About half edited; progress marked by "@c ....", comments/questions in

[deliverable/binutils-gdb.git] / configure.texi

\input texinfo    @c -*-para-*-
@c %**start of header
@setfilename configure.info
@settitle Cygnus Configure
@c %**end of header
@tex
\def\$#1${{#1}}  % Kluge: collect RCS revision info without $...$
\xdef\manvers{\$Revision$}  % For use in headers, footers too
@end tex
@setchapternewpage off

@ifinfo
This document attempts to describe the Cygnus Support version of
@code{configure}.

Copyright (C) 1991 Cygnus Support
Permission is granted to make and distribute verbatim copies of
this manual provided the copyright notice and this permission notice
are preserved on all copies.

@ignore
Permission is granted to process this file through TeX and print the
results, provided the printed document carries copying permission
notice identical to this one except for the removal of this paragraph
(this paragraph not being relevant to the printed manual).

@end ignore
Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided that the entire
resulting derived work is distributed under the terms of a permission
notice identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions,
except that this permission notice may be stated in a translation approved
by Cygnus Support.
@end ifinfo

@titlepage
@sp 10
@title{Cygnus Configure}
@subtitle @manvers, for Cygnus Configure version 1.84
@author{K. Richard Pixley, @code{rich@@cygnus.com}}
@author{Cygnus Support}
@page

@vskip 0pt plus 1filll
Copyright @copyright{} 1991 Cygnus Support

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

Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided that the entire
resulting derived work is distributed under the terms of a permission
notice identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions,
except that this permission notice may be stated in a translation approved
by Cygnus Support.
@end titlepage

@ifinfo
@format
START-INFO-DIR-ENTRY
* configure: (configure.info).  Cygnus configure.
END-INFO-DIR-ENTRY
@end format

@node top, What Configure Does, (dir), (dir)
@top top

This file documents the configuration system used and distributed by
Cygnus Support.

NOTE: support for a Cygnus experimental option, @code{-subdirs} is at
least temporarily suspended.  Most of the code is still in configure but
the option is disabled.  This document describes that feature, but those
parts are prominently marked with NOTE's like this one. FIXME-soon

@menu
* What Configure Does::         What Configure Does
* Invoking::                    Invoking
* Using Configure::             Using Configure
* Porting::                     Porting with Configure
* Reference::                   Gory details described
* Known Bugs::                  Known Bugs
* Variables Index::             Variable Index
* Concept Index::               Concept Index

@end menu

@end ifinfo

@iftex
@unnumbered Preface
NOTE: support for a Cygnus experimental option, @code{-subdirs} is at
least temporarily suspended.  Most of the code is still in configure but
the option is disabled.  This document describes that feature, but those
parts are prominently marked with NOTE's like this one. FIXME-soon
@end iftex

@node What Configure Does, Invoking, top, top
@chapter What Configure Does

@code{configure} prepares source directories for building working
programs.  A program cannot be built until its source has been
configured.  When configure runs, it does the following things for each source
directory for each host and target combination.
@*
NOTE: support for multiple hosts and targets is at least temporarily
suspended.

@table @emph
@item Create build directories
(see @ref{Build Directories}).  When you run @code{configure} with the
@code{-srcdir=} option, it uses the current directory as build
directory, creating under it a directory tree that parallels the
directory structure under the source directory.  (See @ref{Invoking}).

NOTE: support for @code{-subdirs} is at least temporarily suspended.
@*
When you run @code{configure} with the @code{-subdirs} option, it
creates a build subdirectory in each source directory.

If you use both @code{-subdirs} and @code{-srcdir=}, a tree that
parallels the source directory structure is created in the current
directory, and the subdirectories are created in this directory tree
rather than in the source directories.

@item Generate makefiles
A makefile template from the source directory, usually called
@file{Makefile.in}, is copied to an output file in the build directory.
The output file is usually named @file{Makefile}.  @code{configure}
places definitions for a number of standard makefile
macros at the beginning of the output file.  If @code{-prefix=} or
@code{-datadir=} were specified on the @code{configure} command line,
corresponding makefile variables are set accordingly.  If host, target, or
site specific makefile fragments exist, these are inserted into the
output file.  (See @ref{Makefiles, , , make, Makefiles}.)

@item Generate @file{.gdbinit}
If the source directory contains a @file{.gdbinit} file and the build
directory is not the same as the source directory, a @file{.gdbinit}
file is created in the build directory.  (see @ref{Command Files, , ,
gdb, Command Files}.)
@c There doesn't seem to be anything else about this.  Is the build-dir
@c .gdbinit identical with the source-dir one?  If so should say "copy"
@c rather than "create" to make it clear.

@item Make symbolic links
Most directories have some symbolic links with generic names built
pointing to specific files in the source directory.  If the system where
@code{configure} runs cannot support symbolic links, hard links are used
instead.

@item Miscellaneous
If the source directory has special needs, they are handled by shell
script fragments stored with the source.  Usually there are no special
needs, but sometimes they involve changes to the output makefile.

@item Generate @file{config.status}
@code{configure} creates a shell script named @file{config.status} in
the build directory.  This shell script, when run from the build
directory, will reconfigure the build directory (but not its
subdirectories).  This is most often used to have a @code{Makefile} update
itself automatically if a new source directory is available.
(see @ref{Top, , , bash}.)
@c That's a rather extraordinary xref.  What's it meant to clarify
@c ---shell scripts in general??

@item Recursion
If the source directory has subdirectories that should also be
configured, @code{configure} is called for each.
@end table

@node Invoking, Using Configure, What Configure Does, top
@chapter Invoking

The usual way to invoke @code{configure} is as follows:
@example
configure @var{host}
@end example
This prepares the source to be compiled in a
@var{host} environment with programs and files to be installed in
@file{/usr/local}.

NOTE: support for multiple hosts is at least temporarily suspended.
FIXME-soon
@*
If more than one host is specified on the command line, then
configurations are created for each and @code{-subdirs} is assumed.

@code{configure} prepares the source as you specify by selecting and
using script and Makefile fragments prepared in advance, and stored with
the source.  @code{configure}'s command line options also allow you to
specify other aspects of the source configuration:

@table @code
@item -datadir=@var{dir}
Configure the source to install host independent files in @var{dir}.

This option sets the @code{configure} variable @code{datadir}.  If
@code{datadir} is not empty, generated Makefiles will have their
@code{datadir} variables set to this value. (See @ref{Install Details}.)

@item -gas
Configure to use the @sc{GNU} assembler.

@item -help
Display a quick summary of how to invoke @code{configure}.

@item -host=@var{host}
FIXME-soon: I don't think this option should be documented.
@c Then why does it exist?  /Pesch 7jan92

@item -namesubdir=@var{name}
NOTE: support for this @code{-namesubdir=} is at least temporarily
suspended. FIXME-soon

Name any subdirectories created by the @code{-subdirs} option
@file{@var{name}}.   

@emph{Warning:} Avoid using this option if you specify multiple hosts
simultaneously.  There is no way to specify separate names for
subdirectories, when you configure for multiple hosts in a single
invocation of @code{configure}.

@item -nfp
@c singular "target" due to apparent direction of configure.
@emph{No floating point} unit available on the target; configure to
avoid dependencies on hardware floating point.
@c Can we even say "configure to use software floating point support"?

@item -norecursion
Configure only this directory; ignore any subdirectories.  This is used
by the executable shell script @file{config.status} to reconfigure the
current directory.  (see @ref{config.status}).
@c Why *does* that use no recursion?  Speed? geometric combinations
@c under some other script?

@ignore
@c This is complicated enough without "no longer supported" entries.
@c Should really delete this, but for ease of discourse...
@item -objdir=@var{dir}
This option is no longer supported.  Use @code{-srcdir=} instead.
@end ignore

@item -prefix=@var{dir}
Configure the source to install programs and files under directory
@file{@var{dir}}.

This option sets the @code{configure} variable @code{prefix}.  If
@code{prefix} is not empty, generated Makefiles will have their
@code{prefix} variables set to this value. (See @ref{Install Details}.)

@item -recurring
@c Wouldn't it make more sense to call this "-quiet"?
This option is used internally by @code{configure} when recurring on
subdirectories.  Its sole purpose is to suppress status output.  You can
override this effect with the @code{-verbose} option.

@item -rm
@emph{Remove} the configuration specified by @var{host} and the other
command-line options, rather than creating it.

@item -site=@var{site}
Generate Makefiles using site specific Makefile fragments for
@var{site}.  See also @ref{Sites}.

@item -srcdir=@var{_dir}
Build Makefiles to use the sources located in directory @file{@var{dir}}.  The
build directory is assumed to be @file{.}.

@item -subdirs
NOTE: support for this @code{-subdirs} is at least temporary suspended.
FIXME-soon

Place configurations in subdirectories of each build directory.
@code{configure} builds a separate subdirectory for each host specified,
and names it @file{H-@var{host}}.  If a configuration is not native,
(@var{host} is not @var{target}), then the subdirectory is named
@file{X-@var{host}-@var{target}} instead.  You can also name a
subdirectory explicitly using the @samp{-namesubdir} option, but this is
effective only when you specify one configuration at a time.

@item -target=@var{target}
Requests that the sources be configured to target the @var{target}
machine.  If no target is specified explicitly, the target is assumed
to be the same as the host.  

NOTE: support for multiple targets is at least temporarily suspended.
FIXME-soon
@* 
If multiple targets are specified, configurations for each
are created and @code{-subdirs} is assumed.

@item -tmpdir=@var{tmpdir}
Use the directory @var{tmpdir} for @code{configure}'s temporary files.
@c default?

@item -verbose
@itemx -v
Print status lines for each directory configured.  Normally, only the
status lines for the initial working directory are printed.

@item -x
Use @sc{MIT} style @sc{X11} header files and libraries on the host, even
if they are not normally available.
@end table

@node Using Configure, Porting, Invoking, top
@chapter Using Configure

The choices and options available at configuration time
generally have valid defaults, but the defaults do not cover all cases.
The choices available include install locations, build directories,
host, target, and local conventions.

@menu
* Install Locations::           Where to install things once they are built
* Build Directories::           Where to build object files
* Host::                        Telling @code{configure} what will source will
                          be built
* Target::                      Telling @code{configure} what the source will
                          target
* Local Conventions::           Adding information about local conventions
@end menu

@node Install Locations, Build Directories, Using Configure, Using Configure
@section Install Locations
@cindex Where to install

Using the default configuration, @code{make install} creates a
single tree of files, some of which are programs.  The location of this
tree is determined by the value of the variable @code{prefix}.  The
default value of @code{prefix} is @file{/usr/local}.  This is
often correct for native tools installed on only one host.

@menu
* prefix::              Changing the default install directory
* datadir::             How to separate host independent files
                          from host dependent files when
                          installing for multiple hosts
* Install Details::     Full descriptions of all installation
                          subdirectories
@end menu

@node prefix, datadir, Install Locations, Install Locations
@subsection Changing the default install directory
@cindex Changing the default install directory
@cindex Prefix directory

In the default configuration, all files are installed in subdirectories
of @file{/usr/local}.  The location is determined by the value of
the @code{configure} variable @code{prefix}; in turn, this determines the
value of the Makefile variable of the same name (@code{prefix}).

You can also set the value of the Makefile variable @code{prefix}
explicitly each time you invoke @code{make} if you are so inclined; but
because many programs have this location compiled in, you must specify
the @code{prefix} value consistently on each invocation of @code{make},
or you will end up with a broken installation.

To make this easier, the value of the @code{configure} variable
@code{prefix} can be set on the command line to @code{configure}
using the option @code{-prefix=}.  
@c This is self-referential.  What was intended?: (See @ref{prefix}).


@node datadir, Install Details, prefix, Install Locations
@subsection Installing for multiple hosts
@cindex Configuring for multiple hosts
@cindex Sharing host independent files
@cindex The datadir directory
@cindex Installing host independent files

By default, host independent files are installed in subdirectories of
@file{@var{prefix}/lib}.  The location is determined by the value of the
@code{configure} variable @code{datadir}, which determines the value of
the Makefile variable @code{datadir}.  This makes it simpler to install
for a single host, and simplifies changing the default location for the
install tree; but the default doesn't allow for multiple hosts to
effectively share host independent files.

To configure so that multiple hosts can share common files, use
something like:

@example
configure @var{host1} -prefix=/usr/gnu/H-@var{host1} -datadir=/usr/gnu/H-independent
make all info install install-info clean

configure @var{host2} -prefix=/usr/gnu/H-@var{host2} -datadir=/usr/gnu/H-independent
make all info install install-info
@end example

The first line configures the source for @var{host1} to place host
specific programs in subdirectories of @file{/usr/gnu/H-@var{host1}},
and host independent files in @file{/usr/gnu/H-independent}.  
@c Self-ref? (See @ref{datadir}.)

The second line builds and installs all programs for @var{host1},
including both host independent and host specific files.

The third line reconfigures the source for @var{host2} to place host
specific programs in subdirectories of @file{/usr/gnu/H-@var{host2}},
and host independent files (once again) in
@file{/usr/gnu/H-independent}.

The fourth line builds and installs all programs for @var{host2}.  Host
specific files are installed in new directories, but the host
independent files are installed @emph{on top of} the host
independent files installed for @var{host1}.  This results in a single
copy of the host independent files, suitable for use by both hosts.
@c Won't make notice the installed copies aren't out of date and leave
@c 'em alone?

NOTE: support for @code{-subdirs} and multiple hosts is at least
temporarily suspended. FIXME-soon
@*
Previously this was:

@example
configure @var{host1} @var{host2} -prefix=/usr/gnu
@c and make something-or-other, surely?
@end example

@node Install Details,  , datadir, Install Locations
@subsection Full descriptions of all installation subdirectories

During any install, a number of standard directories are created.  Their
names are determined by Makefile variables.  Some of the
defaults for Makefile variables can be changed at configure time using
command line options to @code{configure}.  For more information on the
standard directories or the Makefile variables, please refer to
@cite{standards.text}.

Note that @code{configure} does not create the directory @code{srcdir}
at any time.  @code{srcdir} is not an installation directory.

You can override all makefile variables on the command line to
@code{make}.  (See @ref{Overriding, Overriding Variables, Overriding
Variables, make, Make}.)  If you do so, you will need to specify the
value precisely the same way for each invocation of @code{make}, or you
risk ending up with a broken installation.  This is because many
programs have the locations of other programs or files compiled into
them.  If you find yourself overriding any of the variables frequently,
you should consider site dependent Makefile fragments.  See also
@ref{Sites}.

During @code{make install}, a number of standard directories are
created and populated.  The following Makefile variables define them.
Those whose defaults are set by corresponding @code{configure} variables
are marked ``Makefile and configure''.

@vindex prefix
@defvr {Makefile and configure} prefix
The root of the installation tree.  You can set
its Makefile default with the @code{-prefix=} command line option to
@code{configure}.  (@ref{Invoking}.)  The default value for
@code{prefix} is @file{/usr/local}.
@end defvr

@vindex bindir
@defvr Makefile bindir
A directory for binary programs that users can run.
The default value for @code{bindir} depends on @code{prefix};
@code{bindir} is normally changed only indirectly through @code{prefix}.
The default value for @code{bindir} is @file{$(prefix)/bin}.
@end defvr

@vindex datadir
@defvr {Makefile and configure} datadir
A directory for host independent files.  You can specify the Makefile
default value by using the @code{-datadir=} option to @code{configure}.
(See also @ref{Invoking}.)  The default value for @code{datadir} is
@file{$(prefix)/lib}.
@end defvr

@vindex libdir
@defvr Makefile libdir
A directory for libraries and support programs.  The default value for
@code{libdir} depends on @code{prefix}; @code{libdir} is normally
changed only indirectly through @code{prefix}.  The default value for
@code{libdir} is @file{$(prefix)/lib}.
@end defvr

@vindex mandir
@defvr Makefile mandir
A directory for @code{man} format documentation (``man pages'').  The
default value for @code{mandir} depends on @code{prefix};
@code{mandir} is normally changed only indirectly through @code{prefix}.
The default value for @code{mandir} is @file{$(datadir)/man}.
@end defvr

@vindex man@var{N}dir
@defvr Makefile man@var{N}dir
There are eight variables named @code{man1dir}, @code{man2dir}, etc.
They name the specific directories for each man page section.  For
example, @code{man1dir} holds @file{emacs.1} (the man page for the emacs
program), while @code{man5dir} holds @file{rcsfile.5} (the man page
describing the @code{rcs} data file format).  The default value for any
of the @code{man@var{N}dir} variables depends indirectly on
@code{prefix}, and is normally changed only through @code{prefix}.  The
default value for @code{man@var{N}dir} is
@file{$(mandir)/man@var{N}}.
@end defvr

@vindex manext
@defvr Makefile manext
@emph{Not supported by @code{configure}}.  The @sc{gnu} coding standards
do not call for @code{man1ext}, @code{man2ext}, so the intended use for
@code{manext} is apparently not parallel to @code{mandir}.  Its use is
not clear.  (See also @ref{Makefile Extensions}.)
@end defvr

@vindex infodir
@defvr Makefile infodir
A directory for @emph{info} format documentation.  The default value for
@code{infodir} depends indirectly on @code{prefix}; @code{infodir} is
normally changed only through @code{prefix}.  The default value for
@code{infodir} is @file{$(datadir)/info}.
@end defvr

@vindex docdir
@defvr Makefile docdir
A directory for any documentation that is in a format other than those
used by @code{info} or @code{man}.  The default value for @code{docdir}
depends indirectly on @code{prefix}; @code{docdir} is normally changed only
through @code{prefix}.  The default value for @code{docdir}
is @file{$(datadir)/doc}.  @emph{This variable is an extension to
the @sc{gnu} coding standards}.  (See also @ref{Makefile Extensions}.)
@end defvr

@vindex includedir
@defvr Makefile includedir
A directory for the header files accompanying the libraries installed in
@code{libdir}.  The default value for @code{includedir} depends on
@code{prefix}; @code{includedir} is normally changed only indirectly
through @code{prefix}.  The default value for @code{includedir} is
@file{$(prefix)/include}.
@end defvr

@node Build Directories, Host, Install Locations, Using Configure
@section Build Directories
@cindex Build directories
@cindex objdir
@cindex Object directories
@cindex subdirs
@cindex Building for multiple hosts
@cindex Building for multiple targets

Normally, @code{configure} builds a @file{Makefile} and symbolic links
in the same directory as the source files.  This is the typical
@sc{un*x} way to build programs, but it has limitations.  For instance,
using this approach, you can only build for one host at a time.

@c "Makefile" treated as ordinary word through most of this; I've left it
@c that way since that seems to agree w ordinary usage.  This one was
@c @code'd; if the intent is to emphasize that we're now talking of it
@c as a file, I suggest
@c "...builds @file{Makefile} files"
We refer to the directories where @code{configure} builds a
Makefile as the @emph{build directories} or sometimes as
@emph{objdir} because these are the directories in which @code{make}
will build object files, among other things.

The default build directory is the same as the source directory.
You can use a different build directory with a sequence like the following:

@example
mkdir @var{builddir}
cd @var{builddir}
configure @var{host} -srcdir=@var{sourcedirectory}
@end example

@noindent
where @var{builddir} is the directory where you wish to build,
@var{host} is the host for which you want to build, and
@var{sourcedirectory} is the directory containing the source files.

If you were to do this twice with different values for @var{builddir}
and @var{host}, then you could @code{make} for both at the same time.

@quotation
@emph{NOTE:} The rest of this section describes the @code{-subdirs} feature for
which support is at least temporarily suspended.  FIXME-soon.
@end quotation

Another way to specify the build directory is with the @samp{-subdirs}
option.  For example:

@example
configure @var{host} -subdirs
@end example

Using this option, @code{configure} will create a subdirectory named
@file{H-@var{host}} to act as the build directory for each source
directory.

Since building for multiple hosts is so common, @code{configure}
recognizes this situation as special.  For example:

@example
configure @var{host1} @var{host2}
@end example

is precisely the same as:

@example
configure @var{host1} -subdirs
configure @var{host2} -subdirs
@end example

That is, configuring for multiple hosts or multiple targets implies
@samp{-subdirs}.

When configuring for cross tools (the converse of native tools: when the
host is not the target), as in:

@example
configure @var{host} +target=@var{targ} -subdirs
@end example

@noindent
the subdirectories are named @file{X-@var{host}-@var{targ}}.  This is
especially useful when configuring for multiple targets.

If you use both @samp{-subdirs} and @samp{-srcdir=}, a tree that
parallels the source directory structure is created in the current
directory, and the subdirectories are created in this directory
tree rather than in the source directories.

@emph{NOTE:} previously, @samp{-subdirs} built two-level subdirectories
as @file{./H-@var{host}/T-@var{target}}, created
@file{./H-@var{host}/Makefile} for building across all targets,
@file{./Makefile} for building across all hosts, and
@file{./config.status} and @file{./H-@var{host}/config.status} for
rebuilding these Makefiles.

@node Host, Target, Build Directories, Using Configure
@section Host

@quotation
@emph{NOTE:} support for multiple hosts is at least temporarily suspended.
FIXME-soon.
@end quotation

The arguments to @code{configure} are @emph{hosts}.  By @emph{host} we
mean the environment in which the source will be compiled.  This need
not necessarily be the same as the  physical machine involved,
although it usually is.

For example, if some obscure machine running an operating system other
than @sc{un*x}  had the @sc{gnu} @sc{posix} emulation libraries
available, it would be possible to configure most @sc{gnu} source for a
@sc{posix} system and build it on the obscure host.

For more on this topic, see @ref{Host Environments, , cfg-paper, On
Configuring Development Tools}.

@node Target, Local Conventions, Host, Using Configure
@section Target

For building native development tools, or most of the other @sc{gnu}
tools, you need not worry about the target.  The @emph{target} of a
configuration defaults to the same as the @emph{host}.

For building cross development tools, please see @ref{Building
Development Environments, , cfg-paper, On Configuring Development
Tools}.

@node Local Conventions,  , Target, Using Configure
@section Local Conventions

If you find that a tool does not get configured to your liking, or if
@code{configure}'s conventions differ from your local conventions, you
should probably consider site specific Makefile fragments.  See also
@ref{Sites}.

These are probably not the right choice for options that can be set from
the @code{configure} command line or for differences that are host or
target dependent.

@node Porting, Reference, Using Configure, top
@chapter Porting with Configure
@cindex Porting

This section explains how to add programs, host and target configuration
names, and site-specific information to Cygnus configure.

@menu
* Programs::                    Adding configure to new programs
* Hosts and Targets::           Adding hosts and targets
* Sites::                       Adding site info
@end menu


@node Programs, Hosts and Targets, Porting, Porting
@section Adding Configure To New Programs

If you are writing a new program, you probably shouldn't worry about
porting issues or configure until it is running reasonably on some host.
Then refer back to this section.

If the program in question currently has a configure script that meets
the criteria set out by @cite{standards.text}, please do not add Cygnus
configure.  It should be possible to add this program without change to
a Cygnus configure style source tree.

If the program is not target dependent, please consider using
@code{autoconf} instead of Cygnus configure.  @code{autoconf} will
be available soon from the @sc{fsf}.

@c ..............................pesch rev..............................

To add Cygnus configure to an existing program, do the following.

@table @asis

@item Bring the Makefile up to the standard
The coding standard for @sc{gnu} Makefiles is described in
@cite{standards.text}.

@item Add Cygnus extensions to the Makefile
There are described in @ref{Makefile Extensions}.

@item Move host support from Makefile to fragments
This usually involves finding sections of the Makefile that say things
like ``uncomment these lines for host foo'' and moving them to a new
file call @file{./config/mh-foo}. For more on this, see @ref{Hosts and
Targets}.

@item Choose defaults
If the program has compile time options that determine the way the
program should behave, chose reasonable defaults and make these Makefile
variables.  Be sure the variables are assigned their default values
before the @code{####} line so that they can be overridden with site
specific Makefile fragments.

@item Locate configuration files
If there is configuration information in header files or source files,
separate it in such a way that the files have a generic name.  Then move
the specific instances of those files into the @file{./config}
directory.

@item Separate host and target information
Some programs already have this information separated.  If not, you will
need to do so.  Host specific information is the information needed to
compile the program.  Target specific information it information on the
format of data files that the program will read or write.  This
information should live in separate files in the @file{./config}
directory with names that reflect the configuration for which they are
intended.

At this point you might skip this step and simply move on.  If you do,
you should end up with a program that can be configured only to build
native tools, that is, tools for which the host system is also the
target system.  Later, you could attempt to build a cross tool and
separate out the target specific information by figuring out what went
wrong.  This is often simpler than combing through all of the source
code.

@item Write configure.in
Usually this involves writing shell script fragments to map from
canonical configuration names into the names of the configuration files.
These files will then be linked at configure time from the specific
instances of those files in @file{./config} to file in the build
directory with more generic names.  (see also @ref{Build Directories}).
The format of configure.in is described in @ref{configure.in}.

@item Rename the Makefile to Makefile.in

@end table

At this point you should have a program that can be configured by Cygnus
configure.

@node Hosts and Targets, Sites, Programs, Porting
@section Adding hosts and targets

To add a host or target to a program that currently uses Cygnus
configure, do the following.

@itemize @bullet

@item
Make sure the new configuration name is represented in
@file{config.sub}.  If not, add it.  For more details, see the comments
in the shell script @file{config.sub}.

@item
If you are adding a host configuration, look in @file{configure.in}, in
the per-host section.  Make sure that your configuration name is
represented in the mapping from host configuration names to
configuration files.  If not, add it.  Also see @ref{configure.in}.

@item
If you are adding a target configuration, look in @file{configure.in},
in the per-target section.  Make sure that your configuration name is
represented in the mapping from target configuration names to
configuration files.  If not, add it.  Also see @ref{configure.in}.

@item
Look in @file{configure.in} for the assignments to the variables
@code{files}, @code{links}, @code{host_makefile_frag}, and
@code{target_makefile_frag}.  These are the names of the configuration
files that the program uses.  Make sure that copies of the files exist
for your host.  If not, create them.  See also @ref{Configure
Variables}.

@end itemize

This should be enough to configure for a new host or target
configuration name.  Getting the program to compile and run properly now
is the hard work of the port.


@node Sites,  , Hosts and Targets, Porting
@section Adding site info

If some of the Makefile defaults are not right for your site, you can
build site specific Makefile fragments.  To do this, do the following.

@itemize @bullet

@item
Choose a name for your site.  It must be less than eleven characters for
now.

@item
If the program does not have a @file{./config} directory, create it.

@item
Create a file called @file{./config/ms-@var{site}} where @var{site} is
the name of your site.  In it, set the Makefile variables of your
choice.

@item
Configure the program with:

@example
configure @dots{} +site=@var{site}
@end example

@end itemize
@node Reference, Known Bugs, Porting, top
@chapter Gory details described

@cindex Backends
Here we describe the backend support.

@menu
* Makefile Extensions::         Extensions to the @sc{gnu} coding standards
* configure.in::                The format of the configure.in file
* config.status::               config.status
* Makefile Fragments::          Makefile Fragments
@end menu

@node Makefile Extensions, configure.in, Reference, Reference
@section Extensions to the @sc{gnu} coding standards

@cindex Makefile extensions
@cindex Cygnus extensions

The following additions to the @sc{gnu} coding standards are required
for Cygnus configure to work properly.

@itemize @bullet

@item
The Makefile must contain exactly one line starting with @code{####}.
This line should follow any default macro definitions but precede any
rules.  Host, target, and site specific Makefile fragments will be
inserted immediately after this line.  If the line is missing, the
fragments will not be inserted.

@end itemize

Cygnus adds the following targets to our Makefiles.  Their existence is
not required for Cygnus configure but are documented here for
completeness.

@table @code

@cindex info
@item info
Build all info files from texinfo source.

@cindex install-info
@item install-info
Install all info files.

@cindex clean-info
@item clean-info
Remove all info files and any intermediate files that can be generated
from texinfo source.

@cindex stage1
@item stage1
@cindex stage2
@item stage2
@cindex stage3
@item stage3
@cindex stage4
@item stage4
@cindex de-stage1
@item de-stage1
@cindex de-stage2
@item de-stage2
@cindex de-stage3
@item de-stage3
@cindex de-stage4
@item de-stage4
@cindex bootstrap
@item bootstrap
@cindex comparison
@item comparison
@cindex Makefile
@item Makefile
These targets are in transition and may be removed shortly.

@end table

In addition, the following Makefile targets have revised semantics:

@table @code

@cindex install
@item install
Should @emph{not} depend on the target @code{all}.  If the program is
not already built, @code{make install} should fail.  This allows
programs to be installed even when @code{make} would otherwise determine
them to be out of date.  This can happen when the result of a @code{make
all} is transported via tape to another machine for installation as
well as in a number of other cases.

@cindex clean
@item clean
Should remove any file that can be regenerated by the Makefile,
excepting only the Makefile itself, and any links created by configure.
That is, @code{make all clean} should return all directories to their
original condition.  If this is not done, then:

@example
configure @var{host1} ; make all clean ; configure @var{host2} ; make all
@end example

@noindent
will fail because of intermediate files intended for @var{host1}.

@end table

Cygnus adds the following macros to all Makefile.in's.  Their presence
is not required for Cygnus configure.

@table @code

@cindex docdir
@item docdir
The directory in which to install any documentation that is not either a
man page or an info file.  For man pages, see mandir, for info, see
infodir.

@cindex includedir
@item includedir
The directory in which to install any headers files that should be made
available to users.  This is distinct from the @code{gcc} include
directory which is intended for @code{gcc} only.  Files in
@code{includedir} may be used by @code{cc} as well.

@end table

In addition, the following macros have revised semantics.

@table @code

@cindex manext
@item manext
is not used.  The intended usage is not clear.  For example, if I have a
@file{foo.man} and a @file{bar.man}, and @file{foo.man} is destined for
@file{/usr/local/lib/man/man1/foo.1} while @file{bar.man} is destined
for @file{/usr/local/lib/man/man5/bar.5}, then to what should the value
of @code{manext} be set?  See also @ref{Install Details}.

@cindex datadir
@item datadir
is used for @emph{all} host independent files.  This makes it possible
to share host independent files across multiple hosts without ersorting
to symlinks or multiple mount points.  This also makes it possible
build an install tree that contains multiple host binaries, write
the binaries to tape, and extract any of the hosts without extracting
the others.

@cindex mandir
@item mandir
man pages are host independent so the default path for @code{mandir}
depends on @code{datadir}.

@cindex infodir
@item infodir
info files are host independent so the default path for @code{infodir}
depends on @code{datadir}.

@cindex BISON
@item BISON
is assumed to have a yacc calling convention.  To  use
@code{bison}, use @code{BISON=bison -y}.

@end table

Cygnus also adds the following restrictions on our Makefiles.

@itemize @bullet

@item
When libraries are installed, the line containing the call to
@code{INSTALL_DATA} should always be followed by a line containing a
call to @code{RANLIB} on the installed library.  This is to accomodate
systems that use @code{ranlib}.  Systems that do not use ranlib can set
@code{RANLIB} to @code{echo} in a host specific Makefile fragment.

@end itemize

@node configure.in, config.status, Makefile Extensions, Reference
@section The format of the configure.in file

@cindex configure.in

A configure.in file for Cygnus configure consists of a declarations
section, followed by a per-host section, followed by a per-target
section, optionally followed by a post-target section.  Each section is
a shell script fragment sourced by configure at the appropriate time.
The interface between configure and the shell fragments is through a set
of shell variables.  All sections are sourced in the build directory.

@cindex Per-host section
A line beginning with @code{# Per-host:} begins the per-host section.

@cindex Per-target section
A line beginning with @code{# Per-target:} begins the per-target
section.

@cindex Post-target section
If it exists, the post-target section begins with @code{# Per-target:}.

@menu
* Minimal::                     A minimal configure.in
* Configure Variables::         Variables available to configure.in
* Declarations::                Per invocation
* Per-host::                    On a host basis
* Per-target::                  On a target basis
* Post-target::                 After each target
* Example::                     An example configure.in
@end menu

@node Minimal, Configure Variables, configure.in, configure.in
@subsection A minimal configure.in

@cindex Minimal configure.in example
A minimal @file{configure.in} consists of four lines.

@example
srctrigger=foo.c
srcname="source for the foo program"
# Per-host:
# Per-target:
@end example

The per-host and per-target lines divide the file into the three
required sections.  The srctrigger line names a file.  configure checks
to see that this file exists in the source directory before
configuring.  If the srctrigger file does not exist, configure
uses the value of srcname to print an error message about not finding
the source.

This particular example uses no links, and only the default host,
target, and site specific Makefile fragments if they exist.

@node Configure Variables, Declarations, Minimal, configure.in
@subsection Variables available to configure.in

@cindex Configure.in interface

The following variables are available to the shell fragments in
@file{configure.in}.

@defvar{srctrigger}
Contains the name of a source file that is expected to live in the
source directory.  This is usually set in the declations section of
@file{configure.in}.  Configure tests to see that this file exists.  If
the file does not exist, configure prints an error message.  This is
used as a sanity check that configure.in matches the source directory.
@end defvar

@defvar{srcname}
Contains the name of the source contained in the source directory.  This
is usually set in the declarations section of @file{configure.in}.  If
the file named in @code{srctrigger} does not exist, configure uses the
value of this variable when it prints the error message.
@end defvar

@defvar{configdirs}
Contains the names of any subdirectories on which configure should
recur.  This is usually set in the declarations section of
@file{configure.in}.  If @file{Makefile.in} contains a line starting
with @code{SUBDIRS =}, then it will be replaced with an assignment to
@code{SUBDIRS} using the value of @code{configdirs}.  This can be used
to determine which directories to configure and build depending on the
host and target configurations.
@end defvar

NOTE: support for multiple targets is currently suspended.

@defvar{target_dependent}
If this variable is not empty and @code{-subdirs} is in effect then
configure will create separate build directories for each target.  This
is usually set in the declarations section of @file{configure.in}.  The
default is to assume that a directory is target independent, create only
one real directory with symlinks from the other names.  This means that
a target independent directory will be built exactly once regardless of
how many targets are being built.
@end defvar

@defvar{host}
Contains the  name that the user entered for the host.  Since many
things that the user could enter would map to the same canonical triple,
this variable is innappropriate to use for picking available
configurations.  For that, use @code{host_cpu}, @code{host_vendor},
and/or @code{host_os}.  This variable is useful, however, for error
messages.
@end defvar

@defvar{host_cpu}
Contains the first element of the canonical triple representing the host
as returned by @file{config.sub}.  This is occasionally used to
distinguish between minor variations of a particular vendor's operating
system and sometimes to determine variations in binary format between
the host and the target.
@end defvar

@defvar{host_vendor}
Contains the second element of the canonical triple representing the
host as returned by @file{config.sub}.  This is usually used to
distinguish betwen the numerous variations between @emph{common}
operating systems.
@end defvar

@defvar{host_os}
Contains the the third element of the canonical triple representing the
host as returned by @file{config.sub}.
@end defvar

@defvar{target}
Contains the  name that the user entered for the target.  Since
many things that the user could enter would map to the same canonical
triple, this variable is innappropriate to use for picking available
configurations.  For that, use @code{target_cpu}, @code{target_vendor},
and/or @code{target_os}.  This variable is useful, however, for error
messages.
@end defvar

@defvar{target_cpu}
Contains the first element of the canonical triple representing the
target as returned by @file{config.sub}.  This is used heavily by
programs involved in building programs, like the compiler, assembler,
linker, etc.  Most programs will not need the @code{target} variables at
all, but this one could conceivably be used to build a program, for
instance, that operated on binary data files whose byte order or
alignment are other than that of the system on which the program is
running.
@end defvar

@defvar{target_vendor}
Contains the second element of the canonical triple representing the
target as returned by @file{config.sub}.  This is usually used to
distinguish betwen the numerous variations between @emph{common}
operating systems or object file formats.  Sometimes it is used to
switch between different flavors of user interfaces.
@end defvar

@defvar{target_os}
Contains the the third element of the canonical triple representing the
target as returned by @file{config.sub}.  This variable is used by
development tools to distinguish between subtle variations in object
file formats that some vendors use across operating system releases.  It
might also be use to decide which libraries to build or what user
interface the tool should provide.
@end defvar

@defvar{nfp}
Is set to @code{true} if the user invoked configure with the @code{-nfp}
command line option, otherwise it is empty.  This is a request to target
machines with @emph{no floating point} unit, even if the targets
ordinarily have floating point units available.  This option has no
negation.
@end defvar

@defvar{gas}
Is set to @code{true} if the user invoked configure with the @code{-gas}
command line option, otherwise it is empty.  This is a request to assume
that all target machines have gas available even if they ordinarily do
not.  The converse option, @code{-no-gas} is not available.
@end defvar

@defvar{x}
Is set to @code{true} if the user invoked configure with the @code{-x}
command line option, otherwise it is empty.  This is a request to assume
that @sc{mit x11} compatible headers files and libraries are available
on all hosts, regardless of what is normally available on them.
@end defvar

NOTE: support for @code{-subdirs} is at least temporarily suspended.

@defvar{srcdir}
Is set to the name of the directory containing the source for this
program.  This will be different from @file{.} if the user has specified
either the @code{-srcdir=} or the @code{-subdirs} options.  Note that
@code{srcdir} is not necessarily an absolute path.
@end defvar

@defvar{host_makefile_frag}
Is set to a file name representing to the default Makefile fragment for
this host.  It may be set in @file{configure.in} to overide this
default.
@end defvar

@defvar{target_makefile_frag}
Is set to a file name representing to the default Makefile fragment for
this target.  It may be set in @file{configure.in} to overide this
default.
@end defvar

@defvar{site_makefile_frag}
Is set to a file name representing to the default Makefile fragment for
this host.  It may be set in @file{configure.in} to overide this
default.  Normally @code{site_makefile_frag} is empty, but will have a
value if the user specified @code{-site=} on the command line.  This
variable should probably not be overridden.
@end defvar

@defvar{Makefile}
Is set to the name of the generated @file{Makefile}.  Normally this
value is precisely @file{Makefile} but some programs may want something
else.
@end defvar

@defvar{removing}
Is normally empty but will be set to some non-empty value if the user
specified @code{-rm} on the command line.  That is, if @code{removing}
is non-empty, then configure is @emph{removing} a configuration rather
than creating one.
@end defvar

@defvar{files}
If this variable is non-empty following the @code{per-target:} section,
then each word in it's value will be the target of a symbolic link named
in the @code{links} variable.
@end defvar

@defvar{links}
If the @code{files} variable is non-empty following the
@code{per-target:} section, then symbolic links will be created with the
first word of links pointing to the first word of files, the second word
of links pointing to the second word of files, and so on.
@end defvar


@node Declarations, Per-host, Configure Variables, configure.in
@subsection Per invocation

@cindex Declarations section

Everything from the start of @file{configure.in} up to a line beginning
with @code{# Per-host:} is sourced by configure as a shell script
fragment immediately after parsing command line arguments.  The
variables @code{srctrigger} and @code{srcname} @emph{must} be set here.

Some other things you might want to set here are the variables
@code{configdirs} or @code{target_dependent}.  FIXME-soon.
target_dependent isn't useful without multiple targets.

@node Per-host, Per-target, Declarations, configure.in
@subsection On a host basis

@cindex Per-host section
@cindex Host basis
The per-host section of @file{configure.in} starts with a line beginning
with @code{# Per-host:} and ends before a line beginning with with
@code{# Per-target:}.  Configure sources the per-host section once for
each host.

This section usually contains a big case statement using the variables
@code{host_cpu}, @code{host_vendor}, and @code{host_os} to determine
appropriate values for @code{host_makefile_frag} and @code{files},
although @code{files} is not usually set here.  Usually, it is set
at the end of the per-target section after determining the names of the
target specific configuration files.

@node Per-target, Post-target, Per-host, configure.in
@subsection On a target basis

@cindex Per-target section
@cindex Target basis

The per-target section of @file{configure.in} starts with a line
beginning with @code{# Per-target:} and ends before a line beginning
with @code{# Post-target:} if it exists.  Otherwise the per-target
section extends to the end of the file.  Configure sources the
per-target section once for each target before building any files,
directories, or links.

This section usually contains a big case statement using the variables
@code{target_cpu}, @code{target_vendor}, and @code{target_os} to determine
appropriate values for @code{target_makefile_frag} and @code{files}.
The last lines in the per-target section normally set the variables
@code{files} and @code{links}.

@node Post-target, Example, Per-target, configure.in
@subsection After each target

The post-target section is optional.  If it exists, the post-target
section starts with a line beginning with @code{# Post-target:} and
extends to the end of the file.  If it exists, configure sources this
section once for each target after building all files, directories, or
links.

This section seldom exists but can be used to munge the configure
generated Makefile.

@node Example,  , Post-target, configure.in
@subsection An example configure.in

@cindex Example configure.in
@cindex Bison configure.in
Here is a small example configure.in.

@example
# This file is a shell script fragment that supplies the information
# necessary to tailor a template configure script into the configure
# script appropriate for this directory.  For more information, check
# any existing configure script.

configdirs=
srctrigger=warshall.c
srcname="bison"

# per-host:
case "$@{host_os@}" in
m88kbcs)
        host_makefile_frag=config/mh-delta88
        ;;
esac

# per-target:

files="bison_in.hairy"
links="bison.hairy"

# post-target:
@end example

@node config.status, Makefile Fragments, configure.in, Reference
@section config.status

@cindex config.status

The final step in configuring a directory is to create an executable
shell script call @file{config.status}.  This file is typically used to
rebuild the Makefile for the current directory.  For this reason,
@file{config.status} uses the @code{-norecursion} option to configure
and is therefor probably inappropriate for  reconfiguring a tree
of source code.

@node Makefile Fragments,  , config.status, Reference
@section Makefile Fragments

@cindex Makefile fragments

Cygnus configure uses three types of Makefile fragments.  In a
generated Makefile they occur in the order target fragment, host
fragment, and site fragment.  This is so host fragments can override
target fragments etc.

Host specific Makefile fragments conventionally reside in the
@file{./config} directory with names of the form
@file{mh-@var{host}}.  They are used for hosts that require
odd options to the standard compiler and for compile time options based
on the host configuration.

Target specific Makefile fragments conventionally reside in the
@file{./config} directory with names of the form @file{mt-@var{target}}.
They are used for target dependent compile time options.

Site specific Makefile fragments conventionally reside in the
@file{./config} directory with names of the form @file{ms-@var{site}}.
They are used to override host and target independent compile time
options.  Note that these options can also be overridden on the
@code{make} invocation line.

@node Known Bugs, Variables Index, Reference, top
@chapter Known Bugs

@cindex bugs

The following bugs are known to exist.

@itemize @bullet

@item
There is no way to query about known hosts, known targets, or the
porting or testing status of any configuration.

@item
The negations to the options @code{-gas}, @code{-x}, and @code{-nfp} are
not available.

@end itemize

@page
@node Variables Index, Concept Index, Known Bugs, top
@appendix Variable Index

@printindex vr

@page
@node     Concept Index,  , Variables Index, top
@appendix Concept Index

@printindex cp
@contents
@bye

@c Local Variables:
@c fill-column: 79
@c outline-regexp: "@chap"
@c End:
@c (setq outline-regexp "@chapt\\\|@unnum\\\|@setf\\\|@conte\\\|@sectio\\\|@subsect\\\|@itemize\\\|@defvar{")