OpenBSD-4.6/usr.sbin/crunchgen/crunchgen.8

.\"	$OpenBSD: crunchgen.8,v 1.4 2008/11/24 18:03:22 jmc Exp $
.\"
.\"
.\" Copyright (c) 1994 University of Maryland
.\" All Rights Reserved.
.\"
.\" Permission to use, copy, modify, distribute, and sell this software and its
.\" documentation for any purpose is hereby granted without fee, provided that
.\" the above copyright notice appear in all copies and that both that
.\" copyright notice and this permission notice appear in supporting
.\" documentation, and that the name of U.M. not be used in advertising or
.\" publicity pertaining to distribution of the software without specific,
.\" written prior permission.  U.M. makes no representations about the
.\" suitability of this software for any purpose.  It is provided "as is"
.\" without express or implied warranty.
.\"
.\" U.M. DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL U.M.
.\" BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
.\" WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
.\" ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR
.\" IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
.\"
.\" Author: James da Silva, Systems Design and Analysis Group
.\"			   Computer Science Department
.\"			   University of Maryland at College Park
.\"
.Dd $Mdocdate: November 24 2008 $
.Dt CRUNCHGEN 8
.Os
.Sh NAME
.Nm crunchgen
.Nd generates build environment for a crunched binary
.Sh SYNOPSIS
.Nm crunchgen
.Bk -words
.Op Fl EfMq
.Op Fl c Ar c-file-name
.Op Fl D Ar src-root
.Op Fl e Ar exec-file-name
.Op Fl L Ar lib-dir
.Op Fl m Ar makefile-name
.Op Fl O Ar objdir-name
.Ar conf-file
.Nm crunchgen
.Fl h
.Op Fl f Ar keep-list-file
.Op Fl k Ar keep-symbol
.Ar object-file ...
.Ek
.Sh DESCRIPTION
A crunched binary is a program made up of many other programs linked
together into a single executable.
The crunched binary main() function determines which component program
to run by the contents of argv[0].
The main reason to crunch programs together is for fitting as many programs
as possible onto an installation or system recovery floppy.
.Pp
.Nm
reads in the specifications in
.Ar conf-file
for a crunched binary, and generates a Makefile and accompanying
top-level C source file that when built create the crunched executable
file from the component programs.
For each component program,
.Nm
can optionally attempt to determine the object (.o) files that make up
the program from its source directory Makefile.
This information is cached in a file named
.Pa <conf-name>.cache
between runs.
.Pp
.Nm
is later run again with the
.Fl h
flag to eliminate link-time conflicts between the component programs by
hiding all unnecessary symbols.
Some symbols may be left visible via the
.Fl k Ar keep-symbol
and
.Fl f Ar keep-list-file
options.
The
.Ar keep-list-file
must contain a list of symbols to keep visible, one symbol per line.
Note that the C compiler prepends an underscore in front of
symbols, so to keep the C function ``foo'' visible, the option
.Dq -k _foo
must be used.
.Pp
After
.Nm
is run, the crunched binary can be built by running
.Dq make -f <conf-name>.mk .
The component programs' object files must already be built.
An
.Dq objs
target, included in the output makefile,
will run make in each component program's source dir to build the object
files for the user.
This is not done automatically since in release engineering circumstances
it is generally not desirable to be modifying objects in other directories.
.Pp
The options are as follows:
.Bl -tag -width indent
.It Fl c Ar c-file-name
Set output C file name to
.Ar c-file-name .
The default name is
.Dq Ao conf-name Ac Ns \&.c .
.It Fl D Ar src-root
Assume that relative source directory specifications begin with
.Ar src-root .
.It Fl E
Don't prepend stub names with an underscore.
Used for architectures that don't have underscore prepended to symbol names,
such as ELF architectures.
.It Fl e Ar exec-file-name
Set crunched binary executable file name to
.Ar exec-file-name .
The default name is
.Dq Aq conf-name .
.It Fl f
Flush cache.
Forces the recalculation of cached parameters.
.It Fl h
Hide all unnecessary symbols.
Note that this is done on some ELF architectures by marking the symbol
local, while the
.Fl M
option causes it to mangle the symbol name to hide the symbol.
It is therefore not advisable to try to run
.Xr nm 1
on a crunched object file.
This is due to the nature of the ELF symbol table
and how some architectures use the symbol attributes for their GOT build.
.It Fl L Ar lib-dir
Try to obtain libraries from
.Ar lib-dir .
.It Fl M
On ELF architectures mangle the symbol instead of marking it global;
necessary for some architectures due to GOT usage.
.It Fl m Ar makefile-name
Set output Makefile name to
.Ar makefile-name .
The default name is
.Dq Ao conf-name Ac Ns \&.mk .
.It Fl O Ar objdir-name
Specify an object directory to use.
It defaults to
.Dq obj ,
though for cross building purposes it can be used to specify
obj.${HOST}.${MACHINE}.
Normally used with the make variable ${MAKEOBJDIR}.
.It Fl q
Quiet operation.
Status messages are suppressed.
.El
.Sh CRUNCHGEN CONFIGURATION FILE COMMANDS
.Nm
reads specifications from the
.Ar conf-file
that describe the components of the crunched binary.
In its simplest use, the component program names are merely listed
along with the top-level source directories in which their sources
can be found.
.Nm
then calculates (via the source makefiles) and caches the
list of object files and their locations.
For more specialized situations, the user can specify by hand
all the parameters that
.Nm
needs.
.Pp
The
.Ar conf-file
commands are as follows:
.Bl -tag -width indent
.It srcdirs Ar dirname ...
A list of source trees in which the source directories of the
component programs can be found.
These dirs are searched using the BSD
.Dq <source-dir>/<progname>/
convention.
Multiple srcdirs lines can be specified.
The directories are searched in the order they are given.
.It libdirs Ar dirname
A list of source trees in which the source directories for supplementary
libraries can be found.
.It progs Ar progname ...
A list of programs that make up the crunched binary.
Multiple progs lines can be specified.
.It libs Ar libspec ...
A list of library specifications to be included in the crunched binary link.
Multiple libs lines can be specified.
.It ln Ar progname linkname
Causes the crunched binary to invoke
.Ar progname
whenever
.Ar linkname
appears in argv[0].
This allows programs that change their behavior when
run under different names to operate correctly.
.El
.Pp
To handle specialized situations, such as when the source is not
available or not built via a conventional Makefile, the following
.Ic special
commands can be used to set
.Nm
parameters for a component program.
.Bl -tag -width indent
.It special Ar progname No srcdir Ar pathname
Set the source directory for
.Ar progname .
This is normally calculated by searching the specified srcdirs
for a directory named
.Ar progname .
.It special Ar progname No objdir Ar pathname
Set the obj directory for
.Ar progname .
This is normally calculated by looking for a directory named
.Dq obj
under the
.Ar srcdir ,
and if that is not found, the
.Ar srcdir
itself becomes the objdir.
.It special Ar progname No objs Ar object-file-name ...
Set the list of object files for program
.Ar progname .
This is normally calculated by constructing a temporary makefile that includes
.Dq srcdir/Makefile
and outputs the value of $(OBJS).
.It special Ar progname No objpaths Ar full-pathname-to-object-file ...
Sets the pathnames of the object files for program
.Ar progname .
This is normally calculated by prepending the objdir
pathname to each file in the objs list.
.El
.Pp
Only the objpaths parameter is actually needed by
.Nm crunchgen ,
but it is calculated from objdir and objs,
which are in turn calculated from srcdir,
so it is sometimes convenient to specify the earlier parameters and let
.Nm
calculate forward from there if it can.
.Pp
The makefile produced by
.Nm
contains an optional
.Ar objs
target that will build the object files for each component program by
running make inside that program's source directory.
For this to work the srcdir and objs parameters must also be valid.
If they are not valid for a particular program, that program is skipped in the
.Ar objs
target.
.Sh EXAMPLES
Here is an example
.Nm
input conf file, named
.Pa kcopy.conf :
.Bd -literal -offset indent
srcdirs /usr/src/bin /usr/src/sbin

progs test cp echo sh fsck halt init mount umount myinstall
ln test [       # test can be invoked via [
ln sh -sh       # init invokes the shell with "-sh" in argv[0]

special myprog objpaths /homes/leroy/src/myinstall.o # no sources

libs -lutil -lcrypt
.Ed
.Pp
This conf file specifies a small crunched binary consisting of some
basic system utilities plus a home-grown install program
.Dq myinstall ,
for which no source directory is specified, but its object file is
specified directly with the
.Ic special
line.
.Pp
The crunched binary
.Dq kcopy
can be built as follows:
.Bd -literal -offset indent
% crunchgen -m Makefile kcopy.conf    # gen Makefile and kcopy.c
% make objs		# build the component programs' .o files
% make			# build the crunched binary kcopy
% kcopy sh		# test that this invokes a sh shell
$			# it works!
.Ed
.Pp
At this point the binary
.Dq kcopy
can be copied onto an install floppy
and hard-linked to the names of the component programs.
.Sh AUTHORS
.Nm
was written by James da Silva
.Aq jds@cs.umd.edu .
.Pp
Copyright (c) 1994 University of Maryland.  All Rights Reserved.
.Sh CAVEATS
While
.Nm
takes care to eliminate link conflicts between the component programs
of a crunched binary, conflicts are still possible between the
libraries that are linked in.
Some shuffling in the order of libraries may be required,
and in some rare cases two libraries may
have an unresolvable conflict and thus cannot be crunched together.
.Pp
Some versions of the BSD build environment do not by default build the
intermediate object file for single-source file programs.
The
.Dq make objs
target must then be used to get those object files built,
or some other arrangements made.