Error building package python3-3.6.1.p1 Sage 8.1 on Mac

[Christelle Vincent]

Hi,

I am trying to build Sage 8.1 from source on Mac (High Sierra 10.13.4, just updated Xcode and re-installed Command Line Tools).

I get the error "Error building package python3-3.6.1.p1"

I am attaching the whole log but it ends with

Undefined symbols for architecture x86_64:

  "_libintl_bindtextdomain", referenced from:

      _PyIntl_bindtextdomain in _localemodule.o

  "_libintl_dcgettext", referenced from:

      _PyIntl_dcgettext in _localemodule.o

  "_libintl_dgettext", referenced from:

      _PyIntl_dgettext in _localemodule.o

  "_libintl_gettext", referenced from:

      _PyIntl_gettext in _localemodule.o

  "_libintl_textdomain", referenced from:

      _PyIntl_textdomain in _localemodule.o

ld: symbol(s) not found for architecture x86_64

clang: error: linker command failed with exit code 1 (use -v to see invocation)

make[3]: *** [Programs/_freeze_importlib] Error 1

Error building Python.

Anyone can help me fix this? Thanks!

Christelle

[John H Palmieri]

In my experience, it can help to run Xcode once after upgrading so that it can install some "additional components". Did you re-install command line tools by upgrading through the app store, or some other way?

[Christelle Vincent]

I upgraded through the Apple Developer website. I also did open Xcode again after the installation, but I can't remember if any additional components were installed.

[Dima Pasechnik]

I guess you have Homebrew/MacPorts/Anaconda installed, and a wrong library is picked up from one of these.

[Christelle Vincent]

I do have Homebrew installed! Is there any way to stop it from seeing the wrong library? Can I uninstall Homebrew? I originally installed it at the suggestion of someone who was trying to help me fix a problem I had installing Sage 8.0 (which never worked by the way, I've been having these problems for a while...)

[Dima Pasechnik]

On Wednesday, April 25, 2018 at 3:05:13 PM UTC+3, Christelle Vincent wrote:

I do have Homebrew installed! Is there any way to stop it from seeing the wrong library? Can I uninstall Homebrew? I originally installed it at the suggestion of someone who was trying to help me fix a problem I had installing Sage 8.0 (which never worked by the way, I've been having these problems for a while...)

I don't know enough about Homebrew to say whether it is setting some environment variables that you can simply unset, or whether it's more to this.

Anyhow,  https://docs.brew.sh/FAQ has info on uninstalling.

[Samuel Lelievre]

Typically Homebrew installs stuff in /usr/local or /opt/local.

Personally I have Homebrew installed, and I just get it out

of the way when I build Sage, by doing

    sudo mv /usr/local /usr/localbackup

    sudo mv /opt/local /opt/localbackup

and then when I'm done building Sage I revert that with

    sudo mv /usr/localbackup /usr/local

    sudo mv /opt/localbackup /opt/local

To make my life easier I have aliases in my .bash_profile to

- check which state things are in (local or localbackup),

- rename "local" to "localbackup" in /opt and /usr

- rename "localbackup" to "local" in /opt and /usr

The aliases look like this:

alias backuplocal='sudo mv /opt/local /opt/localbackup && sudo mv /usr/local /usr/localbackup'

alias restorelocal='sudo mv /opt/localbackup /opt/local && sudo mv /usr/localbackup /usr/local'

alias islocalthere='ls /opt | grep oca && ls /usr | grep oca'

[Christelle Vincent]

I just uninstalled Homebrew completely (I don't think I need it), deleted my old attempt at making Sage, restarted my laptop and started making Sage again. That did not help any, I still get exactly the same error.

[Dima Pasechnik]

Perhaps it did not uninstall cleanly. Could you post the output of

set

(run it in terminal at the shell prompt)

[Dima Pasechnik]

it is probably better to use the latest Sage pre-release, 8.2.rc4.

note that 8.1 was not tested on osx 10.13, so it really is not supported...

[Erik Bray]

I think you're all barking up the wrong tree. Her build log contains:

./spkg-build: line 66: ![: command not found

Line 66 (really line 34 in the unwrapped sources) of spkg-build for
Python 3 shows:

if ![ -z "$OPENSSL_INCLUDE" ]; then

I think that's clearly a typo. Looks like it's been there for a long
time too. I don't any shell that allows such syntax but I could be
mistasken. Christelle's shell certainly doesn't support it.

On Thu, Apr 26, 2018 at 10:58 AM, Dima Pasechnik <dim...@gmail.com> wrote:
> it is probably better to use the latest Sage pre-release, 8.2.rc4.
>
> note that 8.1 was not tested on osx 10.13, so it really is not supported...
>

> --
> You received this message because you are subscribed to the Google Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.
> To post to this group, send email to sage-...@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/d/optout.

[Jeroen Demeyer]

On 2018-04-26 14:45, Erik Bray wrote:
> Line 66 (really line 34 in the unwrapped sources) of spkg-build for
> Python 3 shows:
>
> if ![ -z "$OPENSSL_INCLUDE" ]; then
>
> I think that's clearly a typo. Looks like it's been there for a long
> time too.

Since #21944 for the record. The question is: is this what causes the
problem? The reported problem doesn't look OpenSSL-related.

[Jeroen Demeyer]

More precisely, the shell simply interprets

if ![ -z "$OPENSSL_INCLUDE" ]

as

if false

so the typo just breaks support for "OPENSSL_INCLUDE".

[Christelle Vincent]

I don't know enough about OpenSSL to weigh in on that part of the discussion. Maybe that would make it impossible to download some python3 package?

In any case, I am attaching for Dima (thank you for all your help by the way) the output of set. I also downloaded and just started making sage.8.2.rc4, so I'll update on that if it works. I'm not super optimistic, I haven't been able to make Sage since 7.4 (7.4 is the last one that worked on this laptop) and now I really need the latest Sage to work...

Christelle

[Dima Pasechnik]

You have /usr/local/bin first (or at all) in your PATH. This might be asking for trouble while building/running Sage, depending upon what you have installed there.

[Samuel Lelièvre]

Tips for building under macOS:

- Check the number of cores you have with the command

      $ sysctl -n hw.ncpu

- Before running `make`, run the following two lines:

      $ export PATH='/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin:/Library/TeX/texbin'

  (this is to remove /usr/local/bin from your PATH while you build Sage)

      $ export MAKE='make -j4'

  (this is to build in parallel, using 4 cores; you can change 4

  to whatever you get as the output of `sysctl -n hw.ncpu`).

​

[Christelle Vincent]

Just started over with Samuel's instructions, will update when it's done!

[Christelle Vincent]

It still didn't work. To recap, this time I tried to install sage-8.2.rc4, and first I ran 

$ export PATH='/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin:/Library/TeX/texbin'

Make still failed at installing python3-3.6.1.p1, the log looks the same but I am attaching it again in case I'm missing something that is actually different.

To me, the money shot is:

Undefined symbols for architecture x86_64:

  "_libintl_bindtextdomain", referenced from:

      _PyIntl_bindtextdomain in _localemodule.o

  "_libintl_dcgettext", referenced from:

      _PyIntl_dcgettext in _localemodule.o

  "_libintl_dgettext", referenced from:

      _PyIntl_dgettext in _localemodule.o

  "_libintl_gettext", referenced from:

      _PyIntl_gettext in _localemodule.o

  "_libintl_textdomain", referenced from:

      _PyIntl_textdomain in _localemodule.o

ld: symbol(s) not found for architecture x86_64

and if I remember correctly this is exactly what went wrong last year when I was trying to install Sage 7.5. There is something about those variables that is just messed up on my computer.

[John H Palmieri]

One difference between your OS X installation and mine: near the start of your log, I see

checking libintl.h usability... yes
checking libintl.h presence... yes
checking for libintl.h... yes

whereas these say "no" on my machine. What does "locate libintl" say (when you run it from the terminal)?

[Christelle Vincent]

It says: Should I install the database then?

WARNING: The locate database (/var/db/locate.database) does not exist.

To create the database, run the following command:

  sudo launchctl load -w /System/Library/LaunchDaemons/com.apple.locate.plist

Please be aware that the database can take some time to generate; once

the database has been created, this message will no longer appear.

Should I install the database then?

[Dima Pasechnik]

On Friday, April 27, 2018 at 8:16:32 PM UTC+1, Christelle Vincent wrote:

It says: Should I install the database then?

that's OK to do so.

I bet the offending library is in /usr/local

(headers in /usr/local/include, the library in /usr/local/lib, have a look)

Move them out of the way.

(e.g. just rename /usr/local to /usr/local_bak)

[John H Palmieri]

On Friday, April 27, 2018 at 12:56:37 PM UTC-7, Dima Pasechnik wrote:

On Friday, April 27, 2018 at 8:16:32 PM UTC+1, Christelle Vincent wrote:

It says: Should I install the database then?

that's OK to do so.

I bet the offending library is in /usr/local

(headers in /usr/local/include, the library in /usr/local/lib, have a look)

Move them out of the way.

(e.g. just rename /usr/local to /usr/local_bak)

On my system, you can't rename /usr/local – you get a message 'mv: rename local to local_bak: Operation not permitted'. So instead you could do

mv /usr/local/lib /usr/local/lib_bak
mv /usr/local/include /usr/local/include_bak

and maybe the same with /usr/local/bin, too.

--
John
 

[Samuel Lelièvre]

2018-04-27 22:18 GMT+02:00 John H Palmieri:
>
> On Friday, April 27, 2018 at 12:56:37 PM UTC-7, Dima Pasechnik wrote:
>>
>> On Friday, April 27, 2018 at 8:16:32 PM UTC+1, Christelle Vincent wrote:
>>>
>>> It says: Should I install the database then?
>>
>> that's OK to do so.
>>
>> I bet the offending library is in /usr/local
>> (headers in /usr/local/include, the library in /usr/local/lib, have a look)
>> Move them out of the way.
>> (e.g. just rename /usr/local to /usr/local_bak)
>
>
> On my system, you can't rename /usr/local – you get a message

> 'mv: rename local to local_bak: Operation not permitted'.

It's a matter of permissions, you need sudo for that.

> So instead you could do
>
> mv /usr/local/lib /usr/local/lib_bak
> mv /usr/local/include /usr/local/include_bak
>
> and maybe the same with /usr/local/bin, too.

It used to be that running the following in the terminal before

running `make` was enough to build SageMath in my setup:

    export PATH='/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin:/Library/TeX/texbin'

but since Sage 8.2.beta2, it is no longer the case, see

    https://groups.google.com/d/msg/sage-release/eQM0qwn2XW8/f2X1wj6cAAAJ

and I now need to additionally do:

    sudo mv /usr/local /usr/localbackup

before building with `make`, and to revert that after `make`

has completed, by doing:

    sudo mv /usr/localbackup /usr/local

It would be interesting to check whether the changed behaviour

can be tracked to a particular commit.

[John H Palmieri]

On Friday, April 27, 2018 at 4:07:52 PM UTC-7, Samuel Lelievre wrote:

2018-04-27 22:18 GMT+02:00 John H Palmieri:
>
> On Friday, April 27, 2018 at 12:56:37 PM UTC-7, Dima Pasechnik wrote:
>>
>> On Friday, April 27, 2018 at 8:16:32 PM UTC+1, Christelle Vincent wrote:
>>>
>>> It says: Should I install the database then?
>>
>> that's OK to do so.
>>
>> I bet the offending library is in /usr/local
>> (headers in /usr/local/include, the library in /usr/local/lib, have a look)
>> Move them out of the way.
>> (e.g. just rename /usr/local to /usr/local_bak)
>
>
> On my system, you can't rename /usr/local – you get a message

> 'mv: rename local to local_bak: Operation not permitted'.

It's a matter of permissions, you need sudo for that.

Sorry, I should have said that even with sudo, I get this message.

[Christelle Vincent]

I was able to do

$ mv /usr/local/lib /usr/local/lib_bak
$ mv /usr/local/include /usr/local/include_bak

$ mv /usr/local/bin /usr/local/bin_bak

(with sudo) and I was able to successfully install python3-3.6.1.... but then the build failed on pyzmq-17.0.0b3. I'm attaching the log but the problem appears to be:

      File "<string>", line 1, in <module>

      File "/private/var/folders/tm/3mshjypj00xg_v3pcxts2p6w0000gn/T/pip-n0dv8L-build/setup.py", line 1273, in <module>

        setup(**setup_args)

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/core.py", line 151, in setup

        dist.run_commands()

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/dist.py", line 953, in run_commands

        self.run_command(cmd)

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/dist.py", line 972, in run_command

        cmd_obj.run()

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/site-packages/setuptools/command/install.py", line 61, in run

        return orig.install.run(self)

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/command/install.py", line 563, in run

        self.run_command('build')

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/cmd.py", line 326, in run_command

        self.distribution.run_command(command)

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/dist.py", line 972, in run_command

        cmd_obj.run()

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/command/build.py", line 127, in run

        self.run_command(cmd_name)

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/cmd.py", line 326, in run_command

        self.distribution.run_command(command)

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/dist.py", line 972, in run_command

        cmd_obj.run()

      File "/private/var/folders/tm/3mshjypj00xg_v3pcxts2p6w0000gn/T/pip-n0dv8L-build/setup.py", line 1115, in run

        self.distribution.run_command('configure')

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/dist.py", line 971, in run_command

        cmd_obj.ensure_finalized()

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/distutils/cmd.py", line 109, in ensure_finalized

        self.finalize_options()

      File "/private/var/folders/tm/3mshjypj00xg_v3pcxts2p6w0000gn/T/pip-n0dv8L-build/setup.py", line 294, in finalize_options

        self.config = discover_settings(self.build_base)

      File "buildutils/config.py", line 163, in discover_settings

        merge(settings, get_cfg_args())

      File "buildutils/config.py", line 90, in get_cfg_args

        cfg.read('setup.cfg')

      File "/Users/vincent/Applications/sage-8.2.rc4/local/lib/python2.7/site-packages/backports/configparser/__init__.py", line 704, in read

        with open(filename, encoding=encoding) as fp:

    TypeError: bad argument type for built-in operation

    Running setup.py install for pyzmq: finished with status 'error'

Trying Samuel's suggestion, I was able to type in $ export PATH='/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin:/Library/TeX/texbin'

but $ sudo mv /usr/local /usr/localbackup gives me Operation not permitted.

With just the export PATH statement, the build still fails at installing pyzmq-17.0.0b3.

[Dima Pasechnik]

after these changes in /usr/local, did you do

make distclean

without it, stuff in /usr/local that you moved away could be still linked to, and could cause all sorts of errors.

[Christelle Vincent]

I had not typed that. In general if something was not explicitly said then I didn't do it, because while I can develop for Sage I'm not good enough with the guts of a computer to know to do anything extra.

Anyway, after make distclean, I couldn't build gcc-7.2.0, which at the very beginning wouldn't build but when I installed the latest command line tools I could. So that feels like a step back... A lot fewer packages built this time than last time. I cannot attach the log because for some reason it's 13MB (??) but it seems that the problem is:

/Users/vincent/Applications/sage-8.2.rc4/local/var/tmp/sage/build/gcc-7.2.0/src/libstdc++-v3/include/precompiled/stdc++.h:118:10: fatal error: unordered_set: No such file or directory

 #include <unordered_set>

[Dima Pasechnik]

Try 8.2.rc2 with https://trac.sagemath.org/ticket/25118

on top of it. This would not build gcc, but rather use Xcode's C compiler.

In his infinite wisdom the release manager decided to wait till 8.3 with it...

[Christelle Vincent]

Thanks!

I actually went ahead and tried and built 8.1 and that worked!! So I think that whatever you suggested before with exporting the path and renaming local/bin, local/share and local/lib worked!

Thank you so much to everyone who helped with this, I really appreciate finally being able to build the newest Sage!

Best,

Christelle

[Samuel Lelièvre]

Congratulations for building SageMath!

You can now rename /local/bin_bak, /local/share_bak and /local/lib_bak

back to their original names. (And just temporarily rename them to the

_bak version when you need to build SageMath.

Happy SageMath usage and development!

--
You received this message because you are subscribed to a topic in the Google Groups "sage-devel" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/sage-devel/D06JbNlM0u0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to sage-devel+unsubscribe@googlegroups.com.

[Christelle Vincent]

Done! Thanks!

On Saturday, April 28, 2018 at 6:45:31 PM UTC-4, Samuel Lelievre wrote:

Congratulations for building SageMath!

You can now rename /local/bin_bak, /local/share_bak and /local/lib_bak

back to their original names. (And just temporarily rename them to the

_bak version when you need to build SageMath.

Happy SageMath usage and development!

2018-04-29 0:20 GMT+02:00 Christelle Vincent <christell...@uvm.edu>:

Thanks!

I actually went ahead and tried and built 8.1 and that worked!! So I think that whatever you suggested before with exporting the path and renaming local/bin, local/share and local/lib worked!

Thank you so much to everyone who helped with this, I really appreciate finally being able to build the newest Sage!

Best,

Christelle

On Saturday, April 28, 2018 at 3:37:00 PM UTC-4, Dima Pasechnik wrote:

Try 8.2.rc2 with https://trac.sagemath.org/ticket/25118

on top of it. This would not build gcc, but rather use Xcode's C compiler.

In his infinite wisdom the release manager decided to wait till 8.3 with it...

--
You received this message because you are subscribed to a topic in the Google Groups "sage-devel" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/sage-devel/D06JbNlM0u0/unsubscribe.

To unsubscribe from this group and all its topics, send an email to sage-devel+...@googlegroups.com.

[Dima Pasechnik]

On Sunday, April 29, 2018 at 12:22:18 AM UTC+1, Christelle Vincent wrote:

Done! Thanks!

On Saturday, April 28, 2018 at 6:45:31 PM UTC-4, Samuel Lelievre wrote:

Congratulations for building SageMath!

You can now rename /local/bin_bak, /local/share_bak and /local/lib_bak

back to their original names. (And just temporarily rename them to the

_bak version when you need to build SageMath.

Happy SageMath usage and development!

Seconded :-)

I'd also recommend running

make ptest

to make sure that it all works, and there are no more kinks to iron out

[Christelle Vincent]

Well, I don't know if you want to know... But make ptest did end with an error:

Doctests interrupted: 2203/3630 files tested

----------------------------------------------------------------------

Total time for all tests: 12716.1 seconds

    cpu time: 7631.7 seconds

    cumulative wall time: 13780.4 seconds

Traceback (most recent call last):

  File "/Users/vincent/Applications/sage-8.1/src/bin/sage-runtests", line 125, in <module>

    err = DC.run()

  File "/Users/vincent/Applications/sage-8.1/local/lib/python2.7/site-packages/sage/doctest/control.py", line 1144, in run

    self.run_doctests()

  File "/Users/vincent/Applications/sage-8.1/local/lib/python2.7/site-packages/sage/doctest/control.py", line 872, in run_doctests

    self.dispatcher.dispatch()

  File "/Users/vincent/Applications/sage-8.1/local/lib/python2.7/site-packages/sage/doctest/forker.py", line 1774, in dispatch

    self.parallel_dispatch()

  File "/Users/vincent/Applications/sage-8.1/local/lib/python2.7/site-packages/sage/doctest/forker.py", line 1598, in parallel_dispatch

    w.kill()

  File "/Users/vincent/Applications/sage-8.1/local/lib/python2.7/site-packages/sage/doctest/forker.py", line 2079, in kill

    os.killpg(self.pid, signal.SIGQUIT)

OSError: [Errno 1] Operation not permitted

make: *** [ptest] Error 1

[Dima Pasechnik]

On Sunday, April 29, 2018 at 3:54:08 PM UTC+1, Christelle Vincent wrote:

Well, I don't know if you want to know... But make ptest did end with an error:

Doctests interrupted: 2203/3630 files tested

Is this reproducible?

(i.e., does it fail at the same spot if you re-run this?)

To me this looks like your computer suddenly decided it had more important stuff to run...

[Dima Pasechnik]

On Tuesday, May 1, 2018 at 10:59:55 AM UTC+1, Dima Pasechnik wrote:

On Sunday, April 29, 2018 at 3:54:08 PM UTC+1, Christelle Vincent wrote:

Well, I don't know if you want to know... But make ptest did end with an error:

Doctests interrupted: 2203/3630 files tested

Is this reproducible?

(i.e., does it fail at the same spot if you re-run this?)

If it is, then it would be interesting to have a look at

logs/ptest.log

to see if there are some strange failures there (there should not be any errors, perhaps apart from timeouts)

[Jeroen Demeyer]

On 2018-04-29 16:54, Christelle Vincent wrote:
> Doctests interrupted: 2203/3630 files tested

You didn't accidentally press CTRL-C, did you?