This can result in @code{gcc} being unable to compile Fortran
source files,
because when the older version of @code{gcc} is invoked,
it complains that it does not
recognize the language, or the file name suffix.

So, determine whether @code{gcc} already is installed on your system,
and, if so, @emph{where} it is installed, and prepare to configure the
new version of @code{gcc} you'll be building so that it installs
over the existing version of @code{gcc}.

You might want to back up your existing copy of @file{/usr/bin/gcc}, and
the entire @file{/usr/lib} directory, before
you perform the actual installation (as described in this manual).

Existing @code{gcc} installations typically are
found in @file{/usr} or @file{/usr/local}.
(This means the commands are installed in @file{/usr/bin} or
@file{/usr/local/bin},
the libraries in @file{/usr/lib} or @file{/usr/local/lib},
and so on.)

If you aren't certain where the currently
installed version of @code{gcc} and its
related programs reside, look at the output
of this command:

@example
gcc -v -o /tmp/delete-me -xc /dev/null -xnone
@end example

All sorts of interesting information on the locations of various
@code{gcc}-related programs and data files should be visible
in the output of the above command.
(The output also is likely to include a diagnostic from
the linker, since there's no @samp{main_()} function.)
However, you do have to sift through it yourself; @code{gcc}
currently provides no easy way to ask it where it is installed
and where it looks for the various programs and data files it
calls on to do its work.

Just @emph{building} @code{g77} should not overwrite any installed
programs---but, usually, after you build @code{g77}, you will want
to install it, so backing up anything it might overwrite is
a good idea.
(This is true for any package, not just @code{g77},
though in this case it is intentional that @code{g77} overwrites
@code{gcc} if it is already installed---it is unusual that
the installation process for one distribution intentionally
overwrites a program or file installed by another distribution,
although, in this case, @code{g77} is an augmentation of the
@code{gcc} distribution.)

Another reason to back up the existing version first,
or make sure you can restore it easily, is that it might be
an older version on which other users have come to depend
for certain behaviors.
However, even the new version of @code{gcc} you install
will offer users the ability to specify an older version of
the actual compilation programs if desired, and these
older versions need not include any @code{g77} components.
@xref{Target Options,,Specifying Target Machine and Compiler Version,
gcc,Using and Porting GNU CC}, for information on the @samp{-V}
option of @code{gcc}.

@node Configuring gcc
@subsection Configuring GNU CC

@code{g77} is configured automatically when you configure
@code{gcc}.
There are two parts of @code{g77} that are configured in two
different ways---@code{g77}, which ``camps on'' to the
@code{gcc} configuration mechanism, and @code{libg2c}, which
uses a variation of the GNU @code{autoconf} configuration
system.

Generally, you shouldn't have to be concerned with
either @code{g77} or @code{libg2c} configuration, unless
you're configuring @code{g77} as a cross-compiler.
In this case, the @code{libg2c} configuration, and possibly the
@code{g77} and @code{gcc} configurations as well,
might need special attention.
(This also might be the case if you're porting @code{gcc} to
a whole new system---even if it is just a new operating system
on an existing, supported CPU.)

To configure the system, see
@ref{Installation,,Installing GNU CC,gcc,Using and Porting GNU CC},
following the instructions for running @file{./configure}.
Pay special attention to the @samp{--prefix=} option, which
you almost certainly will need to specify.

(Note that @code{gcc} installation information is provided
as a plain-text file in @file{gcc/INSTALL}.)

The information printed by the invocation of @file{./configure}
should show that the @file{f} directory (the Fortran language)
has been configured.
If it does not, there is a problem.

@emph{Note:} Configuring with the @samp{--srcdir} argument,
or by starting in an empty directory
and typing a command such as @kbd{../gcc/configure} to
build with separate build and source directories,
is known to work with GNU @code{make},
but it is known to not work with other variants of @code{make}.
Irix5.2 and SunOS4.1 versions of @code{make} definitely
won't work outside the source directory at present.

@code{g77}'s portion of the @file{configure} script
used to issue a warning message about this
when configuring for building binaries outside the source directory,
but no longer does this as of version 0.5.23.

Instead, @code{g77} simply rejects most common attempts
to build it using a non-GNU @code{make} when the
build directory is not the same as the source directory,
issuing an explanatory diagnostic.

@node Building gcc
@subsection Building GNU CC
@cindex building @code{gcc}
@cindex building @code{g77}

@vindex LANGUAGES
Building @code{g77} requires building enough of @code{gcc} that
these instructions assume you're going to build all of
@code{gcc}, including @code{g++}, @code{protoize}, and so on.
You can save a little time and disk space by changes the
@samp{LANGUAGES} macro definition in @code{gcc/Makefile.in}
or @code{gcc/Makefile}, but if you do that, you're on your own.
One change is almost @emph{certainly} going to cause failures:
removing @samp{c} or @samp{f77} from the definition of the
@samp{LANGUAGES} macro.

After configuring @code{gcc}, which configures @code{g77} and
@code{libg2c} automatically, you're ready to start the actual
build by invoking @code{make}.

@pindex configure
@emph{Note:} You @strong{must} have run the @file{configure}
script in @code{gcc} before you run @code{make},
even if you're using an already existing @code{gcc} development directory,
because @file{./configure} does the work to recognize that you've added
@code{g77} to the configuration.

There are two general approaches to building GNU CC from
scratch:

@table @dfn
@item bootstrap
This method uses minimal native system facilities to
build a barebones, unoptimized @code{gcc}, that is then
used to compile (``bootstrap'') the entire system.

@item straight
This method assumes a more complete native system
exists, and uses that just once to build the entire
system.
@end table

On all systems without a recent version of @code{gcc}
already installed, the @i{bootstrap} method must be
used.
In particular, @code{g77} uses extensions to the C
language offered, apparently, only by @code{gcc}.

On most systems with a recent version of @code{gcc}
already installed, the @i{straight} method can be
used.
This is an advantage, because it takes less CPU time
and disk space for the build.
However, it does require that the system have fairly
recent versions of many GNU programs and other
programs, which are not enumerated here.

@menu
* Bootstrap Build::  For all systems.
* Straight Build::   For systems with a recent version of @code{gcc}.
@end menu

@node Bootstrap Build
@subsubsection Bootstrap Build
@cindex bootstrap build
@cindex build, bootstrap

A complete bootstrap build is done by issuing a command
beginning with @samp{make bootstrap @dots{}}, as
described in @ref{Installation,,Installing GNU CC,
gcc,Using and Porting GNU CC}.
This is the most reliable form of build, but it does require
the most disk space and CPU time, since the complete system
is built twice (in Stages 2 and 3), after an initial build
(during Stage 1) of a minimal @code{gcc} compiler using
the native compiler and libraries.

You might have to, or want to, control the way a bootstrap
build is done by entering the @code{make} commands to build
each stage one at a time, as described in the @code{gcc}
manual.
For example, to save time or disk space, you might want
to not bother doing the Stage 3 build, in which case you
are assuming that the @code{gcc} compiler you have built
is basically sound (because you are giving up the opportunity
to compare a large number of object files to ensure they're
identical).

To save some disk space during installation, after Stage 2
is built, you can type @samp{rm -fr stage1} to remove the
binaries built during Stage 1.

Also, @xref{Installation,,Installing GNU CC,gcc,Using and Porting GNU CC},
for important information on building @code{gcc} that is
not described in this @code{g77} manual.
For example, explanations of diagnostic messages
and whether they're expected, or indicate trouble,
are found there.

@node Straight Build
@subsubsection Straight Build
@cindex straight build
@cindex build, straight

If you have a recent version of @code{gcc}
already installed on your system, and if you're
reasonably certain it produces code that is
object-compatible with the version of @code{gcc}
you want to build as part of building @code{g77},
you can save time and disk space by doing a straight
build.

To build just the compilers along with the
necessary run-time libraries, issue the following
command:

@example
make -k CC=gcc
@end example

If you run into problems using this method, you have
two options:

@itemize @bullet
@item
Abandon this approach and do a bootstrap build.

@item
Try to make this approach work by diagnosing the
problems you're running into and retrying.
@end itemize

Especially if you do the latter, you might consider
submitting any solutions as bug/fix reports.
@xref{Trouble,,Known Causes of Trouble with GNU Fortran}.

However, understand that many problems preventing a
straight build from working are not @code{g77} problems,
and, in such cases, are not likely to be addressed in
future versions of @code{g77}.
Consider treating them as @code{gcc} bugs instead.

@node Pre-installation Checks
@subsection Pre-installation Checks
@cindex pre-installation checks
@cindex installing, checking before

Before installing the system, which includes installing
@code{gcc}, you might want to do some minimum checking
to ensure that some basic things work.

Here are some commands you can try, and output typically
printed by them when they work:

@example
sh# @kbd{cd /usr/src/gcc}
sh# @kbd{./g77 -B./ -v}
g77 version @value{version-g77}
Driving: ./g77 -B./ -v -c -xf77-version /dev/null -xnone
Reading specs from ./specs
gcc version @value{version-gcc}
 cpp -lang-c -v -isystem ./include -undef -D__GNUC__=2 @dots{}
GNU CPP version @value{version-gcc} (Alpha GNU/Linux with ELF)
#include "..." search starts here:
#include <...> search starts here:
 include
 /usr/alpha-linux/include
 /usr/lib/gcc-lib/alpha-linux/@value{version-gcc}/include
 /usr/include
End of search list.
 ./f771 -fnull-version -quiet -dumpbase g77-version.f -version @dots{}
GNU F77 version @value{version-gcc} (alpha-linux) compiled @dots{}
GNU Fortran Front End version @value{version-g77}
 as -nocpp -o /tmp/cca14485.o /tmp/cca14485.s
 ld -m elf64alpha -G 8 -O1 -dynamic-linker /lib/ld-linux.so.2 @dots{}
 /tmp/cca14485
__G77_LIBF77_VERSION__: @value{version-g77}
@@(#)LIBF77 VERSION 19970919
__G77_LIBI77_VERSION__: @value{version-g77}
@@(#) LIBI77 VERSION pjw,dmg-mods 19980405
__G77_LIBU77_VERSION__: @value{version-g77}
@@(#) LIBU77 VERSION 19970919
sh# @kbd{./xgcc -B./ -v -o /tmp/delete-me -xc /dev/null -xnone}
Reading specs from ./specs
gcc version @value{version-gcc}
 ./cpp -lang-c -v -isystem ./include -undef @dots{}
GNU CPP version @value{version-gcc} (Alpha GNU/Linux with ELF)
#include "..." search starts here:
#include <...> search starts here:
 include
 /usr/alpha-linux/include
 /usr/lib/gcc-lib/alpha-linux/@value{version-gcc}/include
 /usr/include
End of search list.
 ./cc1 /tmp/cca18063.i -quiet -dumpbase null.c -version @dots{}
GNU C version @value{version-gcc} (alpha-linux) compiled @dots{}
 as -nocpp -o /tmp/cca180631.o /tmp/cca18063.s
 ld -m elf64alpha -G 8 -O1 -dynamic-linker /lib/ld-linux.so.2 @dots{}
/usr/lib/crt1.o: In function `_start':
../sysdeps/alpha/elf/start.S:77: undefined reference to `main'
../sysdeps/alpha/elf/start.S:77: undefined reference to `main'
sh#
@end example

(Note that long lines have been truncated, and @samp{@dots{}}
used to indicate such truncations.)

The above two commands test whether @code{g77} and @code{gcc},
respectively, are able to compile empty (null) source files,
whether invocation of the C preprocessor works, whether libraries
can be linked, and so on.

If the output you get from either of the above two commands
is noticeably different, especially if it is shorter or longer
in ways that do not look consistent with the above sample
output, you probably should not install @code{gcc} and @code{g77}
until you have investigated further.

For example, you could try compiling actual applications and
seeing how that works.
(You might want to do that anyway, even if the above tests
work.)

To compile using the not-yet-installed versions of @code{gcc}
and @code{g77}, use the following commands to invoke them.

To invoke @code{g77}, type:

@example
/usr/src/gcc/g77 -B/usr/src/gcc/ @dots{}
@end example

To invoke @code{gcc}, type:

@example
/usr/src/gcc/xgcc -B/usr/src/gcc/ @dots{}
@end example

@node Installation of Binaries
@subsection Installation of Binaries
@cindex installation of binaries
@cindex @code{g77}, installation of
@cindex @code{gcc}, installation of

After configuring, building, and testing @code{g77} and @code{gcc},
when you are ready to install them on your system, type:

@example
make -k CC=gcc install
@end example

As described in @ref{Installation,,Installing GNU CC,
gcc,Using and Porting GNU CC}, the values for
the @samp{CC} and @samp{LANGUAGES} macros should
be the same as those you supplied for the build
itself.

So, the details of the above command might vary
if you used a bootstrap build (where you might be
able to omit both definitions, or might have to
supply the same definitions you used when building
the final stage) or if you deviated from the
instructions for a straight build.

If the above command does not install @file{libg2c.a}
as expected, try this:

@example
make -k @dots{} install install-libf77
@end example

We don't know why some non-GNU versions of @code{make} sometimes
require this alternate command, but they do.
(Remember to supply the appropriate definition for @samp{CC}
where you see @samp{@dots{}} in the above command.)

Note that using the @samp{-k} option tells @code{make} to
continue after some installation problems, like not having
@code{makeinfo} installed on your system.
It might not be necessary for your system.

@emph{Note:} @code{g77} no longer installs
files not directly part of @code{g77},
such as @file{/usr/bin/f77}, @file{/usr/lib/libf2c.a},
and @file{/usr/include/f2c.h}, or their
@file{/usr/local} equivalents.

@xref{Distributing Binaries}, for information on
how to accommodate systems with no existing non-@code{g77}
@code{f77} compiler and systems with @code{f2c} installed.

@node Updating Documentation
@subsection Updating Your Info Directory
@cindex updating info directory
@cindex info, updating directory
@cindex directory, updating info
@pindex /usr/info/dir
@pindex g77.info
@cindex texinfo
@cindex documentation

As part of installing @code{g77}, you should make sure users
of @code{info} can easily access this manual on-line.

@code{g77} does this automatically by
invoking the @code{install-info} command
when you use @samp{make install} to install @code{g77}.

If that fails, or if the @code{info} directory
it updates is not the one normally accessed by users,
consider invoking it yourself.
For example:

@smallexample
install-info --info-dir=/usr/info /usr/info/g77.info
@end smallexample

The above example assumes the @code{g77} documentation
already is installed in @file{/usr/info}
and that @file{/usr/info/dir} is the file
you wish to update.
Adjust the command accordingly,
if those assumptions are wrong.

@node Missing tools?
@subsection Missing tools?
@cindex command missing
@cindex command not found
@cindex file not found
@cindex not found

A build of @code{gcc} might fail due to one or more tools
being called upon by @code{make}
(during the build or install process),
when those tools are not installed on your system.

This situation can result from any of the following actions
(performed by you or someone else):

@itemize @bullet
@item
Changing the source code or documentation yourself
(as a developer or technical writer).

@item
Applying a patch that changes the source code or documentation
(including, sometimes, the official patches distributed by
the FSF).

@item
Deleting the files that are created by the (missing) tools.

The @samp{make maintainer-clean} command is supposed
to delete these files, so invoking this command without
having all the appropriate tools installed is not recommended.

@item
Creating the source directory using a method that
does not preserve the date-time-modified information
in the original distribution.

For example, the UNIX @samp{cp -r} command copies a
directory tree without preserving the date-time-modified
information.
Use @samp{cp -pr} instead.
@end itemize

The reason these activities cause @code{make} to try and
invoke tools that it probably wouldn't when building
from a perfectly ``clean'' source directory containing
@code{gcc} and @code{g77} is that some files in the
source directory (and the corresponding distribution)
aren't really source files, but @emph{derived} files
that are produced by running tools with the corresponding
source files as input.
These derived files @dfn{depend}, in @code{make} terminology,
on the corresponding source files.

@code{make} determines that a file that depends on another
needs to be updated if the date-time-modified information for
the source file shows that it is newer than the corresponding
information for the derived file.

If it makes that determination, @code{make} runs the appropriate
commands (specified in the ``Makefile'') to update the
derived file, and this process typically calls upon one or
more installed tools to do the work.

The ``safest'' approach to dealing with this situation
is to recreate the @code{gcc} and @code{g77} source
directories from complete @code{gcc} and @code{g77} distributions
known to be provided by the FSF.

Another fairly ``safe'' approach is to simply install
the tools you need to complete the build process.
This is especially appropriate if you've changed the
source code or applied a patch to do so.

However, if you're certain that the problem is limited
entirely to incorrect date-time-modified information,
that there are no discrepancies between the contents of
source files and files derived from them in the source
directory, you can often update the date-time-modified
information for the derived files to work around the
problem of not having the appropriate tools installed.

On UNIX systems, the simplest way to update the date-time-modified
information of a file is to use the use the @samp{touch}
command.

How to use @samp{touch} to update the derived files
updated by each of the tools is described below.
@emph{Note:} New versions of @code{g77} might change the set of
files it generates by invoking each of these tools.
If you cannot figure
out for yourself how to handle such a situation, try an
older version of @code{g77} until you find someone who can
(or until you obtain and install the relevant tools).

@menu
* autoconf: Missing autoconf?.
* bison: Missing bison?.
* gperf: Missing gperf?.
* makeinfo: Missing makeinfo?.
@end menu

@node Missing autoconf?
@subsubsection Missing @code{autoconf}?
@cindex @code{autoconf}
@cindex missing @code{autoconf}

If you cannot install @code{autoconf}, make sure you have started
with a @emph{fresh} distribution of @code{gcc} and @code{g77},
do @emph{not} do @samp{make maintainer-clean}, and, to ensure that
@code{autoconf} is not invoked by @code{make} during the build,
type these commands:

@example
sh# @kbd{cd gcc/f/runtime}
sh# @kbd{touch configure libU77/configure}
sh# @kbd{cd ../../..}
sh#
@end example

@node Missing bison?
@subsubsection Missing @code{bison}?
@cindex @code{bison}
@cindex missing @code{bison}

If you cannot install @code{bison}, make sure you have started
with a @emph{fresh} distribution of @code{gcc}, do @emph{not}
do @samp{make maintainer-clean}, and, to ensure that
@code{bison} is not invoked by @code{make} during the build,
type these commands:

@example
sh# @kbd{cd gcc}
sh# @kbd{touch bi-parser.c bi-parser.h c-parse.c c-parse.h cexp.c}
sh# @kbd{touch cp/parse.c cp/parse.h objc-parse.c}
sh# @kbd{cd ..}
sh#
@end example

@node Missing gperf?
@subsubsection Missing @code{gperf}?
@cindex @code{gperf}
@cindex missing @code{gperf}

If you cannot install @code{gperf}, make sure you have started
with a @emph{fresh} distribution of @code{gcc}, do @emph{not}
do @samp{make maintainer-clean}, and, to ensure that
@code{gperf} is not invoked by @code{make} during the build,
type these commands:

@example
sh# @kbd{cd gcc}
sh# @kbd{touch c-gperf.h}
sh# @kbd{cd ..}
sh#
@end example

@node Missing makeinfo?
@subsubsection Missing @code{makeinfo}?
@cindex @code{makeinfo}
@cindex missing @code{makeinfo}
@cindex @code{libg2c.a} not found
@cindex missing @code{libg2c.a}

If @code{makeinfo} is needed but unavailable
when installing (via @code{make install}),
some files, like @file{libg2c.a},
might not be installed,
because once @code{make} determines that it cannot
invoke @code{makeinfo}, it cancels any further processing.

If you cannot install @code{makeinfo}, an easy work-around is to
specify @samp{MAKEINFO=true} on the @code{make} command line,
or to specify the @samp{-k} option (@kbd{make -k install}).

Another approach is to force the relevant files to be up-to-date
by typing these commands and then re-trying the installation step:

@example
sh# @kbd{cd gcc}
sh# @kbd{touch f/g77.info f/BUGS f/INSTALL f/NEWS}
sh# @kbd{cd ..}
sh#
@end example

@node Distributing Binaries
@section Distributing Binaries
@cindex binaries, distributing
@cindex code, distributing

If you are building @code{g77} for distribution to others in binary form,
first make sure you are aware of your legal responsibilities (read
the file @file{gcc/COPYING} thoroughly).

Then, consider your target audience and decide where @code{g77} should
be installed.

For systems like GNU/Linux that have no native Fortran compiler (or
where @code{g77} could be considered the native compiler for Fortran and
@code{gcc} for C, etc.), you should definitely configure
@code{g77} for installation
in @file{/usr/bin} instead of @file{/usr/local/bin}.
Specify the
@samp{--prefix=/usr} option when running @file{./configure}.

You might also want to set up the distribution
so the @file{f77} command is a link to @file{g77},
although a script that accepts ``classic'' UNIX @code{f77}
options and translates the command-line to the
appropriate @code{g77} command line would be more appropriate.
If you do this, @emph{please} also provide a ``man page'' in
@file{man/man1/f77.1} describing the command.
(A link to @file{man/man1/g77.1} is appropriate
if @file{bin/f77} is a link to @file{bin/g77}.)

For a system that might already have @code{f2c} installed,
consider whether inter-operation with @code{g77} will be
important to users of @code{f2c} on that system.
If you want to improve the likelihood
that users will be able to use both @code{f2c} and @code{g77}
to compile code for a single program
without encountering link-time or run-time incompatibilities,
make sure that,
whenever they intend to combine @code{f2c}-produced code
with @code{g77}-produced code in an executable, they:

@itemize @bullet
@item
Use the @file{lib/gcc-lib/@dots{}/include/g2c.h} file
generated by the @code{g77} build
in place of the @file{f2c.h} file
that normally comes with @code{f2c}
(or versions of @code{g77} prior to 0.5.23)
when compiling @emph{all} of the @code{f2c}-produced C code

@item
Link to the @code{lib/gcc-lib/@dots{}/libg2c.a} library
built by the @code{g77} build
instead of the @file{libf2c.a} library
that normally comes with @code{f2c}
(or versions of @code{g77} prior to 0.5.23)
@end itemize

How you choose to effect the above depends on whether
the existing installation of @code{f2c} must be
maintained.

In any case, it is important to try and ensure that
the installation keeps working properly even after
subsequent re-installation of @code{f2c},
which probably involves overwriting
@file{/usr/local/lib/libf2c.a} and
@file{/usr/local/include/f2c.h},
or similar.

At least, copying @file{libg2c.a} and @file{g2c.h} into
the appropriate ``public'' directories
allows users to more easily select the version of
@code{libf2c} they wish to use for a particular
build.
The names are changed by @code{g77} to make this
coexistence easier to maintain;
even if @code{f2c} is installed later,
the @code{g77} files normally installed
by its installation process aren't disturbed.
Use of symbolic links from one set of files to
another might result in problems after a subsequent
reinstallation of either @code{f2c} or @code{g77},
so be sure to alert users of your distribution
accordingly.

(Make sure you clearly document, in the description of
your distribution, how installation of your distribution will
affect existing installations of @code{gcc}, @code{f2c},
@code{f77}, @file{libf2c.a}, and so on.
Similarly, you should clearly document any requirements
you assume will be met by users of your distribution.)

For other systems with native @code{f77} (and @code{cc}) compilers,
configure @code{g77} as you (or most of your audience) would
configure @code{gcc} for their installations.
Typically this is for installation in @file{/usr/local},
and would not include a new version of @file{/usr/bin/f77}
or @file{/usr/local/bin/f77},
so users could still use the native @code{f77}.

In any case, for @code{g77} to work properly, you @strong{must} ensure
that the binaries you distribute include:

@table @file
@item bin/g77
This is the command most users use to compile Fortran.

@item bin/gcc
This is the command some users use to compile Fortran,
typically when compiling programs written in other languages
at the same time.
The @file{bin/gcc} executable file must have been built
from a @code{gcc} source tree into which a @code{g77} source
tree was merged and configured, or it will not know how
to compile Fortran programs.

@item info/g77.info*
This is the documentation for @code{g77}.
If it is not included, users will have trouble understanding
diagnostics messages and other such things, and will send
you a lot of email asking questions.

Please edit this documentation (by editing @file{gcc/f/*.texi}
and doing @samp{make doc} from the @file{/usr/src/gcc} directory)
to reflect any changes you've made to @code{g77}, or at
least to encourage users of your binary distribution to
report bugs to you first.

Also, whether you distribute binaries or install @code{g77}
on your own system, it might be helpful for everyone to
add a line listing this manual by name and topic to the
top-level @code{info} node in @file{/usr/info/dir}.
That way, users can find @code{g77} documentation more
easily.
@xref{Updating Documentation,,Updating Your Info Directory}.

@item man/man1/g77.1
This is the short man page for @code{g77}.
It is not always kept up-to-date,
but you might as well include it
for people who really like ``man'' pages.

@cindex gcc-lib directory
@cindex directories, gcc-lib
@item lib/gcc-lib
This is the directory containing the ``private'' files
installed by and for @code{gcc}, @code{g77}, @code{g++},
and other GNU compilers.

@item lib/gcc-lib/@dots{}/f771
This is the actual Fortran compiler.

@item lib/gcc-lib/@dots{}/libg2c.a
This is the run-time library for @code{g77}-compiled programs.
@end table

Whether you want to include the slightly updated (and possibly
improved) versions of @file{cc1}, @file{cc1plus}, and whatever other
binaries get rebuilt with the changes the GNU Fortran distribution
makes to the GNU back end, is up to you.
These changes are highly unlikely to break any compilers,
because they involve doing things like adding to the
list of acceptable compiler options
(so, for example, @file{cc1plus} accepts, and ignores,
options that only @file{f771} actually processes).

Please assure users that unless
they have a specific need for their existing,
older versions of @file{gcc} command,
they are unlikely to experience any problems by overwriting
it with your version---though they could certainly protect
themselves by making backup copies first!

Otherwise, users might try and install your binaries
in a ``safe'' place, find they cannot compile Fortran
programs with your distribution (because, perhaps, they're
invoking their old version of the @file{gcc} command,
which does not recognize Fortran programs), and assume
that your binaries (or, more generally, GNU Fortran
distributions in general) are broken, at least for their
system.

Finally, @strong{please} ask for bug reports to go to you first, at least
until you're sure your distribution is widely used and has been
well tested.
This especially goes for those of you making any
changes to the @code{g77} sources to port @code{g77}, e.g. to OS/2.
@email{fortran@@gnu.org} has received a fair number of bug
reports that turned out to be problems with other peoples' ports
and distributions, about which nothing could be done for the
user.
Once you are quite certain a bug report does not involve
your efforts, you can forward it to us.