libgcc: error: unknown value '10.10' of -mmacosx-version-min

[AP1010]

Ateempt to build libgcc on Yosemite fails with the following error:

configure: error: cannot compute suffix of object files: cannot compile

In the config.log (attached) the following message:

conftest.c:1:0: error: unknown value '10.10' of -mmacosx-version-min

suggests that the Yosemite OS X version (10.10) may not be handled correctly

[ryandesign (Ryan Carsten Schmidt)]

Jeremy had pointed this out by private mail on the 4th:

From: Jeremy Huddleston Sequoia
Subject: gcc support for larger deployment targets
Date: June 4, 2014 at 12:01:42 AM CDT
To: Markus Weissmann, David Evans, Ryan Schmidt

Could one of you please update the gcc ports to support deployment targets >= 10.10?

I suggest that you just make this change in gcc47, gcc48, and gcc49 and mark gcc46 and earlier as not supported on Yosemite, since upstream no longer supports gcc-4.6.

$ gcc-mp-4.8 -mmacosx-version-min=10.12.13 -dM -E -o - - < /dev/null | grep MAC_OS_X
<built-in>:0:0: error: unknown value '10.12.13' of -mmacosx-version-min
#define __ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ 1000

$ clang-mp-3.4 -mmacosx-version-min=10.12.13 -dM -E -o - - < /dev/null | grep MAC_OS_X
#define __ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ 101213

See r120636, dports/lang/llvm-3.4/files/yosemite-deployment-target.patch​ as reference.

Thanks,
Jeremy

[skymoo (Adam Mercer)]

Cc Me!

[pturing@…]

Cc Me!

[platipodium (Carsten Lemmen)]

Cc Me!

[borodiychuk (Andriy Borodiychuk)]

Cc Me!

[borodiychuk (Andriy Borodiychuk)]

Cc Me!

[borodiychuk (Andriy Borodiychuk)]

As far as I can understand, this file (at least) has to be patched:

gcc-4.8.2/gcc/config/darwin-c.c, around lines 573-597:

/* Return the value of darwin_macosx_version_min suitable for the
   __ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ macro,
   so '10.4.2' becomes 1040.  The lowest digit is always zero.
   Print a warning if the version number can't be understood.  */
static const char *
version_as_macro (void)
{
  static char result[] = "1000";

  if (strncmp (darwin_macosx_version_min, "10.", 3) != 0)
    goto fail;
  if (! ISDIGIT (darwin_macosx_version_min[3]))
    goto fail;
  result[2] = darwin_macosx_version_min[3];
  if (darwin_macosx_version_min[4] != '\0'    //  <--- This is the problem. The code does not expect minor version to be 2-chars, rather only one-char.
      && darwin_macosx_version_min[4] != '.') //  <---  So for 10.X it works, bit for 10.XX it gives an error
    goto fail;

  return result;

 fail:
  error ("unknown value %qs of -mmacosx-version-min",
	 darwin_macosx_version_min);
  return "1000";
}

[jeremyhu (Jeremy Huddleston Sequoia)]

I'll put together a patch for apple-gcc42, and you guys can take care of merging it into the GPL3 ports.

[jeremyhu (Jeremy Huddleston Sequoia)]

See r122725

[larryv (Lawrence Velázquez)]

Cc Me!

[ryandesign (Ryan Carsten Schmidt)]

Has duplicate #44510.

I am working on this as part of updating the gcc ports (#43789, #43502, #43529).

[davidchambers (David Chambers)]

Cc Me!

[ryandesign (Ryan Carsten Schmidt)]

gcc48, gcc49 and gcc410 have a fix for this as of r123168. I'll work on earlier gcc versions next.

[borodiychuk (Andriy Borodiychuk)]

Yosemite, 10.10, libgcc 4.9.1

[borodiychuk (Andriy Borodiychuk)]

Replying to ryandesign@…:

gcc48, gcc49 and gcc410 have a fix for this as of r123168. I'll work on earlier gcc versions next.

Tried to build it again Yosemite, after upgrading ports. Now it tries to build libgcc 4.9.1, but also fails with an error. I have attached a logfile in the previous comment.

[ryandesign (Ryan Carsten Schmidt)]

All that I'm planning on dealing with in this ticket is the handling of the 10.10 version number.

[ryandesign (Ryan Carsten Schmidt)]

Replying to ab@…:

Tried to build it again Yosemite, after upgrading ports. Now it tries to build libgcc 4.9.1, but also fails with an error. I have attached a logfile in the previous comment.

This new problem is now filed at #44596.

[eirnym (Eir Nym)]

Cc Me!

[tonior@…]

Cc Me!

[dave@…]

Cc Me!

[larryv (Lawrence Velázquez)]

Here’s the ​official upstream fix on the GCC 4.9 branch.

[ryandesign (Ryan Carsten Schmidt)]

Has duplicate #45485.

[macports_org@…]

Am finding I cannot install octave because the same issue is occurring for gcc47.

[arthurprat@…]

Hello, I needed gcc47 so I manually applied the patch from r123168 (which was designed for gcc48, gcc49 and gcc410) to the gcc47 files. It seems to work, but is there anything I should worry about? A reason this patch was not applied to gcc47 too in the first place?

[arthurprat@…]

Cc Me!

[jeremyhu (Jeremy Huddleston Sequoia)]

AFAIK, upstream is not supporting gcc 4.7 and older any more. What do you need it for?

[ryandesign (Ryan Carsten Schmidt)]

Replying to arthurprat@…:

A reason this patch was not applied to gcc47 too in the first place?

Because I haven't gotten around to it.

[jbgoette (Jörg Götte)]

Replying to macports_org@…:

Am finding I cannot install octave because the same issue is occurring for gcc47.

The default variant of octave is +atlas which in turn has +gcc47 as default, but can be installed with +gcc48 or +gcc49. This should remove the necessity for gcc47 in the octave installation.

[jbgoette (Jörg Götte)]

Cc Me!

[mf2k (Frank Schima)]

Replying to jbgoette@…:

The default variant of octave is +atlas which in turn has +gcc47 as default, but can be installed with +gcc48 or +gcc49. This should remove the necessity for gcc47 in the octave installation.

While that is unfortunate and probably should be changed in the atlas port, there is nothing "necessary" about using gcc47 and installing octave. Simple manually install atlas with +gcc48 and then you no longer need it for installing octave.

[tristancollins (Tristan Collins)]

Cc Me!

[larryv (Lawrence Velázquez)]

Has duplicate #45549.

[enxbw3@…]

Replying to larryv@…:

Has duplicate #45549.

Hi, I am totally a newbee to this stuff. I need to use gcc46, which seems unavailable now..... Can I use gcc47 or 48 instead? Or I should do something else. Many thanks.

[ryandesign (Ryan Carsten Schmidt)]

gcc48, gcc49 and gcc5 have already been fixed to work on Yosemite. gcc47 and earlier have not yet.

[enxbw3@…]

Replying to ryandesign@…:

gcc48, gcc49 and gcc5 have already been fixed to work on Yosemite. gcc47 and earlier have not yet.

So that means I have to wait for the fixing of gcc46 as far as I can understand. Anyway, thanks for all your work. : )

[ryandesign (Ryan Carsten Schmidt)]

Why can't use use gcc48 or newer?

[enxbw3@…]

Replying to ryandesign@…:

Why can't use use gcc48 or newer?

Of course I can if that can replace gcc46! Sorry for the misunderstanding. I thought it can't.

[enxbw3@…]

Replying to ryandesign@…:

Why can't use use gcc48 or newer?

Sorry sir, but while I am try to install my software with gcc48, following error showed up:

/bin/sh: g++-mp-4.6: command not found

Is this because I used gcc48 instead of gcc46?

[ryandesign (Ryan Carsten Schmidt)]

Replying to enxbw3@…:

/bin/sh: g++-mp-4.6: command not found

If you are manually runing g++-mp-4.6, then run g++-mp-4.8 instead. If this is happening because of some other script you're running, figure out where in that script it's referencing g++-mp-4.6 and change it to g++-mp-4.8.

[rubendibattista (Ruben Di Battista)]

Cc me please!

[tonior@…]

Cc Me!

[larryv (Lawrence Velázquez)]

Replying to rubendibattista@…:

Cc me please!

You need to click the “Cc” button next to the Cc list above to actually be Cc’d. I’ve added you.

[rubendibattista (Ruben Di Battista)]

Thanks! Sorry I'm new here...

Anyway I'm trying to compile OpenFOAM with Gcc48 and I'm getting segmentation fault: 11 when running the executables. Can this be connected to this issue?

[ryandesign (Ryan Carsten Schmidt)]

Doesn't sound like it. Please file a new ticket or ask on the mailing list.

[arthurprat@…]

Replying to jeremyhu@…:

AFAIK, upstream is not supporting gcc 4.7 and older any more. What do you need it for?

I may have been a bit quick actually, I thought I needed it because of some things that would not compile with 4.8 but actually everything works fine, thanks

[arthurprat@…]

Replying to rubendibattista@…:

Thanks! Sorry I'm new here...

Anyway I'm trying to compile OpenFOAM with Gcc48 and I'm getting segmentation fault: 11 when running the executables. Can this be connected to this issue?

Make sure you got the last version of the gcc48 port : 4.8.3_1. That _1 is specifically there to get rid of segmentation faults on Yosemite

[ryandesign (Ryan Carsten Schmidt)]

Has duplicate #45603.

[eirnym (Eir Nym)]

Good news.

Upstream fixes: ​https://gcc.gnu.org/viewcvs/gcc?view=revision&revision=215251

Homebrew has been closed issue: ​https://github.com/Homebrew/homebrew-versions/issues/450

[larryv (Lawrence Velázquez)]

Replying to eirnym@…:

Upstream fixes

We’re aware. It’s been fixed on upstream trunk and the 4.9 branch for over a month.

Homebrew has been closed issue

Homebrew only provides GCC 4.9. Our gcc48, gcc49, and gcc5 ports have been fixed for a while.

[larryv (Lawrence Velázquez)]

gcc47 patched in r127365.

[anddam (Andrea D'Amore)]

Cc Me!

[larryv (Lawrence Velázquez)]

Has duplicate #45632.

[eirnym (Eir Nym)]

Homebrew only provides GCC 4.9. Our gcc48, gcc49, and gcc5 ports have been fixed for a while.

I haven't seen messages where told they've been fixed.

I need any version of fortran compiler, not GCC itself. I need any available compiler I can use for scipy. I really need it.

[ryandesign (Ryan Carsten Schmidt)]

Replying to eirnym@…:

I haven't seen messages where told they've been fixed.

See comment:13 above.

[larryv (Lawrence Velázquez)]

Have you actually tried to install any of them? They compile and install fine. Run sudo port selfupdate first to make sure you have the latest ports.

[larryv (Lawrence Velázquez)]

gcc46 fixed in r127393.

[eirnym (Eir Nym)]

Replying to ryandesign@…:

See comment:13 above.

comment:14 has stopped me think this is final fix.

Replying to larryv@…:

Have you actually tried to install any of them?

Yes, before comment:13.

I need alternate Fortran compiler. I don't want install gcc with bugs every year. Do you know good compiler to port (or to use)?

[ryandesign (Ryan Carsten Schmidt)]

Replying to eirnym@…:

Replying to ryandesign@…:

See comment:13 above.

comment:14 has stopped me think this is final fix.

That was a separate problem, which was separately fixed. See comment:16.

I need alternate Fortran compiler. I don't want install gcc with bugs every year. Do you know good compiler to port (or to use)?

Try the gcc49 port. It is the latest stable version of gcc.

[eirnym (Eir Nym)]

Replying to ryandesign@…:

That was a separate problem, which was separately fixed. See comment:16.

I see

Replying to ryandesign@…:

I need alternate Fortran compiler. I don't want install gcc with bugs every year. Do you know good compiler to port (or to use)?

Try the gcc49 port. It is the latest stable version of gcc.

I've tied to have delay in developing for 3-5 months every year. I need alternate fortran compiler codebase.

[ryandesign (Ryan Carsten Schmidt)]

This ticket is about fixing a specific problem with the gcc ports on Yosemite, so let's keep the comments relevant to that. For suggestions of other fortran compilers, you could ask on the macports-users mailing list.

[jeremyhu (Jeremy Huddleston Sequoia)]

I think I may have accidentally seen the source changes for this change when I looked at Jack's patch attached to #45449

If the changes being made to the gccXY ports are the same as the patch Jack attached to #45449, then this is an incorrect change.

Please look at the change I referenced in this ticket three months ago (r122725) as the correct solution and inform upstream if they are doing it wrong.

[elfprince13 (Thomas Dickerson)]

Replying to larryv@…:

gcc46 fixed in r127393.

Any idea if that will work with gcc45, or how much work might be needed to backport the fix to gcc45?

[jeremyhu (Jeremy Huddleston Sequoia)]

Why do you want such a dated compiler? IMO, we should mark gcc 4.6 and earlier as unavailable on OS X 10.10 like I've been doing for dated llvm.

[larryv (Lawrence Velázquez)]

Replying to elfprince13@…:

Any idea if that will work with gcc45, or how much work might be needed to backport the fix to gcc45?

I haven’t looked at gcc45 yet. Even after applying this fix, there’s no guarantee that it’ll function correctly on Yosemite without further work.

[larryv (Lawrence Velázquez)]

Replying to jeremyhu@…:

Why do you want such a dated compiler? IMO, we should mark gcc 4.6 and earlier as unavailable on OS X 10.10 like I've been doing for dated llvm.

I’m inclined to agree.

[lzkelley (Luke)]

Cc Me!

[lzkelley (Luke)]

Is there any workaround for this? almost everything needs libgcc.

[larryv (Lawrence Velázquez)]

Not sure what you mean.

1. This has already been fixed for libgcc and gcc46 through gcc5. Are you still having problems installing any of those ports? Have you run sudo port selfupdate lately to retrieve the latest ports?
2. Nearly nothing requires libgcc, except a few ports that use GCC for specific things.

[lzkelley (Luke)]

Replying to larryv@…:

Not sure what you mean.

1. This has already been fixed for libgcc and gcc46 through gcc5. Are you still having problems installing any of those ports? Have you run sudo port selfupdate lately to retrieve the latest ports?
2. Nearly nothing requires libgcc, except a few ports that use GCC for specific things.

I just reinstalled macports yesterday (after installing 10.10), so it should all be up to date. I can't install gcc48 (e.g.) because it fails on libgcc dependency. There are a bunch of python ports which seem to depend on it, and a couple of other science-specific ports.

[lzkelley (Luke)]

Replying to lkelley@…:

Replying to larryv@…:

Not sure what you mean.

1. This has already been fixed for libgcc and gcc46 through gcc5. Are you still having problems installing any of those ports? Have you run sudo port selfupdate lately to retrieve the latest ports?
2. Nearly nothing requires libgcc, except a few ports that use GCC for specific things.

I just reinstalled macports yesterday (after installing 10.10), so it should all be up to date. I can't install gcc48 (e.g.) because it fails on libgcc dependency. There are a bunch of python ports which seem to depend on it, and a couple of other science-specific ports.

Yeah... it still fails, same error as #44510 which was removed as duplicate of this ticket.

[lzkelley (Luke)]

#44510 like error

[larryv (Lawrence Velázquez)]

Doesn’t look the same to me. The error in #44510 was the same as in this ticket (“cannot compute suffix of object files”, etc.). This is the error you’re seeing:

:info:build /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/gcc-4.9.1/gcc/fortran/module.c:113:1: error: conflicting declaration 'typedef enum pointer_t pointer_t'
:info:build  pointer_t;
:info:build  ^
:info:build In file included from /usr/include/mach/memory_object_types.h:77:0,
:info:build                  from /usr/include/mach/mach_types.h:85,
:info:build                  from /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/prev-gcc/include-fixed/malloc.h:37,
:info:build                  from /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/gcc-4.9.1/gcc/system.h:466,
:info:build                  from /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/gcc-4.9.1/gcc/fortran/module.c:68:
:info:build /usr/include/mach/vm_types.h:40:26: note: previous declaration as 'typedef vm_offset_t pointer_t'
:info:build  typedef vm_offset_t      pointer_t;
:info:build                           ^
:info:build make[3]: *** [fortran/module.o] Error 1

Could you open a new ticket for this?

[lzkelley (Luke)]

Replying to larryv@…:

Doesn’t look the same to me. The error in #44510 was the same as in this ticket (“cannot compute suffix of object files”, etc.). This is the error you’re seeing:

:info:build /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/gcc-4.9.1/gcc/fortran/module.c:113:1: error: conflicting declaration 'typedef enum pointer_t pointer_t'
:info:build  pointer_t;
:info:build  ^
:info:build In file included from /usr/include/mach/memory_object_types.h:77:0,
:info:build                  from /usr/include/mach/mach_types.h:85,
:info:build                  from /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/build/prev-gcc/include-fixed/malloc.h:37,
:info:build                  from /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/gcc-4.9.1/gcc/system.h:466,
:info:build                  from /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_lang_gcc49/libgcc/work/gcc-4.9.1/gcc/fortran/module.c:68:
:info:build /usr/include/mach/vm_types.h:40:26: note: previous declaration as 'typedef vm_offset_t pointer_t'
:info:build  typedef vm_offset_t      pointer_t;
:info:build                           ^
:info:build make[3]: *** [fortran/module.o] Error 1

Could you open a new ticket for this?

Opened ticket #45758, sorry for the missplacement!

[larryv (Lawrence Velázquez)]

gcc45 fixed in r128469:128470.

[larryv (Lawrence Velázquez)]

gcc44 fixed in r128477:128478.

[larryv (Lawrence Velázquez)]

gcc43 fixed in r128487.

gcc42 apparently does not build on anything newer than Leopard, so I’m done here. See #45449 for apple-gcc42 and llvm-gcc42, which I have to handle differently due to copyright/licensing concerns.

[ryandesign (Ryan Carsten Schmidt)]

Has duplicate #45561.