Segmentation fault on OSX anaconda python

[RasmusFonseca]

Just did a clean install of anaconda and I'm now getting segfaults when importing:

MBPro:~/Programs $ conda activate py27
(py27) MBPro:~/Programs $ python
Python 2.7.14 |Anaconda, Inc.| (default, Dec  7 2017, 11:07:58)
[GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import vmd
Segmentation fault: 11
(py27) MBPro:~/Programs $ conda activate base
(base) MBPro:~/Programs $ python
Python 3.6.4 |Anaconda, Inc.| (default, Jan 16 2018, 12:04:33)
[GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import vmd
Segmentation fault: 11

[Eigenstate]

It looks like OSX + Python 2.7 is broken in some way, as my Travis CI builds do this too. I can't debug this without a stack trace at minimum, and probably can't at all without a mac. Can you use Python 3.6?

[RasmusFonseca]

The base environment is 3.6.4. Would this be a useful stack-trace:

MBPro:~/Programs $ python --version
Python 3.6.4 :: Anaconda, Inc.
MBPro:~/Programs $ lldb -- python -c "import vmd"
(lldb) target create "python"
Current executable set to 'python' (x86_64).
(lldb) settings set -- target.run-args  "-c" "import vmd"
(lldb) r
Process 75708 launched: '/Users/rfonseca/anaconda3/bin/python' (x86_64)
Process 75708 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x25)
    frame #0: 0x0000000208f2eabb libpython3.6m.dylib`PyUnicode_InternInPlace + 107
libpython3.6m.dylib`PyUnicode_InternInPlace:
->  0x208f2eabb <+107>: movb   0x25(%rax), %r15b
    0x208f2eabf <+111>: movq   0x1d1aa2(%rip), %rax      ; _PyThreadState_Current
    0x208f2eac6 <+118>: movb   $0x1, 0x25(%rax)
    0x208f2eaca <+122>: movq   %rbx, %rsi
Target 0: (libvmd.so) stopped.
(lldb) bt
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x25)
  * frame #0: 0x0000000208f2eabb libpython3.6m.dylib`PyUnicode_InternInPlace + 107
    frame #1: 0x0000000208fafc5f libpython3.6m.dylib`PyImport_Import + 351
    frame #2: 0x0000000208fb106c libpython3.6m.dylib`PyImport_ImportModule + 28
    frame #3: 0x000000020597cee4 libvmd.so`get_vmdapp() + 16
    frame #4: 0x00000002059854d6 libvmd.so`PyInit_vmd + 22
    frame #5: 0x000000010018e749
    frame #6: 0x000000010018dddc
    frame #7: 0x000000010008b562
    frame #8: 0x0000000100161a10
    frame #9: 0x0000000100154589
    frame #10: 0x00000001001636aa
    frame #11: 0x00000001001632ac
    frame #12: 0x0000000100160f4c
    frame #13: 0x00000001001635fc
    frame #14: 0x00000001001632ac
    frame #15: 0x0000000100160f4c
    frame #16: 0x00000001001635fc
    frame #17: 0x00000001001632ac
    frame #18: 0x0000000100160f4c
    frame #19: 0x00000001001635fc
    frame #20: 0x00000001001632ac
    frame #21: 0x0000000100160f4c
    frame #22: 0x00000001001635fc
    frame #23: 0x00000001001632ac
    frame #24: 0x0000000100160f4c
    frame #25: 0x0000000100154589
    frame #26: 0x00000001001636aa
    frame #27: 0x00000001001632ac
    frame #28: 0x0000000100160f4c
    frame #29: 0x0000000100154589
    frame #30: 0x000000010014ed99
    frame #31: 0x000000010008b562
    frame #32: 0x0000000100161a10
    frame #33: 0x0000000100154589
    frame #34: 0x00000001001636aa
    frame #35: 0x00000001001632ac
    frame #36: 0x0000000100160f4c
    frame #37: 0x00000001001635fc
    frame #38: 0x00000001001632ac
    frame #39: 0x0000000100160f4c
    frame #40: 0x00000001001635fc
    frame #41: 0x00000001001632ac
    frame #42: 0x0000000100160f4c
    frame #43: 0x00000001001635fc
    frame #44: 0x00000001001632ac
    frame #45: 0x0000000100160f4c
    frame #46: 0x0000000100163d4f
    frame #47: 0x0000000100009e80
    frame #48: 0x00000001000130da
    frame #49: 0x0000000100188674
    frame #50: 0x000000010015ee17
    frame #51: 0x0000000100154589
    frame #52: 0x00000001001af599
    frame #53: 0x00000001001af437
    frame #54: 0x00000001001d1b16
    frame #55: 0x0000000100001e7d
    frame #56: 0x00007fffa8b23235
(lldb)

[RasmusFonseca]

If not I'll come by with a mac tomorrow.

[Eigenstate]

Okay I have added a flag to the build process that will build the main library (not the plugins though) with debug symbols.

git clean -fdx
git pull
python setup.py build --debug
python setup.py install

Load that up into the debugger and you'll get a more helpful stack trace. All of the source files live in vmd/vmd_src/src.

[RasmusFonseca]

Still haven't managed to get debug symbols or resolve the issue, but I noticed something strange when building:

   LIB: libtcl8.5.dylib -> /usr/lib
   INC: tcl.h -> /Users/rfonseca/anaconda3/envs/py36_np1.13.3/include
   LIB: libsqlite3.dylib -> /Users/rfonseca/anaconda3/envs/py36_np1.13.3/lib
   INC: sqlite3.h -> /Users/rfonseca/anaconda3/envs/py36_np1.13.3/include
   LIB: libexpat.dylib -> /Users/rfonseca/anaconda3/envs/py36_np1.13.3/lib
   INC: expat.h -> /Users/rfonseca/anaconda3/envs/py36_np1.13.3/include
   LIB: libnetcdf.dylib -> /Users/rfonseca/anaconda3/envs/py36_np1.13.3/lib
   INC: netcdf.h -> /Users/rfonseca/anaconda3/envs/py36_np1.13.3/include
   LIB: libpython3.6*.dylib -> /Users/rfonseca/anaconda3/envs/py36_np1.13.3/lib

Isn't this weird given there's a perfectly valid /Users/rfonseca/anaconda3/envs/py36_np1.13.3/lib/libtcl8.6.dylib?

[RasmusFonseca]

I think that was the problem. setting up my conda environment with tk=8.5 solves the problem.

[RasmusFonseca]

Ok, so for the record, creating a conda-environment the following way:

conda create -n vmd-python python=3.6 netcdf4 numpy expat tk=8.5
conda activate vmd-python
cd path/to/vmd-python
python setup.py build
python setup.py install

.. works. I guess the problem was that it was detecting the specific tcl version (8.5) in /usr/lib but not caring where it got the includes from, so they didn't match because the most recent conda version was 8.6. Specifically installing tk 8.5 in conda solved this.