## # Darwin (Mac OS) hints # Wilfredo Sanchez ## ## # Paths ## # Configure hasn't figured out the version number yet. Bummer. perl_revision=`awk '/define[ ]+PERL_REVISION/ {print $3}' $src/patchlevel.h` perl_version=`awk '/define[ ]+PERL_VERSION/ {print $3}' $src/patchlevel.h` perl_subversion=`awk '/define[ ]+PERL_SUBVERSION/ {print $3}' $src/patchlevel.h` version="${perl_revision}.${perl_version}.${perl_subversion}" # Pretend that Darwin doesn't know about those system calls in Tiger # (10.4/darwin 8) and earlier [perl #24122] case "$osvers" in [1-8].*) d_setregid='undef' d_setreuid='undef' d_setrgid='undef' d_setruid='undef' ;; esac # This was previously used in all but causes three cases # (no -Ddprefix=, -Dprefix=/usr, -Dprefix=/some/thing/else) # but that caused too much grief. # vendorlib="/System/Library/Perl/${version}"; # Apple-supplied modules # BSD paths case "$prefix" in '') # Default install; use non-system directories prefix='/usr/local'; siteprefix='/usr/local'; ;; '/usr') # We are building/replacing the built-in perl prefix='/'; installprefix='/'; bin='/usr/bin'; siteprefix='/usr/local'; # We don't want /usr/bin/HEAD issues. sitebin='/usr/local/bin'; sitescript='/usr/local/bin'; installusrbinperl='define'; # You knew what you were doing. privlib="/System/Library/Perl/${version}"; sitelib="/Library/Perl/${version}"; vendorprefix='/'; usevendorprefix='define'; vendorbin='/usr/bin'; vendorscript='/usr/bin'; vendorlib="/Network/Library/Perl/${version}"; # 4BSD uses ${prefix}/share/man, not ${prefix}/man. man1dir='/usr/share/man/man1'; man3dir='/usr/share/man/man3'; # But users' installs shouldn't touch the system man pages. # Transient obsoleted style. siteman1='/usr/local/share/man/man1'; siteman3='/usr/local/share/man/man3'; # New style. siteman1dir='/usr/local/share/man/man1'; siteman3dir='/usr/local/share/man/man3'; ;; *) # Anything else; use non-system directories, use Configure defaults ;; esac ## # Tool chain settings ## # Since we can build fat, the archname doesn't need the processor type archname='darwin'; # nm isn't known to work after Snow Leopard and XCode 4; testing with OS X 10.5 # and Xcode 3 shows a working nm, but pretending it doesn't work produces no # problems. usenm='false'; case "$optimize" in '') # Optimizing for size also mean less resident memory usage on the part # of Perl. Apple asserts that this is a more important optimization than # saving on CPU cycles. Given that memory speed has not increased at # pace with CPU speed over time (on any platform), this is probably a # reasonable assertion. if [ -z "${optimize}" ]; then case "`${cc:-gcc} -v 2>&1`" in *"gcc version 3."*) optimize='-Os' ;; *) optimize='-O3' ;; esac else optimize='-O3' fi ;; esac # -fno-common because common symbols are not allowed in MH_DYLIB # -DPERL_DARWIN: apparently the __APPLE__ is not sanctioned by Apple # as the way to differentiate Mac OS X. (The official line is that # *no* cpp symbol does differentiate Mac OS X.) ccflags="${ccflags} -fno-common -DPERL_DARWIN" # At least on Darwin 1.3.x: # # # define INT32_MIN -2147483648 # int main () { # double a = INT32_MIN; # printf ("INT32_MIN=%g\n", a); # return 0; # } # will output: # INT32_MIN=2.14748e+09 # Note that the INT32_MIN has become positive. # INT32_MIN is set in /usr/include/stdint.h by: # #define INT32_MIN -2147483648 # which seems to break the gcc. Defining INT32_MIN as (-2147483647-1) # seems to work. INT64_MIN seems to be similarly broken. # -- Nicholas Clark, Ken Williams, and Edward Moy # # This seems to have been fixed since at least Mac OS X 10.1.3, # stdint.h defining INT32_MIN as (-INT32_MAX-1) # -- Edward Moy # case "$(grep '^#define INT32_MIN' /usr/include/stdint.h)" in *-2147483648) ccflags="${ccflags} -DINT32_MIN_BROKEN -DINT64_MIN_BROKEN" ;; esac # Avoid Apple's cpp precompiler, better for extensions if [ "X`echo | ${cc} -no-cpp-precomp -E - 2>&1 >/dev/null`" = "X" ]; then cppflags="${cppflags} -no-cpp-precomp" # This is necessary because perl's build system doesn't # apply cppflags to cc compile lines as it should. ccflags="${ccflags} ${cppflags}" fi # Known optimizer problems. case "`cc -v 2>&1`" in *"3.1 20020105"*) toke_cflags='optimize=""' ;; esac # Shared library extension is .dylib. # Bundle extension is .bundle. ld='cc'; so='dylib'; dlext='bundle'; usedl='define'; # 10.4 can use dlopen. # 10.4 broke poll(). case "$osvers" in [1-7].*) dlsrc='dl_dyld.xs'; ;; *) dlsrc='dl_dlopen.xs'; d_poll='undef'; i_poll='undef'; ;; esac case "$ccdlflags" in # If passed in from command line, presume user knows best '') cccdlflags=' '; # space, not empty, because otherwise we get -fpic ;; esac # Perl bundles do not expect two-level namespace, added in Darwin 1.4. # But starting from perl 5.8.1/Darwin 7 the default is the two-level. case "$osvers" in 1.[0-3].*) lddlflags="${ldflags} -bundle -undefined suppress" ;; 1.*) ldflags="${ldflags} -flat_namespace" lddlflags="${ldflags} -bundle -undefined suppress" ;; [2-6].*) ldflags="${ldflags} -flat_namespace" lddlflags="${ldflags} -bundle -undefined suppress" ;; *) lddlflags="${ldflags} -bundle -undefined dynamic_lookup" case "$ld" in *MACOSX_DEVELOPMENT_TARGET*) ;; *) ld="env MACOSX_DEPLOYMENT_TARGET=10.3 ${ld}" ;; esac ;; esac ldlibpthname='DYLD_LIBRARY_PATH'; # useshrplib=true results in much slower startup times. # 'false' is the default value. Use Configure -Duseshrplib to override. cat > UU/archname.cbu <<'EOCBU' # This script UU/archname.cbu will get 'called-back' by Configure # after it has otherwise determined the architecture name. case "$ldflags" in *"-flat_namespace"*) ;; # Backward compat, be flat. # If we are using two-level namespace, we will munge the archname to show it. *) archname="${archname}-2level" ;; esac EOCBU # 64-bit addressing support. Currently strictly experimental. DFD 2005-06-06 case "$use64bitall" in $define|true|[yY]*) case "$osvers" in [1-7].*) cat <&4 *** 64-bit addressing is not supported for Mac OS X versions *** below 10.4 ("Tiger") or Darwin versions below 8. Please try *** again without -Duse64bitall. (-Duse64bitint will work, however.) EOM exit 1 ;; *) case "$osvers" in 8.*) cat <&4 *** Perl 64-bit addressing support is experimental for Mac OS X *** 10.4 ("Tiger") and Darwin version 8. System V IPC is disabled *** due to problems with the 64-bit versions of msgctl, semctl, *** and shmctl. You should also expect the following test failures: *** *** ext/threads-shared/t/wait (threaded builds only) EOM [ "$d_msgctl" ] || d_msgctl='undef' [ "$d_semctl" ] || d_semctl='undef' [ "$d_shmctl" ] || d_shmctl='undef' ;; esac case `uname -p` in powerpc) arch=ppc64 ;; i386) arch=x86_64 ;; *) cat <&4 *** Don't recognize processor, can't specify 64 bit compilation. EOM ;; esac for var in ccflags cppflags ld ldflags do eval $var="\$${var}\ -arch\ $arch" done ;; esac ;; esac ## # System libraries ## # vfork works usevfork='true'; # malloc wrap works case "$usemallocwrap" in '') usemallocwrap='define' ;; esac # our malloc works (but allow users to override) case "$usemymalloc" in '') usemymalloc='n' ;; esac # However sbrk() returns -1 (failure) somewhere in lib/unicore/mktables at # around 14M, so we need to use system malloc() as our sbrk() malloc_cflags='ccflags="-DUSE_PERL_SBRK -DPERL_SBRK_VIA_MALLOC $ccflags"' # Locales aren't feeling well. LC_ALL=C; export LC_ALL; LANG=C; export LANG; # # The libraries are not threadsafe as of OS X 10.1. # # Fix when Apple fixes libc. # case "$usethreads$useithreads" in *define*) case "$osvers" in [12345].*) cat <&4 *** Warning, there might be problems with your libraries with *** regards to threading. The test ext/threads/t/libc.t is likely *** to fail. EOM ;; *) usereentrant='define';; esac esac # Fink can install a GDBM library that claims to have the ODBM interfaces # but Perl dynaloader cannot for some reason use that library. We don't # really need ODBM_FIle, though, so let's just hint ODBM away. i_dbm=undef; # Configure doesn't detect ranlib on Tiger properly. # NeilW says this should be acceptable on all darwin versions. ranlib='ranlib' ## # Build process ## # Case-insensitive filesystems don't get along with Makefile and # makefile in the same place. Since Darwin uses GNU make, this dodges # the problem. firstmakefile=GNUmakefile;