?RCS: $Id$ ?RCS: ?RCS: Copyright (c) 1991-1997, 2004-2006, Raphael Manfredi ?RCS: ?RCS: You may redistribute only under the terms of the Artistic License, ?RCS: as specified in the README file that comes with the distribution. ?RCS: You may reuse parts of this distribution only within the terms of ?RCS: that same Artistic License; a copy of which may be found at the root ?RCS: of the source tree for dist 4.0. ?RCS: ?RCS: $Log: libc.U,v $ ?RCS: Revision 3.0.1.10 1997/02/28 15:56:48 ram ?RCS: patch61: replaced .a with $_a all over the place ?RCS: patch61: added support for HPUX-10 nm output ?RCS: ?RCS: Revision 3.0.1.9 1995/07/25 14:11:56 ram ?RCS: patch56: now knows about OS/2 platforms ?RCS: ?RCS: Revision 3.0.1.8 1995/05/12 12:20:47 ram ?RCS: patch54: made sure only most recent version of shared lib is picked ?RCS: patch54: final "nm -p" check now uses xscan and xrun like everybody ?RCS: patch54: can now grok linux nm output with lead __IO (ADO) ?RCS: patch54: added support for linux ELF output, using 'W' for alias (ADO) ?RCS: ?RCS: Revision 3.0.1.7 1994/10/29 16:23:40 ram ?RCS: patch36: now looks for shared libraries before anything else (ADO) ?RCS: patch36: added new nm output format (ADO) ?RCS: ?RCS: Revision 3.0.1.6 1994/08/29 16:28:10 ram ?RCS: patch32: added I-type symbols for nm output parsing on Linux ?RCS: ?RCS: Revision 3.0.1.5 1994/06/20 07:03:24 ram ?RCS: patch30: checks are now presented by succession of if/elif ?RCS: patch30: uniformized checks for shared objects with new so symbol ?RCS: ?RCS: Revision 3.0.1.4 1994/05/06 15:06:33 ram ?RCS: patch23: added shared library knowledge (ADO and WED) ?RCS: ?RCS: Revision 3.0.1.3 1994/01/24 14:12:17 ram ?RCS: patch16: can now export nm_extract as an internal-use only variable ?RCS: ?RCS: Revision 3.0.1.2 1993/09/13 16:09:03 ram ?RCS: patch10: added special handling for Apollo systems (WAD) ?RCS: ?RCS: Revision 3.0.1.1 1993/08/27 14:40:03 ram ?RCS: patch7: added entry for /usr/shlib/libc.so (OSF/1 machines) ?RCS: ?RCS: Revision 3.0 1993/08/18 12:08:57 ram ?RCS: Baseline for dist 3.0 netwide release. ?RCS: ?MAKE:libc libnames +nm_extract: echo n c rm test grep Getfile Myread \ Oldconfig Loc sed libs incpath libpth ar runnm nm nm_opt nm_so_opt \ contains xlibpth so osname trnl tr sort uniq _o _a ?MAKE: -pick add $@ %< ?S:libc: ?S: This variable contains the location of the C library. ?S:. ?S:libnames: ?S: The constructed list of library names. Normally empty, but hint files ?S: may set this for later perusal by Configure. ?S:. ?S:nm_extract: ?S: This variable holds the name of the extraction command used to process ?S: the output of nm and yield the list of defined symbols. It is used ?S: internally by Configure. ?S:. ?T:thislib try libnames xxx xscan xrun thisname com tans file ?F:!libnames !libc.tmp !tmp.imp ?LINT:change nm_opt runnm ?INIT:libnames='' : Figure out where the libc is located case "$runnm" in true) ?X: indentation is wrong on purpose--RAM : get list of predefined functions in a handy place echo " " case "$libc" in '') libc=unknown case "$libs" in *-lc_s*) libc=`./loc libc_s$_a $libc $libpth` esac ;; esac case "$libs" in '') ;; *) for thislib in $libs; do case "$thislib" in -lc|-lc_s) : Handle C library specially below. ;; -l*) thislib=`echo $thislib | $sed -e 's/^-l//'` if try=`./loc lib$thislib.$so.'*' X $libpth`; $test -f "$try"; then : elif try=`./loc lib$thislib.$so X $libpth`; $test -f "$try"; then : elif try=`./loc lib$thislib$_a X $libpth`; $test -f "$try"; then : elif try=`./loc $thislib$_a X $libpth`; $test -f "$try"; then : elif try=`./loc lib$thislib X $libpth`; $test -f "$try"; then : elif try=`./loc $thislib X $libpth`; $test -f "$try"; then : elif try=`./loc Slib$thislib$_a X $xlibpth`; $test -f "$try"; then : else try='' fi libnames="$libnames $try" ;; *) libnames="$libnames $thislib" ;; esac done ;; esac ?X: ?X: Some systems (e.g. DG/UX) use "environmental" links, which make the test ?X: -f fail. Ditto for symbolic links. So in order to reliably check the ?X: existence of a file, we use test -r. It will still fail with DG/UX links ?X: though, but at least it will detect symbolic links. At some strategic ?X: points, we make use of (test -h), using a sub-shell in case builtin test ?X: does not implement the -h check for symbolic links. This makes it ?X: possible to preset libc in a hint file for instance and have it show up ?X: as-is in the question. ?X: xxx=normal case "$libc" in unknown) ?X: ?X: The sed below transforms .so.9 .so.12 into something like .so.0009 .so.0012, ?X: then sorts on it to allow keeping .so.12 instead of .so.9 as the latest ?X: up-to-date library. The initial filename (before sed munging, saved in hold ?X: space via 'h') is appended via 'G' before sorting, then the leading munged ?X: part is removed after sorting. Nice efficient work from Tye McQueen. ?X: The initial blurfl is here to prevent the trailing pipe from producing an ?X: empty string, causing Configure to output all its set variables! ?X: set /lib/libc.$so for xxx in $libpth; do $test -r $1 || set $xxx/libc.$so : The messy sed command sorts on library version numbers. $test -r $1 || \ set `echo blurfl; echo $xxx/libc.$so.[0-9]* | \ tr ' ' $trnl | egrep -v '\.[A-Za-z]*$' | $sed -e ' h s/[0-9][0-9]*/0000&/g s/0*\([0-9][0-9][0-9][0-9][0-9]\)/\1/g G s/\n/ /' | \ $sort | $sed -e 's/^.* //'` eval set \$$# done $test -r $1 || set /usr/ccs/lib/libc.$so $test -r $1 || set /lib/libsys_s$_a ;; *) ?X: ensure the test below for the (shared) C library will fail set blurfl ;; esac if $test -r "$1"; then echo "Your (shared) C library seems to be in $1." libc="$1" elif $test -r /lib/libc && $test -r /lib/clib; then ?X: ?X: Apollo has its C library in /lib/clib AND /lib/libc ?X: not to mention its math library in /lib/syslib... ?X: echo "Your C library seems to be in both /lib/clib and /lib/libc." xxx=apollo libc='/lib/clib /lib/libc' if $test -r /lib/syslib; then echo "(Your math library is in /lib/syslib.)" ?X: Put syslib in libc -- not quite right, but won't hurt libc="$libc /lib/syslib" fi elif $test -r "$libc" || (test -h "$libc") >/dev/null 2>&1; then echo "Your C library seems to be in $libc, as you said before." ?X: For mips, and... elif $test -r $incpath/usr/lib/libc$_a; then libc=$incpath/usr/lib/libc$_a; echo "Your C library seems to be in $libc. That's fine." elif $test -r /lib/libc$_a; then libc=/lib/libc$_a; echo "Your C library seems to be in $libc. You're normal." else if tans=`./loc libc$_a blurfl/dyick $libpth`; $test -r "$tans"; then : elif tans=`./loc libc blurfl/dyick $libpth`; $test -r "$tans"; then libnames="$libnames "`./loc clib blurfl/dyick $libpth` elif tans=`./loc clib blurfl/dyick $libpth`; $test -r "$tans"; then : elif tans=`./loc Slibc$_a blurfl/dyick $xlibpth`; $test -r "$tans"; then : elif tans=`./loc Mlibc$_a blurfl/dyick $xlibpth`; $test -r "$tans"; then : else tans=`./loc Llibc$_a blurfl/dyick $xlibpth` fi if $test -r "$tans"; then echo "Your C library seems to be in $tans, of all places." libc=$tans else libc='blurfl' fi fi if $test $xxx = apollo -o -r "$libc" || (test -h "$libc") >/dev/null 2>&1; then dflt="$libc" cat < libpath cat >&4 < libnames set X `cat libnames` shift xxx=files case $# in 1) xxx=file; esac echo "Extracting names from the following $xxx for later perusal:" >&4 echo " " $sed 's/^/ /' libnames >&4 echo " " $echo $n "This may take a while...$c" >&4 ?X: ?X: Linux may need the special Dynamic option to nm for shared libraries. ?X: In general, this is stored in the nm_so_opt variable. ?X: Unfortunately, that option may be fatal on non-shared libraries. ?X: for file in $*; do case $file in *$so*) $nm $nm_so_opt $nm_opt $file 2>/dev/null;; *) $nm $nm_opt $file 2>/dev/null;; esac done >libc.tmp $echo $n ".$c" ?X: ?X: To accelerate processing, we look at the correct 'sed' command ?X: by using a small subset of libc.tmp, i.e. fprintf function. ?X: When we know which sed command to use, do the name extraction ?X: $grep fprintf libc.tmp > libc.ptf ?X: ?X: In order to enhance readability and save some space, we define ?X: some variables that will be "eval"ed. ?X: xscan='eval "libc.list"; $echo $n ".$c" >&4' xrun='eval "libc.list"; echo "done." >&4' ?X: BSD-like output, I and W types added for Linux ?X: Some versions of Linux include a leading __IO in the symbol name. ?X: HPUX10 reportedly has trailing spaces, though I'm surprised it has ?X: BSD-like output. (AD). xxx='[ADTSIW]' if com="$sed -n -e 's/__IO//' -e 's/^.* $xxx *_[_.]*//p' -e 's/^.* $xxx *//p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun ?X: SYSV-like output elif com="$sed -n -e 's/^__*//' -e 's/^\([a-zA-Z_0-9$]*\).*xtern.*/\1/p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun elif com="$sed -n -e '/|UNDEF/d' -e '/FUNC..GL/s/^.*|__*//p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun elif com="$sed -n -e 's/^.* D __*//p' -e 's/^.* D //p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun elif com="$sed -n -e 's/^_//' -e 's/^\([a-zA-Z_0-9]*\).*xtern.*text.*/\1/p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun elif com="$sed -n -e 's/^.*|FUNC |GLOB .*|//p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun elif com="$grep '|' | $sed -n -e '/|COMMON/d' -e '/|DATA/d' \ -e '/ file/d' -e 's/^\([^ ]*\).*/\1/p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun elif com="$sed -n -e 's/^.*|FUNC |GLOB .*|//p' -e 's/^.*|FUNC |WEAK .*|//p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun ?X: mips nm output (sysV) elif com="$sed -n -e 's/^__//' -e '/|Undef/d' -e '/|Proc/s/ .*//p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun elif com="$sed -n -e 's/^.*|Proc .*|Text *| *//p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun elif com="$sed -n -e '/Def. Text/s/.* \([^ ]*\)\$/\1/p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun ?X: OS/2 nm output elif com="$sed -n -e 's/^[-0-9a-f ]*_\(.*\)=.*/\1/p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun elif com="$sed -n -e 's/.*\.text n\ \ \ \.//p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun ?X: AIX nm output elif com="sed -n -e 's/^__.*//' -e 's/[ ]*D[ ]*[0-9]*.*//p'";\ eval $xscan;\ $contains '^fprintf$' libc.list >/dev/null 2>&1; then eval $xrun else $nm -p $* 2>/dev/null >libc.tmp $grep fprintf libc.tmp > libc.ptf if com="$sed -n -e 's/^.* [ADTSIW] *_[_.]*//p' -e 's/^.* [ADTSIW] //p'";\ eval $xscan; $contains '^fprintf$' libc.list >/dev/null 2>&1 then nm_opt='-p' eval $xrun else echo " " echo "$nm didn't seem to work right. Trying $ar instead..." >&4 com='' if $ar t $libc > libc.tmp && \ $contains '^fprintf$' libc.tmp >/dev/null 2>&1 then for thisname in $libnames $libc; do $ar t $thisname >>libc.tmp done $sed -e "s/\\$_o\$//" < libc.tmp > libc.list echo "Ok." >&4 elif test "X$osname" = "Xos2" && $ar tv $libc > libc.tmp; then ?X: Repeat libc to extract forwarders to DLL entries too for thisname in $libnames $libc; do $ar tv $thisname >>libc.tmp ?X: Revision 50 of EMX has bug in ar: it will not extract forwarders ?X: to DLL entries. Use emximp which will extract exactly them. emximp -o tmp.imp $thisname \ 2>/dev/null && \ $sed -e 's/^\([_a-zA-Z0-9]*\) .*$/\1/p' \ < tmp.imp >>libc.tmp $rm -f tmp.imp done $sed -e "s/\\$_o\$//" -e 's/^ \+//' < libc.tmp > libc.list echo "Ok." >&4 else echo "$ar didn't seem to work right." >&4 echo "Maybe this is a Cray...trying bld instead..." >&4 if bld t $libc | \ $sed -e 's/.*\///' -e "s/\\$_o:.*\$//" > libc.list && $test -s libc.list then for thisname in $libnames; do bld t $libnames | \ $sed -e 's/.*\///' -e "s/\\$_o:.*\$//" >>libc.list $ar t $thisname >>libc.tmp done echo "Ok." >&4 else echo "That didn't work either." >&4 echo "No problem, I'll be compiling test programs then..." >&4 runnm=false fi fi fi fi nm_extract="$com" if $test -f /lib/syscalls.exp; then echo " " echo "Also extracting names from /lib/syscalls.exp for good ole AIX..." >&4 $sed -n 's/^\([^ ]*\)[ ]*syscall[0-9]*[ ]*$/\1/p' \ /lib/syscalls.exp >>libc.list fi ?X: remember, indentation is wrong on purpose--RAM ;; esac $rm -f libnames libpath