ld64-127 causes build failures with Leopard/ppc (G4)

[zzanderr]

The default choice for OS X 10.5.8, at least on 32-bit G4's, is ld64 +ld64_127. Out of the box, libgcc @4.9.2_1 will not build using this variant.

You might try adjusting the defaults for various particular platforms and OS versions (for all I know, ld64 +ld64_127 may work on a 64-bit G5).

While posting to an open ticket for libgcc, I noticed other people are having the same problem: #46557 and #45174. These failures might also be confounded with a recurring libgcc bootstrap regression issue (see ​https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63703).

NB: I am in the middle of rebuilding MacPorts from scratch after the registry became corrupted, prior to the failure of my PowerBook's DC-power logic board, for which I waited two weeks for a replacement. After the libgcc @4.9.2_1 build failed, I tried compiling libgcc @4.9.1_1, which worked back in November--again, failure. Then, following a vague suspicion that the root lay in the new division of ld64, I installed ld64 +ld64_97 and tried a third time--finally success. The failed attempts both boinked near the end of the build--over 6 hours on a PowerBook G4 @1.67 GHz. So I enjoyed an entire day of head-scratching and teeth-gnashing, all with little hope of success.

(Sorry, I needed to vent.) I realize you are a volunteer, and I sincerely thank you for your effort.

ld64_97's log unfortunately clobbered ld64-127's log, which consisted mostly of warnings, even though ld64-127 installed without so much as a burp. So below are logs of libgcc's failed and successful builds. I will also post them here: #46557.

[jeremyhu (Jeremy Huddleston Sequoia)]

Thanks for the detailed report. It is very much expected that ld64-127 should work with Leopard/ppc. Indeed, we're trying to also get it to work with Tiger, so we can eventually drop the ancient ld64-97 port.

[jeremyhu (Jeremy Huddleston Sequoia)]

:info:build ld: 32-bit absolute address out of range (0x140D68FBC max is 4GB): from vtable for (anonymous namespace)::pass_branch_target_load_optimize2 + 0x0000001C 

[jeremyhu (Jeremy Huddleston Sequoia)]

Would you be able to figure out the exact invocation of the failing command (run the failing gcc command with -v -Wl,-v to get verbose details) and send me a tarball containing all the the object files and libraries involved, so I can try reproducing this?

[zzanderr]

Replying to jeremyhu@…:

Would you be able to figure out the exact invocation of the failing command (run the failing gcc command with -v -Wl,-v to get verbose details) and send me a tarball containing all the the object files and libraries involved, so I can try reproducing this?

Could you clarify this? I'm not sure I understand.

[jeremyhu (Jeremy Huddleston Sequoia)]

This is the failing command:

/mp2/local/var/macports/build/_mp2_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/./prev-gcc/xg++ -B/mp2/local/var/macports/build/_mp2_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/./prev-gcc/ -B/mp2/local/ppc-apple-darwin9/bin/ -nostdinc++ -B/mp2/local/var/macports/build/_mp2_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/prev-ppc-apple-darwin9/libstdc++-v3/src/.libs -B/mp2/local/var/macports/build/_mp2_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/prev-ppc-apple-darwin9/libstdc++-v3/libsupc++/.libs  -I/mp2/local/var/macports/build/_mp2_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/prev-ppc-apple-darwin9/libstdc++-v3/include/ppc-apple-darwin9  -I/mp2/local/var/macports/build/_mp2_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/prev-ppc-apple-darwin9/libstdc++-v3/include  -I/mp2/local/var/macports/build/_mp2_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/gcc-4.9.2/libstdc++-v3/libsupc++ -L/mp2/local/var/macports/build/_mp2_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/prev-ppc-apple-darwin9/libstdc++-v3/src/.libs -L/mp2/local/var/macports/build/_mp2_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/prev-ppc-apple-darwin9/libstdc++-v3/libsupc++/.libs   -g -O2 -mdynamic-no-pic -gtoggle -DIN_GCC    -fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -pedantic -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings   -DHAVE_CONFIG_H -static-libstdc++ -static-libgcc  -o cc1 c/c-lang.o c-family/stub-objc.o attribs.o c/c-errors.o c/c-decl.o c/c-typeck.o c/c-convert.o c/c-aux-info.o c/c-objc-common.o c/c-parser.o c/c-array-notation.o c-family/c-common.o c-family/c-cppbuiltin.o c-family/c-dump.o c-family/c-format.o c-family/c-gimplify.o c-family/c-lex.o c-family/c-omp.o c-family/c-opts.o c-family/c-pch.o c-family/c-ppoutput.o c-family/c-pragma.o c-family/c-pretty-print.o c-family/c-semantics.o c-family/c-ada-spec.o c-family/c-cilkplus.o c-family/array-notation-common.o c-family/cilk.o c-family/c-ubsan.o darwin-c.o rs6000-c.o cc1-checksum.o libbackend.a main.o  libcommon-target.a libcommon.a ../libcpp/libcpp.a ../libdecnumber/libdecnumber.a libcommon.a ../libcpp/libcpp.a  -liconv ../libbacktrace/.libs/libbacktrace.a ../libiberty/libiberty.a ../libdecnumber/libdecnumber.a  -L/mp2/local/lib -lcloog-isl -L/mp2/local/lib -lisl -L/mp2/local/lib -L/mp2/local/lib -L/mp2/local/lib -lmpc -lmpfr -lgmp   -lz

Run that at the appropriate location with the additional args: '-v -Wl,-v' that will show more verbose information, including the command line to ld.

Send me all of the relvant files (all the .o and .a files as well as libisl, libcloog-isl, libmpc, libmpfr, libgmp, and libz) and the command line to ld, so I can reproduce the problem.

[jeremyhu (Jeremy Huddleston Sequoia)]

I suspect a subtraction is being done, similar to the comment here:

​http://bugs.freepascal.org/view.php?id=20879

[jeremyhu (Jeremy Huddleston Sequoia)]

Yeah, I see the problem. I'll get a fix in soon.

[zzanderr]

Replying to jeremyhu@…:

I suspect a subtraction is being done, similar to the comment here:

​http://bugs.freepascal.org/view.php?id=20879

Just deactivated libgcc, ld64 and ld64-97, then reinstalled ld64-127 to get a log.

I can try to rebuild libgcc with ld-127--unless you have solved it already...

[jeremyhu (Jeremy Huddleston Sequoia)]

r132725

[zzanderr]

Great! I'll try it out tomorrow night. Thanks for the quick work.

[jeremyhu (Jeremy Huddleston Sequoia)]

Thanks for testing. If you run into the issue again (or something similar), grab the relevant files, so I can reproduce it.

[mojca (Mojca Miklavec)]

Cc Me!

[barracuda156]

Quite surprisingly, I have got an error with gcc48 now towards the very end of the build: turned out, ld64-97 +llvm5 was responsible. Activated +llvm33, and the build proceeded through completion. I have had a similar case when ld64-127 failed building some other gcc version.

(This is just for the record, no fixes needed.)