libgcc11 11.2.0_1 fails to build on Monterey

[jwhowarth]

libgcc11 11.2.0_1 fails to build on Monterey with the error....

:info:build Error: configuring for an unreleased macOS version x86_64-apple-darwin21
:info:build make[2]: *** [configure-stage1-gcc] Error 1
:info:build make[2]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/build'
:info:build make[1]: *** [stage1-bubble] Error 2

[jwhowarth]

patch currently used by home-brew to solve build failures on Monterey

[devernay (Frédéric Devernay)]

port configures fine with attached patch, but then fails with:

/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/build/./gcc/xgcc -B/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/build/./gcc/ -xc++ -nostdinc /dev/null -S -o /dev/null -fself-test=/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/gcc-11.2.0/gcc/testsuite/selftests
xgcc: warning: could not understand version 12.0
xgcc: warning: could not understand version 12.0
<built-in>: error: unknown value '12.0' of '-mmacosx-version-min'
<built-in>: error: unknown value '12.0' of '-mmacosx-version-min'
-fself-test: 64067 pass(es) in 1.185077 seconds
make[3]: *** [s-selftest-c] Error 1

[devernay (Frédéric Devernay)]

Full Monterey patch includes more (from ​https://github.com/Homebrew/homebrew-core/blob/HEAD/Formula/gcc.rb#L55): ​https://github.com/iains/gcc-darwin-arm64/commit/20f61faaed3b335d792e38892d826054d2ac9f15.patch?full_index=1

confirmed working

[Knapoc <Knapoc@…>]

In 7f9e2fb69924a606ed00a7c09d0d07cfd9a0604b/macports-ports (master):

gcc11: add macOS 12 support

Closes: #63677

[imgx64]

I'm still unable to build libgcc11 on macOS 12. But now the error is different.

Note that I'm on M1 Mac Mini and don't have Rosetta installed.

:info:build checking whether we are cross compiling... configure: error: in `/opt/local/var/macports/build/_opt_local_macports-ports_lang_gcc11/libgcc11/work/build/arm64-apple-darwin21/libgomp':
:info:build configure: error: cannot run C compiled programs.
:info:build If you meant to cross compile, use `--host'.
:info:build See `config.log' for more details
:info:build make[2]: *** [configure-stage1-target-libgomp] Error 1
:info:build make[2]: Leaving directory `/opt/local/var/macports/build/_opt_local_macports-ports_lang_gcc11/libgcc11/work/build'
:info:build make[1]: *** [stage1-bubble] Error 2
:info:build make[1]: Leaving directory `/opt/local/var/macports/build/_opt_local_macports-ports_lang_gcc11/libgcc11/work/build'
:info:build make: *** [bootstrap-lean] Error 2
:info:build make: Leaving directory `/opt/local/var/macports/build/_opt_local_macports-ports_lang_gcc11/libgcc11/work/build'
:info:build Command failed:  cd "/opt/local/var/macports/build/_opt_local_macports-ports_lang_gcc11/libgcc11/work/build" && /usr/bin/make -j8 -w bootstrap-lean 
:info:build Exit code: 2
:error:build Failed to build libgcc11: command execution failed
:debug:build Error code: CHILDSTATUS 92846 2
:debug:build Backtrace: command execution failed
:debug:build     while executing
:debug:build "system {*}$notty {*}$callback {*}$nice $fullcmdstring"
:debug:build     invoked from within
:debug:build "command_exec -callback portprogress::target_progress_callback build"
:debug:build     (procedure "portbuild::build_main" line 8)
:debug:build     invoked from within
:debug:build "$procedure $targetname"
:error:build See /opt/local/var/macports/logs/_opt_local_macports-ports_lang_gcc11/libgcc11/main.log for details.

[abey79 (Antoine Beyeler)]

I have the exact same error as imgx64 on a new M1 Max MBP under Monterey.

[ednl (Ewoud Dronkert)]

I only see libgcc11 11.1.0_2, is that correct? Is 11.2.0_1 a development version which is not (yet) seeded to the general repos, or is it a typo in the version number? In any case, same build problem for me on Monterey M1 Mac Mini (with Rosetta).

[mpbarlow (Matt Barlow)]

Copying my comment from #63698

---

I have found a workaround. The error in config.log is that it cannot find /usr/lib/libgcc_s.1.1.dylib or /usr/local/lib/libgcc_s.1.1.dylib. However, this file is present in the work directory for the port, so after the install failed I ran

sudo cp /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/build/arm64-apple-darwin21/libgcc/libgcc_s.1.1.dylib /usr/local/lib

After that the port built without issue—hopefully there are no further implications from doing this.

[abey79 (Antoine Beyeler)]

Matt's workaround worked for me.

[astroboylrx (Rixin Li)]

It seems there was an fix mentioned in #63691 and port selfupdate leads to different error during make:

Comparing stages 2 and 3
warning: gcc/cc1obj-checksum.o differs
warning: gcc/cc1objplus-checksum.o differs
Bootstrap comparison failure!
gcc/tree-ssa-operands.o differs
gcc/tree-ssanames.o differs
gcc/ipa-inline.o differs
gcc/tree-ssa-pre.o differs
...

Any suggestions other than putting library outside /opt/local to /usr/local? Thanks in advance.

[cjones051073 (Chris Jones)]

See ​https://github.com/macports/macports-ports/commit/a32f923fd3e337be4eae3c8dce70e49c2de1c1ed

[astroboylrx (Rixin Li)]

Replying to cjones051073:

See ​https://github.com/macports/macports-ports/commit/a32f923fd3e337be4eae3c8dce70e49c2de1c1ed

Sorry, doesn't seem to work. Same error after this.

[mascguy (Christopher Nielsen)]

This port is also failing to build on our Monterey buildbot, blocking builds of numerous dependent ports.

The latest attempt, from today:

​https://build.macports.org/builders/ports-12_arm64-builder/builds/1059

[Tatsh (Andrew Udvare)]

Seems this only affects ARM64 so the title should be updated.

[astroboylrx (Rixin Li)]

Replying to Tatsh:

Seems this only affects ARM64 so the title should be updated.

It also fails on my Intel Mac. I don't think it only affects ARM64.

[cjones051073 (Chris Jones)]

If it is also affecting Intel builds please can someone running this OS post a full log, just so we can see iff it really is the same error or not. Note the ARM and Intel builds use different versions of gcc 11, from different sources, so could be entirely different issues.

[astroboylrx (Rixin Li)]

Replying to cjones051073:

If it is also affecting Intel builds please can someone running this OS post a full log, just so we can see iff it really is the same error or not. Note the ARM and Intel builds use different versions of gcc 11, from different sources, so could be entirely different issues.

This issue was not meant to be dedicated to ARM64 (#63698 was, but was then closed due to duplicate to this issue). The full log may not help because I had to dive into the log and execute the command that failed and then found what happened was Bootstrap comparison failure! (objects differ between the stage 2 and 3 of building). Anyway, I uploaded my main.log just in case it helps.

EDIT: sorry, this error is indeed in the main log. I might confuse it with some previous bugs.

[astroboylrx (Rixin Li)]

Full log of a failed build on an Intel Mac

[cjones051073 (Chris Jones)]

See ​https://github.com/macports/macports-ports/commit/573b82fd5df538ea2ad75e2c2373f4727ef3430c

I incorrectly read the home-brew recipe, they in fact apply the patch from the arm branch to both intel and arm builds.

Tested on intel only (no access to an arm machine) and libgcc11 built fine.

[Sebastian-Maier (Sebastian Maier)]

Replying to cjones051073:

See ​https://github.com/macports/macports-ports/commit/573b82fd5df538ea2ad75e2c2373f4727ef3430c

I incorrectly read the home-brew recipe, they in fact apply the patch from the arm branch to both intel and arm builds.

Tested on intel only (no access to an arm machine) and libgcc11 built fine.

Also works for me on ARM64.
Both libgcc11 and gcc11 built fine now on macOS Monterey (macOS 12.0.1, Darwin 21.1.0, MacPorts 2.7.1).

[saagarjha (Saagar Jha)]

Replying to cjones051073:

See ​https://github.com/macports/macports-ports/commit/573b82fd5df538ea2ad75e2c2373f4727ef3430c

I incorrectly read the home-brew recipe, they in fact apply the patch from the arm branch to both intel and arm builds.

Tested on intel only (no access to an arm machine) and libgcc11 built fine.

You also probably want to take this patch, otherwise the generated binaries will fail to run: ​https://github.com/iains/gcc-darwin-arm64/pull/46#issuecomment-864925727

[cjones051073 (Chris Jones)]

Binaries run fine for me, on Intel. Used to build a number of ports that require fortran...

I'll take a look though.

[cjones051073 (Chris Jones)]

I suspect that patch is not required, and the one now used supersedes it.

[hakante (HåkanT)]

I'm still having the issue on M1 silicon after syncing and cleaning when building libgcc11, with the error message matching #63698.

[hakante (HåkanT)]

/libgcc11/main.log

[hakante (HåkanT)]

/libgcc11/work/build/arm64-apple-darwin21/libgomp/config.log

[cjones051073 (Chris Jones)]

I cannot help with arm machines as I do not have access to any. Others though seem to claim it works for them, so try comparing your logs to theirs...

[astroboylrx (Rixin Li)]

Replying to cjones051073:

See ​https://github.com/macports/macports-ports/commit/573b82fd5df538ea2ad75e2c2373f4727ef3430c

I incorrectly read the home-brew recipe, they in fact apply the patch from the arm branch to both intel and arm builds.

Tested on intel only (no access to an arm machine) and libgcc11 built fine.

The new tree works for Intel Mac. Thank you all very much for patching this port.

Replying to hakante:

I'm still having the issue on M1 silicon after syncing and cleaning when building libgcc11, with the error message matching #63698.

I can confirm that it still fails to build on ARM Mac and shows the same error as shown in #63698.

[Sebastian-Maier (Sebastian Maier)]

I have no idea why it worked for me on ARM64 but not others.

Maybe because I was pretty aggressive when deleting leftover files from previous build attempts as I used a variation of (1):

sudo rm -rf /opt/local/var/macports/build/*
sudo rm -rf /opt/local/var/macports/distfiles/*
sudo rm -rf /opt/local/var/macports/packages/*

Apart from that, libgcc11 is actually installed in version 11.1.0_2 on ARM64 in accordance with the latest Portfile (and not in version 11.2.0_1 which is used on Intel Macs):

$ port installed libgcc11
The following ports are currently installed:
  libgcc11 @11.1.0_2 (active)

$ uname --processor 
arm64

(1) ​https://superuser.com/a/165670

[manuelmenzella (Manuel Menzella)]

I can't get this to work for arm64, unfortunately. I'm getting the same error as above.

:info:build checking whether we are cross compiling... configure: error: in `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/build/arm64-apple-darwin21/libgomp':
:info:build configure: error: cannot run C compiled programs.
:info:build If you meant to cross compile, use `--host'.
:info:build See `config.log' for more details
:info:build make[2]: *** [configure-stage1-target-libgomp] Error 1
:info:build make[2]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/build'
:info:build make[1]: *** [stage1-bubble] Error 2
:info:build make[1]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/build'
:info:build make: *** [bootstrap-lean] Error 2
:info:build make: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/build'
:info:build Command failed:  cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/work/build" && /usr/bin/make -j8 -w bootstrap-lean
:info:build Exit code: 2
:error:build Failed to build libgcc11: command execution failed
:debug:build Error code: CHILDSTATUS 98007 2
:debug:build Backtrace: command execution failed
:debug:build     while executing
:debug:build "system {*}$notty {*}$callback {*}$nice $fullcmdstring"
:debug:build     invoked from within
:debug:build "command_exec -callback portprogress::target_progress_callback build"
:debug:build     (procedure "portbuild::build_main" line 8)
:debug:build     invoked from within
:debug:build "$procedure $targetname"
:error:build See /opt/local/var/macports/logs/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_lang_gcc11/libgcc11/main.log for details.

Is there any information I can provide to help debug this?

[astroboylrx (Rixin Li)]

Replying to Sebastian-Maier:

I have no idea why it worked for me on ARM64 but not others.

Maybe because I was pretty aggressive when deleting leftover files from previous build attempts as I used a variation of (1):

sudo rm -rf /opt/local/var/macports/build/*
sudo rm -rf /opt/local/var/macports/distfiles/*
sudo rm -rf /opt/local/var/macports/packages/*

This doesn't work for me. The error log complains about cannot run C compiled programs. and says If you meant to cross compile, use '--host'.

I think it may worth trying to add --host=arm or --host=arm64 to the configure step. But I'm not familiar with Portfile. Is there a quick way to do that?

[imgx64]

Replying to Sebastian-Maier:

I have no idea why it worked for me on ARM64 but not others.

Do you have the files /usr/lib/libgcc_s.1.1.dylib or /usr/local/lib/libgcc_s.1.1.dylib as per comment:10? If yes, where did they come from?

[Sebastian-Maier (Sebastian Maier)]

/usr/local/lib/libgcc_s.1.1.dylib does exist on my MBP (while /usr/lib/libgcc_s.1.1.dylib does not).

According to the creation date, it looks like a leftover from Matt's workaround that I tried several days earlier (seemingly without any success).

The build does not work for me either when I delete this file.

Sorry for the confusion.

[astroboylrx (Rixin Li)]

Replying to astroboylrx:

The error log complains about cannot run C compiled programs. and says If you meant to cross compile, use '--host'.

To add some extra info, trying to install gcc with HomeBrew on macOS 12.0.1 (M1) also runs into the same issue/error.

[JohnPritchard (Johnny P)]

I have successfully built libgcc11 (and gcc11) on macOS Monterey arm64 by applying the workaround patch gcc11_patch_macOS12_arm64 (that I have attached to this ticket) which I think addresses the symptoms, rather than curing the problem.

The issues seem to revolve around libgcc_s.1.1.dylib.

The first issue is that when "checking whether we are cross compiling" it is checks for libgcc_s.1.1.dylib only in /usr/lib and /usr/local/lib, so the first part of the patch is to create (and remove) a softlink /usr/local/lib/libgcc_s.1.1.dylib so that it will be found when needed.

The 2nd issue is that once created, libgcc_s.1.1.dylib doesn't have executable permissions, like all the other dylibs created, so the 2nd part of the patch sets the permissions (on all the dylibs, just in case...)

I don't understand why these problems present only for arm64 and not for x86_64 (probably because they are using different versions), I leave that for others to resolve.

[MarcusCalhoun-Lopez (Marcus Calhoun-Lopez)]

I am not sure the ​patch to fix this problem is actually applied. I have created a ​pull request that attempts to prevent the patchfiles-append patch-darwin21-support.diff command from being overridden later.

[MarcusCalhoun-Lopez (Marcus Calhoun-Lopez)]

In 3d0dde8ac41070f5782bf70d082c30d5f3632391/macports-ports (master):

gcc11: ensure all patches are applied

Patch file from
7f9e2fb69924a606ed00a7c09d0d07cfd9a0604b/macports-ports
is lost when patchfile command is later used.

Fixes #63677
No revbump since port either builds correctly or not at all.