Doing a "deep-clean" of boost and reverting to using CLANG with no additional flags produced a working version as the compile finished.
Now on to playing with getting MPI, OMP and everything else working. Thank you
// Gustaf
________________________________
Från: Gustaf Olsson <gustaf.olsson.lnu.se>
Skickat: den 5 maj 2020 18:22:03
Till: AMBER Mailing List
Ämne: Re: [AMBER] AmberTools20 comilation on macOS
Hello David
As per ususal I am trying to figure out what has priority over whatever else and and try to navigate the interesting macOS system. The normal issues getting homebrew installed versions to get priority over the apple alternatives is a struggle at times. Really surprised to see that I am not able to edit the symlink for CC in /usr/bin without having to turn root.
I do have boost installed though I did also unlink the boost installation and this did not solve the issue.
Using the "./quick_cmake_install -mpi -openmp CLANG" and receiving the error:
BLAS_error.c:1:10: fatal error: 'stdio.h' file not found
#include <stdio.h>
^~~~~~~~~
1 error generated.
Was particularly disappointing as I remember this being a problem with the deprecated inculde files with Xcode that caused some problems last fall (?).
I am aware that homebrew does cause some issues and that the latest Xcode update broke the gcc9 installation from homebrew. I'll keep trying until I find a combination that works.
Best regards
// Gustaf
________________________________
Från: David A Case <david.case.rutgers.edu>
Skickat: den 5 maj 2020 14:58:58
Till: AMBER Mailing List
Ämne: Re: [AMBER] AmberTools20 comilation on macOS
On Tue, May 05, 2020, Gustaf Olsson wrote:
>
> "boost::re_detail_107200::raw_storage::insert(unsigned long, unsigned long)", referenced from:
> boost::re_detail_107200::basic_regex_parser<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::parse_repeat(unsigned long, unsigned long) in metatwist.cpp.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[2]: *** [AmberTools/src/moft/metatwist] Error 1
> make[1]: *** [AmberTools/src/moft/CMakeFiles/metatwist.dir/all] Error 2
> make: *** [all] Error 2
>
>This is using the built in (xcode) compilers.
Boy, I'm not real sure how to reproduce this. I'm guessing you may(?)
have boost installed somewhere on your machine (via homebrew? conda?).
Is that a possibility?
Try adding -DFORCE_EXTERNAL_LIBS=boost, if you indeed have another boost
installation.
Does your build report provide any clues? (This is at the bottom of the
cmake.log file). Posting that might help us figure out what is going
on.
To skip metatwist/moft, set -DDISABLE_TOOLS=moft in your run_cmake file,
and see if other problems are arising.
>Applications/CMake.app is not present by default. Hence this test "fails"
>and sets cmake=cmake. Using a package manager (for me homebrew) I have
>cmake in /usr/local/bin/.
I doubt that this would lead to the problems you report. I've used
cmake from both kitware and macports, without any problems. What
compiler was used to build cmake should be irrelevant, as far as I can
see.
>
>
>During the configuration stage, I also notice that the script does
>not find open-mpi or mkl libraries, even if these are installed. I
>assume there might be a need for some "hands on" convincing regarding
>parallelization.
It's worth your while to visit the common options page:
https://ambermd.org/pmwiki/pmwiki.php/Main/CMake-Common-Options
and to read the section in the manual about MKL and MPI (if you have not
already done so.) In brief, MKL is "opt-in": you have to ask for it.
The MPI stack that is used is the one in your PATH: if that is openmpi,
then it should be used. But in my experience, you have to compile your
own MPI to use clang + gfortran.
>
>On another note, is there any way to define which compilers to use on the
>CLI (like before with configure gnu)?
If you have the gnu compilers in your path, just set -DCOMPILER=GNU,
(instead of CLANG), and you'll get the gnu compilers. A homebrew or
macports setup ought to work.
...good luck....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue May 05 2020 - 10:30:02 PDT
