willing to test gcc42 on Mac OS X 10.3.9, PowerBook G4

[yaseppochi (Stephen J. Turnbull)]

If you have anything specific I should look for, please let me know. Here's progress to date

I've uncommented the darwin 7 variant in the Portfile. This gives the following kind of error in the build:

if [ x"" != x ]; then \
  /opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/build/./prev-gcc/xgcc -B/opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/build/./prev-gcc/ -B/opt/local/powerpc-apple-darwin7.9.0/bin/ -c -DHAVE_CONFIG_H -g -O2 -I. -I/opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/gcc-4.2-20060204/libiberty/../include  -W -Wall -pedantic -Wwrite-strings -Wstrict-prototypes -Wc++-compat  /opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/gcc-4.2-20060204/libiberty/floatformat.c -o pic/floatformat.o; \
else true; fi
/opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/build/./prev-gcc/xgcc -B/opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/build/./prev-gcc/ -B/opt/local/powerpc-apple-darwin7.9.0/bin/ -c -DHAVE_CONFIG_H -g -O2 -I. -I/opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/gcc-4.2-20060204/libiberty/../include  -W -Wall -pedantic -Wwrite-strings -Wstrict-prototypes -Wc++-compat /opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/gcc-4.2-20060204/libiberty/floatformat.c -o floatformat.o
/opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/gcc-4.2-20060204/libiberty/floatformat.c: In function 'floatformat_to_double':
/opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/gcc-4.2-20060204/libiberty/floatformat.c:345: warning: floating constant exceeds range of 'float'
/var/tmp//ccZfmQrf.s:480:stfiwx instruction is optional for the PowerPC (not allowed without -force_cpusubtype_ALL option)
/var/tmp//ccZfmQrf.s:535:stfiwx instruction is optional for the PowerPC (not allowed without -force_cpusubtype_ALL option)
make[3]: *** [floatformat.o] Error 1
make[2]: *** [all-stage2-libiberty] Error 2
make[1]: *** [stage2-bubble] Error 2
make: *** [bootstrap] Error 2
Error: Target com.apple.build returned: shell command "cd "/opt/local/var/db/dports/build/_Users_steve_Software_Ports_dports_lang_gcc42/work/build" && make bootstrap -j 2 XCFLAGS=-force_cpusubtype_ALL" returned error 2
Command output: no
make[1]: *** [stage2-bubble] Error 2
make: *** [bootstrap] Error 2

Warning: the following items did not execute (for gcc42): com.apple.build
Error: Status 1 encountered during processing.

This error will occur in several places and stages of the build, depending on which of the CFLAGS variables below are used; the above is just the most recent example (which occurred with only XCFLAGS set of the three). The following spec for darwin7 variant led to a successful build:

# hope springs eternal! uncommented by sjt
# untested on darwin 7
platform darwin 7 {
        depends_build   port:odcctools
        patch {
                reinplace "s|/usr/bin/libtool|${prefix}/bin/odlibtool|g" \
                        ${workpath}/gcc-${version}/gcc/config/darwin.h
        }
        configure.args-append   --with-as=${prefix}/bin/odas \
                                        --with-ld=${prefix}/bin/odld \
                                        --with-ar=${prefix}/bin/odar
        # added by sjt
        build.args-append       XCFLAGS=-force_cpusubtype_ALL
        configure.env           CFLAGS=-force_cpusubtype_ALL \
                                BOOT_CFLAGS="-g -O2 -force_cpusubtype_ALL"
}

The -force_cpusubtype_ALL may not be omitted from any of the three *CFLAGS variables. Of course it's possible (likely?) that there's a better way to achieve this end, but those are the variables I could find that seemed to be more or less appropriate.

[mww@…]

cool, thanks! I've just added your patch to the gcc42 port;

[yaseppochi (Stephen J. Turnbull)]

(In reply to comment #1)

cool, thanks! I've just added your patch to the gcc42 port;

Oops. I'm sorry, but it doesn't actually work from scratch. I'm not sure what I did to get a build, but when I uninstalled and tried again, it died again, in the same way.

I'm in the process of trying again, but it's been 15 hours and I'm still in the middle of libstdc++ (which I guess is probably toward the end but I don't know). I will confirm when I do finish the build, but for now I wanted to let you know that the patch I gave you doesn't actually do the trick.

What I'm currently testing is

# take 2: hope springs eternal! uncommented by sjt
# untested on darwin 7
platform darwin 7 {
        depends_build   port:odcctools
        patch {
                reinplace "s|/usr/bin/libtool|${prefix}/bin/odlibtool|g" \
                        ${workpath}/gcc-${version}/gcc/config/darwin.h
        }
        configure.args-append   --with-as=${prefix}/bin/odas \
                                        --with-ld=${prefix}/bin/odld \
                                        --with-ar=${prefix}/bin/odar
        # added by sjt
        build.args-append       XCFLAGS=-force_cpusubtype_ALL \
                                CFLAGS=-force_cpusubtype_ALL \
                                BOOT_CFLAGS="-g -O2 -force_cpusubtype_ALL"
}

I'm not sure which, if any, of the make variables can be omitted; it will take several days to test at this rate. :-)

[yaseppochi (Stephen J. Turnbull)]

Maybe I just should give up.

The build fails trying to execute "libstdc++-v3/scripts/make_exports.pl" because the script hardcode's "nm", and Mac OS X 10.3.9's nm doesn't understand the -P flag. odnm from odcctools does understand -P, so I tried editing the script to use "/opt/local/bin/nm" instead of "nm" in several places, but now I get:

/opt/local/bin/odnm -P .libs/bitmap_allocator.o .libs/pool_allocator.o .libs/mt_allocator.o .libs/codecvt.o .libs/compatibility.o .libs/complex_io.o .libs/ctype.o .libs/debug.o .libs/debug_list.o .libs/functexcept.o .libs/globals_locale.o .libs/globals_io.o .libs/ios.o .libs/ios_failure.o .libs/ios_init.o .libs/ios_locale.o .libs/limits.o .libs/list.o .libs/locale.o .libs/locale_init.o .libs/locale_facets.o .libs/localename.o .libs/stdexcept.o .libs/strstream.o .libs/tree.o .libs/allocator-inst.o .libs/concept-inst.o .libs/fstream-inst.o .libs/ext-inst.o .libs/ios-inst.o .libs/iostream-inst.o .libs/istream-inst.o .libs/istream.o .libs/locale-inst.o .libs/locale-misc-inst.o .libs/misc-inst.o .libs/ostream-inst.o .libs/sstream-inst.o .libs/streambuf-inst.o .libs/streambuf.o .libs/string-inst.o .libs/valarray-inst.o .libs/wlocale-inst.o .libs/wstring-inst.o .libs/atomicity.o .libs/codecvt_members.o .libs/collate_members.o .libs/ctype_members.o .libs/messages_members.o .libs/monetary_members.o .libs/numeric_members.o .libs/time_members.o .libs/basic_file.o .libs/c++locale.o ../libmath/.libs/libmath.a ../libsupc++/.libs/libsupc++convenience.a|/opt/local/bin/odnm: no name list/opt/local/bin/odnm error at ../../../../gcc-4.2-20060218/libstdc++-v3/scripts/make_exports.pl line 130.

from the Perl script (either indirectly by doing "make" or by directly invoking the Perl script). If I cd to the work directory and execute the command line constructed by the script by hand, I get no errors and a long spew of symbols to stdout. Apparently this is Perl damage, but I don't know enough about Perl to make a guess.

[yaseppochi (Stephen J. Turnbull)]

(In reply to comment #3)

The build fails trying to execute "libstdc++-v3/scripts/make_exports.pl" because the script hardcode's "nm", and Mac OS X 10.3.9's nm doesn't understand the -P flag.

With GCC 4.1, I did the following and successfully built GCC 4.1.

Create a new directory /opt/odbinutils/bin.

Link all of the odcctools $TOOL in /opt/local/bin to /opt/odbinutils/bin/powerpc-apple-darwin7.9.0-$TOOL.

Apply this patch to dports/lang/gcc41/Portfile:

Index: dports/lang/gcc41/Portfile
===================================================================
RCS file: /Volumes/src/cvs/od/proj/darwinports/dports/lang/gcc41/Portfile,v
retrieving revision 1.18
diff -u -r1.18 Portfile
--- dports/lang/gcc41/Portfile  29 May 2006 05:18:49 -0000      1.18
+++ dports/lang/gcc41/Portfile  15 Jun 2006 15:32:15 -0000
@@ -57,6 +57,7 @@
                                --with-gmp=${prefix} \
                                --with-mpfr=${prefix}
 
+build.env      PATH=/opt/local/bin:/opt/odbinutils/bin:/usr/bin:/bin
 build.target   profiledbootstrap
 
 post-destroot {

Unfortunately this procedure doesn't work for GCC 4.2. It finds the desired nm executable, but then fails as described in my previous report. It seems that the "no name list" error is actually a warning and not the fatal error; the problem seems to be that the "close $NM or die;" fails, which the GCC build infrastructure treats as fatal. I hacked the perl script to not die in that case, and the build of libstdc++ continued. However, the build eventually failed (and in the process almost ran me out of disk space, so I've deleted the workspace so I can get work done). Sorry for not providing details, but I can't afford the time or effort to rebuild (it also kills interactive responsiveness) right now. I'll try again later, when I'll also have a patch for the perl script.

[markd@…]

This port has been updated. Please try again.

[yaseppochi (Stephen J. Turnbull)]

I no longer support users of Panther, so I've upgraded to Tiger myself. You can close this as far as I'm concerned.

[markd@…]

Closing.

[ryandesign (Ryan Carsten Schmidt)]

Replying to markd@…:

resolution set to fixed

Nothing was fixed; reopening so I can set the resolution properly.

[(none)]

Milestone Port Enhancements deleted