Segfault on MacOS caused by LibGMT.log_to_file

[chandrex]

Hi, I've been trying to install gmt-python for the past day and a half with no avail. I am on macOS 10.13.4, Python 3.6.5 (installed with Anaconda 3). I've been following all of the installation instruction, and everything seems fine except when I get to the very end, with the gmt.test().

I am in the conda environment (using source activate gmt-python), and trying this in the Python interpreter. This is what happens:

(gmt-python) Chandras-MacBook-Pro:~ chandra$ python
Python 3.6.5 | packaged by conda-forge | (default, Apr 6 2018, 13:44:09)
[GCC 4.2.1 Compatible Apple LLVM 6.1.0 (clang-602.0.53)] on darwin
Type "help", "copyright", "credits" or "license" for more information.

import gmt
gmt.test()

returns

============================= test session starts ==============================
platform darwin -- Python 3.6.5, pytest-3.5.0, py-1.5.3, pluggy-0.6.0 -- /anaconda3/envs/gmt-python/bin/python
cachedir: .pytest_cache
Matplotlib: 2.2.2
Freetype: 2.8.1
rootdir: /Users/chandra, inifile:
plugins: mpl-0.9
collected 126 items
gmt/base_plotting.py::gmt.base_plotting.BasePlotting._preprocess <- ../../anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/base_plotting.py PASSED [ 0%]
gmt/figure.py::gmt.figure.Figure <- ../../anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/figure.py FAILED [ 1%]
gmt/clib/core.py::gmt.clib.core.LibGMT <- ../../anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py PASSED [ 2%]
gmt/clib/core.py::gmt.clib.core.LibGMT._check_dtype_and_dim <- ../../anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py PASSED [ 3%]
gmt/clib/core.py::gmt.clib.core.LibGMT.extract_region <- ../../anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py gmt-python-session [ERROR]: Long input record (236 bytes) was truncated to first 126 bytes!
FAILED [ 3%]
gmt/clib/core.py::gmt.clib.core.LibGMT.grid_to_vfile <- ../../anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py FAILED [ 4%]
gmt/clib/core.py::gmt.clib.core.LibGMT.log_to_file <- ../../anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py Segmentation fault: 11

I feel like this might be because of a bash profile issue. I am not sure how to set up a .bashrc profile properly on the Mac (sorry!), but my .bash_profile looks like this (note the 'added by...' is commented out):

added by Anaconda3 5.1.0 installer
export PATH="/anaconda3/bin:$PATH"
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/anaconda3/envs/gmt-python/lib

Any help would be much appreciated! 😄
ps. let me know if any clarifications are needed, sorry first time writing an issue in!
pps. attached a screenshot of terminal, just to be super clear!

[leouieda]

@chandrex thank you reporting this and sorry for all the trouble! I really appreciate the amount of effort you put in to the issue. That really helps! Welcome to the world of open-source! I hope you'll like it and consider becoming involved 😀

OK, this is a bad crash. It normally shouldn't happen but it's really tricky dealing with C code from Python. One possibility is that Python is finding the wrong version of GMT. Do you have GMT installed from another source (homebrew, fink, etc)?

Could you try removing the LD_LIBRARY_PATH line? That's not usually necessary unless you can't get it to work otherwise. Sorry about this, it should be clearer in the docs and I opened #165 to remind me to deal with this.

If that doesn't work, then can you please post the output of conda list? That way I can try to see if everything is being installed in the correct version.

Again, thanks for detailed issue!

[chandrex]

Hiya,

Yes I have GMT installed already (5.4.2), which I installed with homebrew. It won't work with or without the LD_LIBRARY_PATH (I tried!), but I will remove it anyhow.

Here is my conda list

Thanks for your help 😁

[leouieda]

OK, here are a few things you can try:

1. I noticed that your pandas and xarray were installed by pip. Which indicates that they were pulled in when you installed gmt-python. Also, your numpy is a release candidate. We noticed these problems before so we updated the install instructions to match. Could you dele the current environment (conda env remove -n gmt-python) and follow these steps?
2. If that doesn't work, then try setting LD_LIBRARY_PATH using export LD_LIBRARY_PATH=$HOME/anaconda3/envs/gmt-python/lib:$LD_LIBRARY_PATH . This gives anaconda a higher priority.

I'm working on a better way that we can diagnose these problems (#167).

Again, thanks for reporting this!

[leouieda]

@chandrex just to let you know that your issue contributed several changes to the project: #168 #173 #175 #176 and #177

#177 introduces a new way to set the path to the GMT shared library. Once, that's incorporated (it will say "Merged" on the link) please try deleting your conda environment and following the new instructions that will appear in http://www.gmtpython.xyz/install.html

Again, thank you for helping us make things more robust!

[leouieda]

@chandrex please let me know if the new changes work for you. If not we can at least better diagnose the problem.

[chandrex]

Hi @leouieda, alas I have tried everything you suggested and I am still getting the same problem at the final step when I do the gmt.test(), with the failures and segmentation fault.
Steps I took apart from your suggestions:
• Set up a proper .bashrc (my mac did not have one previously)
• I thought it could have been a problem with where anaconda was installed, so uninstalled/reinstalled in where $HOME is (it was not there previously)
Maybe I should uninstall my GMT 5 version and see if I can get it working without it? What do you think?

[leouieda]

:( sorry to hear. You shouldn't have uninstall GMT5. Could you please post the new output of gmt.test()? This should let us diagnose the problem now.

[chandrex]

Hi @leouieda here is the screenshot:

And in writing:

import gmt
gmt.test()

=========================== Currently loaded libgmt ============================
binary dir: /Users/chandra/anaconda3/envs/gmt-python/bin
cores: 4
grid layout: rows
image layout:
library path: /Users/chandra/anaconda3/envs/gmt-python/lib/libgmt.dylib
padding: 2
plugin dir: /Users/chandra/anaconda3/envs/gmt-python/lib/gmt/plugins
share dir: /Users/chandra/anaconda3/envs/gmt-python/share/gmt
version: 6.0.0
============================= test session starts ==============================
platform darwin -- Python 3.6.5, pytest-3.5.0, py-1.5.3, pluggy-0.6.0 -- /Users/chandra/anaconda3/envs/gmt-python/bin/python
cachedir: .pytest_cache
Matplotlib: 2.2.2
Freetype: 2.8.1
rootdir: /Users/chandra, inifile:
plugins: mpl-0.9
collected 132 items

anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/base_plotting.py::gmt.base_plotting.BasePlotting._preprocess PASSED [ 0%]
anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/figure.py::gmt.figure.Figure FAILED [ 1%]
anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py::gmt.clib.core.LibGMT PASSED [ 2%]
anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py::gmt.clib.core.LibGMT._check_dtype_and_dim PASSED [ 3%]
anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py::gmt.clib.core.LibGMT.extract_region gmt-python-session [ERROR]: Long input record (236 bytes) was truncated to first 126 bytes!
FAILED [ 3%]
anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py::gmt.clib.core.LibGMT.grid_to_vfile PASSED [ 4%]
anaconda3/envs/gmt-python/lib/python3.6/site-packages/gmt/clib/core.py::gmt.clib.core.LibGMT.log_to_file Segmentation fault: 11

[mikehagerty]

Hi.
I just followed your wiki download/install instructions on mac os 10.10.5 and I get the exact same error:

gmt.test()
...
gmt/clib/core.py::gmt.clib.core.LibGMT.get_libgmt_func PASSED [ 4%]
gmt/clib/core.py::gmt.clib.core.LibGMT.grid_to_vfile PASSED [ 5%]
gmt/clib/core.py::gmt.clib.core.LibGMT.log_to_file Segmentation fault: 11

My dependencies (e.g., xarray, pandas) were all installed with conda.
Everything looks good (no complaints) except for the seg fault.

Please let me know if anyone makes any progress on solving this.

Cheers!
-Mike

[leouieda]

Hi @mikehagerty and @chandrex this is a very strange failure. I'll try to get my hands on a Mac to investigate.

@mikehagerty out of curiosity, do you also have GMT5 installed?

[leouieda]

OK, I managed to replicate the issue using Travis CI on #180. It's not a problem with your installation or with the pre-installed GMT5. I'll investigate further to try to get to the bottom of this.

Thanks for bringing this to my attention! It's hard to track these issues when I'm not using a Mac.