Vous souhaitez optimiser la place. Une des façons de le faire est de choisir les commandes que vous voulez laisser. Par exemple, je pense que ping, ifconfig ou tail ne sont pas forcément utiles pour un lecteur multimédia.
Pour cela, il suffit de prendre les commandes que l’on veut et les mettre dans /bin/ et /sbin/. C’est cool mais il y a un pb!!!!
Il faut les librairies pour pouvoir faire fonctionner ces commandes. Pour les trouver, vous pouvez taper:
| ldd + nom_commande |
et copier à la main les librairies.
Ou alors, vous pouvez utiliser le script mklibs.sh
J’ai eu du mal à le trouver sur le net. C’est un script qui fait parti du projet Debian. Vous pouvez donc le trouver sur leur site. Au détail près que ce que j’ai trouvé, c’est du python. Et c’est pas vraiment ce que je cherchais.
J’ai mis les sources de ce srcipt à la fin de ce post.
Pour utiliser ce script, c’est très simple:
| mklibs -v -d DOSSIER_LIB/ DOSSIER_COMMANDES/* |
Par exemple, si je veux trouver toutes les librairies nécessaires au fonctionnement des commandes qui sont dans /mnt/emb/bin et /mnt/emb/sbin/ et mettre ces librairies dans /mnt/emb/lib/, il faut taper:
mklibs -v -d /mnt/emb/lib /mnt/emb/bin/* /mnt/emb/sbin/*
Autre conseil: Si vous le faîtes à la main, pensez surtout à mettre les permissions en exécution des librairies par un ‘chmod a+x /mnt/emb/lib/*’
et surtout bon courage
Une idée si vous voulez vraiment gagner de l’espace, est d’utiliser la busybox. Elle permet de générer un certain nombre de commandes pesant beaucoup moins que leurs cousines.
Voilà donc le code du script ‘mklibs.sh’.
Pour l’utilser, c’est très simple, vous copiez le contenu du tableau dans un fichier que vous renommez ‘mklibs.sh’ et vous l’autorisez en exécution et voilou. Tout chaud pour êter utilisé.
| mklibs.sh |
| #!/bin/bash
# # — NO-ROCK-COPYRIGHT-NOTE — # # Found in the debian boot-floppies source package from: # http://www.kernel.org/debian/dists/potato/main/source/admin/ # # mklibs.sh: An automated way to create a minimal /lib/ directory. # # Copyright 1999 by Marcus Brinkmann # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 of the License, or # (at your option) any later version. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License # along with this program; if not, write to the Free Software # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA # # Introduction: # When creating boot floppies, there is never enough room on the disks. # So it is important not to waste bytes on unnecessary code. # Shared Libraries contain many functions that are probably not used in the # binaries included in the boot disks, and copying the whole library is a # waste of space. # This utilitiy helps to reduce the necessary libraries to only include the # symbols needed to run a given set of executables. # # Features: # * Automatic detection of all necessary libraries, even for inter-library # dependencies, for a given set of executables. # * Automatic installation of all needed libraries and soname links. # * Automatic reduction of all libraries to minimal size for which PIC # libraries are provided. # # Requirements: # * Beside the shared libraries, you need libfoo_pic.a files for all # libraries you want to reduce. # * You need binutils (notably objdump and objcopy) installed. # A GENERAL NOTE ABOUT LANGUAGE ABUSE # # If you believe this program had better not been written in shell script # language, I invite you to reimplement it in the language of your # preference. # The reasons I chose shell are: # * Shell scripts are very portable and available even on minimal systems as # well as boot disks. # * Shell scripts can be run without compilation. # * Shell scripts provide a very easy interface to the various system # commands I need to get the library dependencies and sort through them. # Perl is lacking good data types, so implementing this would be equally # cumbersome in perl (I need trees and queues, for example). # C and C++ are lacking easy access to the system commands. # # Of course, shell scripting has many problems: # * Use of temporary files for string arrays. # * Slow. # * Hard to debug. # # I think for hand written code, I hit a limit with the size and execution # time of this program of what is still acceptable as a shell script. I also # tried to improve the situation with many comments. # TODO: # * Make sure versioned symbols get correct version number. # This seems to work now, however, we always include # all versions of a symbol. This is not a problem. To do # it properly, we had to parse the version information in # objdump, which is hard. # * Use –dynamic-syms on so lib instead –syms on pic file. # * Autodetect that libc needs ld (should be possible from # output of objdump –privat-headers| grep NEEDD). # * Code to create libs in cycles !!! # HISTORY: # # 1999-09-13 Marcus Brinkmann # # * Initial release (v0.1). # # STATIC DATA SECTION # usage= »Usage: $0 [OPTION]… -d DEST FILE … » try= »Try « \` »$0 –help’ for more information » version= »$0 0.1, Copyright 1999 Marcus Brinkmann » PATH=/bin:/usr/bin default_src_path=/lib:/usr/lib dest= » » exec= » » action= » » verbose= »false » gcc=${GCC-gcc} objdump=${OBJDUMP-objdump} objcopy=${OBJCOPY-objcopy} # ================= # GRAPH ABSTRACTION # ================= # # Because we do some hairy graph operations, we provide some # abstractions of them. Some functions here are very simple, but # the source is much more readable this way. # check-node NODE … # checks if all NODEs are valid node names. # Used internally for verificaton only. # Return 0 if all NODEs are valid. # Currently, a node is valid if it does not contain a space. check-node () { local node for node in « $@ » ; do if [ « x`echo $node | sed -e ‘/ /d’` » = x ] ; then echo 1>&2 $0: check-node: invalid node \ »$node\ » exit 1 fi done return 0 } # is-graph FILE … # provides a very simple type assertion # Turns FILE into a graph if it isn’t already and returns 0. is-graph () { local file for file in « $@ » ; do if [ ! -e « $file » ] ; then touch « $qfile » fi done } # add-node FILE NODE # add a node NODE to graph FILE. # This is useful if you need to make sure that a node appears # in the graph without actually connecting it to an arrow. # You don’t need to add nodes that are part of an arrow. add-node () { if [ $# != 2 ] ; then echo 1>&2 $0: add-node: internal error: called with invalid number of arguments exit 1 fi check-node « $2 » echo « $2 $2 » >> « $1 » return 0 } # add-arrow FILE NODE1 NODE2 # add an arrow from NODE1 to NODE2 to graph FILE. add-arrow () { if [ $# != 3 ] ; then echo 1>&2 $0: add-arrow: internal error: called with invalid number of arguments exit 1 fi check-node « $2 » « $3 » echo « $2 $3 » >> « $1 » return 0 } # find-cycle FILE # finds a cycle in a graph FILE. # If a cycle is found, it is printed out at stdin, one node each line, # and 0 is returned. Otherwise, nothing is printed on stdout and exit # status is 1. find-cycle () { if [ $# != 1 ] ; then echo 1>&2 $0: find-cycle: internal error: called with invalid number of arguments exit 1 fi tsort « $1 » 2> « $fl_dir/find-cycle » > /dev/null if [ « x`cat $fl_dir/find-cycle` » = x ] ; then return 1 else if [ « x`head -n 1 $fl_dir/find-cycle` » != « xtsort: cycle in data » ] ; then echo 1>&2 $0: find-cycle: internal error: tsort has invalid output format exit 1 fi cat « $fl_dir/find-cycle » | sed -e ‘1d’ -e ‘/tsort: cycle in data/,$d’ -e ‘s/^tsort: //’ fi } # shrink-nodes FILE NODE1 … # shrinks several nodes NODE1 … to a single node in graph FILE. # To hide cycles, we treat a cycle as a single node and replace # each occurence of a node in the cycle with a new node # [NODE1,…] . This change is destructive and can not be undone! # (You would need to store the entry point to the cycle for each arrow # pointing to/from it). # This function does not check if the the nodes NODE1 … exist. # However, if none of these nodes exists already, the new node will # not appear either. This makes this function sort of idem potent. # It does not check if NODE1 … are a cycle. We will assume this # later in the library dependency analysis, but nothing in the code # relies on it. # Always shrink all cycles, or you may get unresolved symbols. # # Example: # N1 —> N2 N1 ——-> /————\ # | « shrink-nodes N2 N4 » | _ | [N2,N4] | # v ——————-> v _____/| \————/ # N3 —> N4 N3 / # A small helper function will aid us… # equal-match STRING STRING1 … # return 0 if STRING is among STRING1 …, 1 otherwise. equal-match () { local string local stringk string= »$1″ shift for stringk in « $@ » ; do if [ « x$string » = « x$stringk » ] ; then return 0 fi done return 1 } shrink-nodes () { local head local lnode local rnode local graph= »$1″ shift is-graph « $graph » check-node « $@ » local cnode= »[`echo « $@ » | sed ‘s/ /,/g’`] » # Okay, it’s a hack. We treat the graph as a queue. I am just too # lazy to copy the relevant code here. Of course, we exploit several # properties of the graph and queue file format here (for example, # that graphs never can contain a QUEUE_SEPERATOR, and that a graph is # really a simple file with « a b » entries). cat /dev/null > « $fl_dir/shrink-cycle » while head=`get-top-of-queue « $graph »` ; do lnode=`echo $head|sed ‘s/ [^ ]*$//’` if equal-match « $lnode » « $@ » ; then lnode= »$cnode » fi rnode=`echo $head|sed ‘s/^[^ ]* //’` if equal-match « $rnode » « $@ » ; then rnode= »$cnode » fi echo « $lnode $rnode » >> « $fl_dir/shrink-cycle » done cat « $fl_dir/shrink-cycle » | sort -u > « $graph » } # ================= # QUEUE ABSTRACTION # ================= # # I added an abstract interface for queues to make the code more readable. # Queue operations usually consist of several atomic file operations, which # can get quite messy. # # You can use queues to simply loop through all lines of a file, but you # also can add stuff to the queue while processing it. # # Implementation: All queues consist of a QUEUE_FILE which has two parts: # the remaining entries in the queue (QUEUE) and the already processed # entries (BUCKET). # The two parts are seperated by a line containing only QUEUE_SEPERATOR. QUEUE_SEPERATOR=SEPERATOR___ABOVE_IS_QUEUE__BELOW_IS_BUCKET___SEPERATOR # check-queue-entry QENTRY … # checks if all queue entries QENTRY are valid. # Used internally for verificaton only. # Return 0 if all QENTRYs are valid. # Currently, a node is valid if it does not match the QUEUE_SEPERATOR. check-queue-entry () { local qentry for qentry in « $@ » ; do if [ « x`echo $qentry | sed « /^$QUEUE_SEPERATOR$/d »` » = x ] ; then echo 1>&2 $0: check-queue-entry: invalid qentry name \ »$qentry\ » exit 1 fi done return 0 } # is-queue QUEUE_FILE … # provides a very simple type assertion # Turns QUEUE_FILE into a queue if it isn’t already and returns 0. is-queue () { local qfile for qfile in « $@ » ; do if [ ! -e « $qfile » ] ; then echo « $QUEUE_SEPERATOR » > « $qfile » else if ! grep -q « ^$QUEUE_SEPERATOR$ » « $qfile » ; then echo « $QUEUE_SEPERATOR » >> « $qfile »; fi fi done } # get-top-of-queue QUEUE_FILE # processes a queue one more time. # If QUEUE of QUEUE_FILE is empty, exit status is 1 and no output is given. # Otherwise, top of QUEUE is removed, returned on stdout and # appended to the end of the BUCKET part of QUEUE_FILE. get-top-of-queue () { if [ $# != 1 ] ; then echo 1>&2 $0: get-top-of-queue: internal error: called with invalid number of arguments exit 1 fi is-queue « $1 » local head=`head -n 1 « $1″` if [ « x$head » = « x$QUEUE_SEPERATOR » ] ; then return 1 else sed -e 1d « $1 » > « $fl_dir/get-top-of-queue » echo « $head » | tee –append « $fl_dir/get-top-of-queue » cat « $fl_dir/get-top-of-queue » > « $1 » return 0 fi } # add-to-queue-if-not-there QUEUE_FILE QENTRY … # add queue entries QENTRY … to the beginning of the # QUEUE of QUEUE_FILE if it is neither in QUEUE nor in BUCKET # of QUEUE_FILE. # Return with exit status 0. # Note: If you want to add QENTRY to the *end* of QUEUE, you would do # something like the following: # sed -e s/^$QUEUE_SEPERATOR$/$head »‘\ # ‘ »$QUEUE_SEPERATOR/ » # which is necessary to pass the newline to sed. I think we can take the # easy way out. add-to-queue-if-not-there () { local qentry local qfile= »$1″ shift check-queue-entry « $@ » is-queue « $qfile » for qentry in « $@ » ; do if ! grep -q « ^$qentry\$ » « $qfile » ; then echo « $qentry » > « $fl_dir/add-to-queue-if-not-there » cat « $qfile » >> « $fl_dir/add-to-queue-if-not-there » cat « $fl_dir/add-to-queue-if-not-there » > « $qfile » fi done return 0 } # ================== # LIBRARY PROCESSING # ================== # # The following helper functions mess around with the actual # processing and installation of libraries. # # get-library-depends OBJ1 … # get all libraries the objects OBJ1 … depend on. # OBJs can be binaries or shared libraries. # The list is neither sort’ed nor uniq’ed. get-library-depends () { if [ $# = 0 ] ; then echo 1>&2 $0: get-library-depends: internal error: no arguments exit 1 fi $objdump –private-headers « $@ » 2> /dev/null \ | sed -n ‘s/^ *NEEDED *\([^ ]*\)$/\1/p’ } # get-undefined-symbols OBJ1 … # get all unresolved symbols in OBJ1 … # The list is neither sort’ed nor uniq’ed. get-undefined-symbols () { if [ $# = 0 ] ; then echo 1>&2 $0: get-undefined-symbols: internal error: no arguments exit 1 fi # ash has undefined reference to sys_siglist if .bss is not mentioned # here. Reported by Joel Klecker. # All symbols are epxosed, so we just catch all. Suggested by Roland # McGrath. Another thing to try is to investigate –dynamic-reloc. $objdump –dynamic-syms « $@ » 2> /dev/null \ | sed -n ‘s/^.* \([^ ]*\)$/\1/p’ # | sed -n ‘s/^.*[\*UND\*|.bss].* \([^ ]*\)$/\1/p’ } # get-provided-symbols LIB1 LIB2 … # get all symbols available from libraries LIB1 … . # Does only work for pic libraries. # # v Watch the tab stop here. # 00000000 w F .text 00000000 syscall_device_write_request # 00000000 g F .text 0000056c __strtoq_internal get-provided-symbols () { if [ $# = 0 ] ; then echo 1>&2 $0: get-provided-symbols: internal error: no arguments exit 1 fi $objdump –syms « $@ » 2>/dev/null | grep -v ‘\*UND\*’ \ | sed -n ‘s/^[0-9a-f]\+ \(g \| w\) .. .* [0-9a-f]\+ \(0x8[08]\)\? *\([^ ]*\)$/\3/p’ } # Crude hack (?) only used for diagnostic. get-provided-symbols-of-so-lib () { if [ $# = 0 ] ; then echo 1>&2 $0: get-provided-symbols: internal error: no arguments exit 1 fi $objdump –dynamic-syms « $@ » 2>/dev/null \ | sed -e ‘/\*UND\*/d’ | sed -n ‘s/^.* \([^ ]*\)$/\1/p’ } # get-common-symbols FILE1 FILE2 # returns a list of all symbols in FILE1 that appear also in FILE2 # Note: When get-common-symbols returns, FILE1 and FILE2 are « sort -u »‘ed. # Note: Version Information in FILE1 is ignored when comparing. get-common-symbols () { if [ $# != 2 ] ; then echo 1>&2 $0: get-common-symbols: internal error: called with invalid number of arguments exit 1 fi # Not needed anymore, but we go for compatibility. # (Somewhere we HAVE to clean FILE2 up). sort -u « $1 » > $fl_dir/get-common-symbols cat $fl_dir/get-common-symbols > « $1 » sort -u « $2 » > $fl_dir/get-common-symbols cat $fl_dir/get-common-symbols > « $2 » local symbol= while symbol=`get-top-of-queue $fl_dir/get-common-symbols` ; do grep ^$symbol\$\\\|^$symbol@ « $1 » done } # create-link TARGET LINK_NAME # creates a soft link if there isn’t one already. create-link () { if [ $# != 2 ] ; then echo 1>&2 $0: create-link: internal error: called with invalid number of arguments exit 1 fi if [ ! -e « $2 » ] ; then $action ln -s « $1 » « $2 » fi } # find-file PATH FILE # search all directories in PATH for file FILE, return absolute path # FILE can be a relative path and a filename. # PATH is a list, seperator is ‘:’. find-file () { if [ $# != 2 ] ; then echo 1>&2 $0: find-file: internal error: exactly two arguments required exit 1 fi local path=$1 local dir=`echo $path | sed -e ‘s/:.*$//’` until [ « x$path » = x ] ; do if [ « x$dir » != x ] ; then if [ -e « $dir/$2 » ] ; then echo « $dir/$2 » return 0 fi fi path=`echo $path | sed -e ‘s/^[^:]*:*//’` dir=`echo $path | sed -e ‘s/:.*$//’` done return 1 } # find-files PATH FILE1 FILE2 … # search all directories in PATH for file FILE1, FILE2… # FILE can be a relative path and a filename. # PATH is a list, seperator is ‘:’. # Return value is a white space seperated list of absolute filenames. find-files () { if [ $# -lt 2 ] ; then echo 1>&2 $0: find-files: internal error: too few arguments exit 1 fi local path= »$1″ ; shift while [ $# != 0 ] ; do find-file $path $1 shift done } # get-pic-file LIB # returns the filename of the pic archive for LIB. # Note: There doesn’t seem to be any convention, *ick*. get-pic-file () { if [ $# != 1 ] ; then echo 1>&2 $0: get-pic-file: internal error: called with invalid number of arguments exit 1 fi if [ « x$1 » = « xlibc-2.0.7.so » ] ; then # Order does matter! First init, then lib, then fini! echo `find-files $src_path libc_pic/soinit.so libc_pic.a libc_pic/sofini.so` return 0 fi if [ « x$1 » = « xlibc-2.1.2.so » -o « x$1 » = « xlibc-2.1.3.so » \ -o « x$1 » = « xlibc-2.2.so » -o « x$1 » = « xlibc-2.2.1.so » \ -o « x$1 » = « xlibc-2.2.2.so » ] ; then # Order does matter! First init, then lib, then fini! echo `find-files $src_path libc_pic/soinit.o libc_pic.a libc_pic/sofini.o libc_pic/interp.o` return 0 fi if [ « x$1 » = « xlibm-2.1.2.so » -o « x$1 » = « xlibm-2.1.3.so » \ -o « x$1 » = « xlibm-2.2.so » -o « x$1 » = « xlibm-2.2.1.so » \ -o « x$1 » = « xlibm-2.2.2.so » ] ; then echo `find-file « $src_path » libm_pic.a` return 0 fi if [ « x$1 » = « xlibslang.so.1.3.9 » ] ; then echo `find-file $src_path libslang1.3.9_pic.a` return 0 fi if [ « x$1 » = « xlibslang.so.1.4.1 » ] ; then echo `find-file $src_path libslang1.4.1_pic.a` return 0 fi local libname=`echo $1 | sed -e ‘s/^lib\(.*\).so.*/\1/’` echo `find-file « $src_path » lib${libname}_pic.a` return 0 } get-extra-flags () { if [ $# != 1 ] ; then echo 1>&2 $0: get-extra-flags: internal error: called with invalid number of arguments exit 1 fi if [ « x$1 » = « xlibc-2.0.7.so » ] ; then echo `find-file $src_path ld-2.0.7.so` -lgcc return 0 fi if [ « x$1 » = « xlibc-2.1.2.so » ] ; then echo « `find-file $src_path ld-2.1.2.so` -lgcc -Wl,–version-script=`find-file $src_path libc_pic.map` » return 0 fi if [ « x$1 » = « xlibm-2.1.2.so » -o « x$1 » = « xlibm-2.1.3.so » ] ; then echo « -Wl,–version-script=`find-file $src_path libm_pic.map` » return 0 fi if [ « x$1 » = « xlibc-2.1.3.so » ] ; then echo « `find-file $src_path ld-2.1.3.so` -lgcc -Wl,–version-script=`find-file $src_path libc_pic.map` » return 0 fi if [ « x$1 » = « xlibc-2.2.so » ] ; then echo « `find-file $src_path ld-2.2.so` -lgcc -Wl,–version-script=`find-file $src_path libc_pic.map` » return 0 fi if [ « x$1 » = « xlibc-2.2.1.so » ] ; then echo « `find-file $src_path ld-2.2.1.so` -lgcc -Wl,–version-script=`find-file $src_path libc_pic.map` » return 0 fi if [ « x$1 » = « xlibc-2.2.2.so » ] ; then echo « `find-file $src_path ld-2.2.2.so` -lgcc -Wl,–version-script=`find-file $src_path libc_pic.map` » return 0 fi return 0 } # install-small-lib LIB_SONAME # makes a small version of library LIB_SONAME # # This happens the following way: # 0. Make exception for the linker ld. # 1. Try to figure out complete path of pic library. # 2. If no found, copy the shared library, else: # a. Get shared libraries this lib depends on, transform into a # list of « -lfoo » options. # b. Get a list of symbols both provided by the lib and in the undefined # symbols list. # c. Make the library, strip it. # d. Add symbols that are still undefined to the undefined symbols list. # e. Put library into place. install-small-lib () { if [ $# != 1 ] ; then echo 1>&2 $0: install-small-lib: internal error: called with invalid number of arguments exit 1 fi local src_file=`find-file $src_path $1` if `echo « $1 » | grep -q ^ld` ; then get-provided-symbols « $src_file » >> $fl_dir/provided-symbols $action $objcopy –strip-unneeded -R .note -R .comment « $src_file » « $dest/$1 » return 0 fi local pic_objects=`get-pic-file « $1″` local extra_flags=`get-extra-flags « $1″` local architecture=`dpkg –print-architecture` # some arm bins or libs are improperly linked, force -lgcc if [ « $architecture » = arm ]; then extra_flags= »$extra_flags -lgcc » fi if [ « x$pic_objects » = x ] ; then $verbose 2>&1 No pic archive for library « $1 » found, falling back to simple copy. get-provided-symbols-of-so-lib « $src_file » >> $fl_dir/provided-symbols get-undefined-symbols « $src_file » >> $fl_dir/undefined-symbols $action $objcopy –strip-unneeded -R .note -R .comment « $src_file » « $dest/$1 » else $verbose 2>&1 Make small lib from « $pic_objects » in « $dest/$1 ». # XXX: If ld is NEEDED, we need to include it on the gcc command line get-library-depends « $src_file » \ | sed -n -e ‘s/^lib\(.*\)\.so.*$/\1/p’ > $fl_dir/lib-dependencies get-provided-symbols $pic_objects > $fl_dir/lib-provided-symbols # Argument order does matter: get-common-symbols $fl_dir/lib-provided-symbols \ $fl_dir/undefined-symbols > $fl_dir/lib-symbols-to-include ${gcc} \ -nostdlib -nostartfiles -shared \ « -Wl,-soname=$1 » \ `cat $fl_dir/lib-symbols-to-include | sed ‘s/^/-u/’` \ -o $fl_dir/lib-so \ $pic_objects $extra_flags \ « -L$dest » \ -L`echo $src_path | sed -e ‘s/::*/:/g’ -e ‘s/^://’ -e ‘s/:$//’ \ -e ‘s/:/ -L/g’` \ `cat $fl_dir/lib-dependencies | sed ‘s/^/-l/’` \ && $objcopy –strip-unneeded -R .note -R .comment $fl_dir/lib-so $fl_dir/lib-so-stripped \ || { echo 1>&2 $0: install-small-lib: $gcc or $objcopy failed. exit 1 } get-undefined-symbols $fl_dir/lib-so-stripped \ >> $fl_dir/undefined-symbols get-provided-symbols-of-so-lib $fl_dir/lib-so-stripped >> $fl_dir/provided-symbols $action cp $fl_dir/lib-so-stripped « $dest/$1 » fi } # install-libs-in-sphere [LIB1,…] # extracts the libs in a shrinked node and cycles through them until all # possible symbols are resolved. # Always make sure this can be called recursively (from install-libs)! install-libs-in-sphere () { if [ $# != 1 ] ; then echo 1>&2 $0: install-libs-in-sphere: internal error: called with invalid number of arguments exit 1 fi # Unfortunately, we need a small parser here to do the right thing when # spheres are within spheres etc. RegEx simply can’t count brackets. local string=`echo « $1 » | sed -e ‘s/^\[//’ -e ‘s/\]$//’` local char local result= local depth=0 while [ « x$string » != x ] ; do # Jump to next special char for faster operation. # Don’t be confused by the regex, it matches everything but ],[ char=`echo $string | sed -e ‘s/^\([^],[]*\).*$/\1/’` string=`echo $string | sed -e ‘s/^[^],[]*//’` result= »$result$char »; # Read special char char=`echo $string | sed -e ‘s/^\(.\).*$/\1/’` string=`echo $string | sed -e ‘s/^.//’` case « $char » in [) depth=$(($depth+1));; ]) depth=$(($depth-1));; ,) if [ $depth = 0 ] ; then char=’ ‘; fi;; esac result= »$result$char »; done $verbose 2>&1 « RESOLVING LOOP…`echo $result | md5sum` » echo XXX: CODE NOT FINISHED install-libs $result $verbose 2>&1 « END OF LOOP… `echo $result | md5sum` » } # install-libs LIB1 … # goes through an ordered list of libraries and installs them. # Make sure this can be called recursively, or hell breaks loose. # Note that the code is (almost) tail-recursive. I wish I could # write this in Scheme install-libs () { local cur_lib local lib for cur_lib in « $@ » ; do if echo « $cur_lib » | grep -q ‘^\[‘ ; then install-libs-in-sphere « $cur_lib » else lib=`find-file $src_path $cur_lib` if [ -L « $lib » ] ; then lib=`basename \`readlink $lib\« create-link $lib $dest/$cur_lib else install-small-lib $cur_lib fi fi done } # # MAIN PROGRAM # # 1. Option Processing # 2. Data Initialization # 3. Graph Construction and Reduction # 4. Library Installation # Global Files: # $fl_dir/undefined-symbols # Holds all undefined symbols we consider for inclusion. # Only grows. Does not to be sort’ed and uniq’ed, but will # get occasionally. # $fl_dir/provided-symbols # Holds all defined symbols we included. # Only grows. Should later be a superset of undefined-symbols. # But some weak symbols may be missing! # $fl_dir/library-depends # Queue of all libraries to consider. # # 1. Option Processing # while :; do case « $1 » in -L) src_path= »$src_path:$2″; shift 2;; -d|–dest-dir) dest=$2; shift 2;; -n|–dry-run) action= »echo »; shift;; -v|–verbose) verbose= »echo »; shift;; -V|–version) echo « $version »; exit 1;; -h|–help) echo « $usage » echo « Make a set of minimal libraries for FILE … in directory DEST. » echo » echo « \ Options: -L DIRECTORY Add DIRECTORY to library search path. -n, –dry-run Don’t actually run any commands; just print them. -v, –verbose Print additional progress information. -V, –version Print the version number and exit. -h, –help Print this help and exit. -d, –dest-dir DIRECTORY Create libraries in DIRECTORY. Required arguments for long options are also mandatory for the short options. » exit 0;; -*) echo 1>&2 $0: $1: unknown flag; echo 1>&2 « $usage »; echo 1>&2 « $try »; exit 1;; ?*) exec= »$exec $1″; shift;; *) break;; esac done src_path=${src_path-$default_src_path} if [ « x$exec » = x ] ; then exit 0 fi if [ « x$dest » = x ] ; then echo 1>&2 $0: no destination directory given; echo 1>&2 « $usage »; exit 1 fi # # 2. Data Initialization # $verbose -n 2>&1 « Initializing data objects… « # Temporary directory. fl_dir= »/tmp/,mklibs.$$ » set -e mkdir $fl_dir set +e trap « rm -fr $fl_dir » EXIT # Intialize our symbol array and library queue with the information # from the executables. get-undefined-symbols $exec > $fl_dir/undefined-symbols add-to-queue-if-not-there $fl_dir/library-depends `get-library-depends $exec` $verbose 2>&1 « done. » # # 3.a Graph Construction # # Build the dependency graph, add new library dependencies to the queue on # the way. # If the soname is a link, add the target to the end of the queue and # add a simple arrow to the graph. # If the soname is a real lib, get its dependencies and add them to # the queue. Furthermore, add arrows to the graph. If the lib is not # dependant on any other lib, add the node to make sure it is mentioned # at least once in the graph. $verbose -n 2>&1 « Constructing dependency graph… (« while cur_lib=`get-top-of-queue $fl_dir/library-depends` do lib=`find-file $src_path $cur_lib` if [ -L « $lib » ] ; then $verbose -n 2>&1 L lib=`basename \`readlink $lib\« add-to-queue-if-not-there $fl_dir/library-depends « $lib » add-arrow $fl_dir/dependency-graph « $cur_lib » « $lib » else get-library-depends « $lib » > $fl_dir/backup if [ « x`head -n 1 $fl_dir/backup` » = x ] ; then $verbose -n 2>&1 N add-node $fl_dir/dependency-graph « $cur_lib » else $verbose -n 2>&1 A for lib in `cat $fl_dir/backup` ; do add-to-queue-if-not-there $fl_dir/library-depends « $lib » add-arrow $fl_dir/dependency-graph « $cur_lib » « $lib » done fi fi done $verbose 2>&1 « ) done. » # # 3.b Graph Reduction # # Find and shrink cycles in the graph. $verbose -n 2>&1 « Eliminating cycles… (« while cycle=`find-cycle « $fl_dir/dependency-graph »` ; do $verbose -n 2>&1 C shrink-nodes « $fl_dir/dependency-graph » $cycle done $verbose 2>&1 « ) done. » # # 4. Library Installation # # Let tsort(1) do the actual work on the cycle-free graph. tsort $fl_dir/dependency-graph > $fl_dir/backup # Now the ordered list of libraries (or cycles of them) # can be processed by install-libs. This is indeed the last step. install-libs `cat $fl_dir/backup` #sort -u $fl_dir/provided-symbols > $fl_dir/diag1 #sort -u $fl_dir/undefined-symbols > $fl_dir/diag2 #cat $fl_dir/diag1 $fl_dir/diag2 | sort | uniq -u > $fl_dir/diag3 ## diag3 has now the symmetric difference. #cat $fl_dir/diag3 $fl_dir/diag2 | sort | uniq -d > $fl_dir/diag1 ## diag1 has now all undefined symbols that are not provided. ##cat $fl_dir/diag1 | wc ## Note that some of these symbols are weak and not having them is probably ## not an error. exit 0 |