Could you please try rebasing this on top of master, to see whether tests work with the addition of #219 ?

Sorry for the delay in looking at all this. It looks generally good to me. A few concerns:

• First of all - could you please rebase it on master, to see that tests are passing with all these changes?
• If we want this to properly install on different platforms, we're going to have to rebuild the messenger, huh?
• Could you please provide a couple of tests for the new functionality?

No worries on the delay.

I'll get to work on writing some tests, but the updates on master seem to have broken my build; I'm getting

 Invalid MEX-file
'/home/bbeltr1/Developer/python-matlab-bridge/venv-pymatbridge/lib/python3.4/site-packages/pymatbridge/matlab/messenger.mexa64':
/opt/MATLAB/R2014b/bin/glnxa64/../../sys/os/glnxa64/libstdc++.so.6: version
`GLIBCXX_3.4.20' not found (required by /usr/lib/x86_64-linux-gnu/libzmq.so.3)

I'll have to look into this.

Did you rebuild the messenger after updating?

On Mon, Sep 21, 2015 at 2:42 PM, Bruno Beltran [email protected]
wrote:

No worries on the delay.

I'll get to work on writing some tests, but the updates on master seem to
have broken my build; I'm getting

Invalid MEX-file
'/home/bbeltr1/Developer/python-matlab-bridge/venv-pymatbridge/lib/python3.4/site-packages/pymatbridge/matlab/messenger.mexa64':
/opt/MATLAB/R2014b/bin/glnxa64/../../sys/os/glnxa64/libstdc++.so.6: version
`GLIBCXX_3.4.20' not found (required by /usr/lib/x86_64-linux-gnu/libzmq.so.3)

I'll have to look into this.

—
Reply to this email directly or view it on GitHub
#218 (comment)
.

oh wait, so I realize what happened. I replaced $matlabpath/sys/os/glnxa64/libstdc++.so that MATLAB ships with with a modern, working version of libc as a workaround for unrelated issues ages ago. So the fact that it is currently failing to build on my computer probably has to do with the fact that I reinstaled matlab, not with me rebasing on master. On the other hand, how did you get this working on Linux previously? In its current state, a new libzmq should always complain when it tries to start in an environment with LD_LIBRARY_PATH set to point to matlab's old, broken glibc++. On the other hand, starting matlab without all the libraries it normally ships with in its modified LD_LIBRARY_PATH breaks matlab on my system for every version from 2013a to 2015b.

Thoughts?

should I have rebuilt? I thought the whole point of the update was for the mex file to work on Ubuntu?

Not sure. I am still trying to digest your previous message.

On Mon, Sep 21, 2015 at 2:51 PM, Bruno Beltran [email protected]
wrote:

should I have rebuilt? I thought the whole point of the update was for the
mex file to work on Ubuntu?

—
Reply to this email directly or view it on GitHub
#218 (comment)
.

I probably wasn't very clear.

Specifically, the problem I have with the current setup is that, even before any of my changes, if I build the messenger myself and put everything in the "right" place, then I get the following error:

In [1]: from pymatbridge import Matlab
lab = Matlab(executable='/home/bbeltr1/bin/mat')
lab.start()
   ...:
Invalid MEX-file
'/home/bbeltr1/Developer/python-matlab-bridge/venv-pymatbridge/lib/python3.4/site-packages/pymatbridge/matlab/messenger.mexa64':
/opt/MATLAB/R2014b/bin/glnxa64/../../sys/os/glnxa64/libstdc++.so.6: version
`GLIBCXX_3.4.20' not found (required by /usr/lib/x86_64-linux-gnu/libzmq.so.3)

Error in matlabserver (line 13)
messenger('init', socket_address);

2015-09-21 15:44:51,624 - python-matlab-bridge - WARNING - MATLAB session timed out after 15 seconds
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-1-8fe4fca3a574> in <module>()
      1 from pymatbridge import Matlab
      2 lab = Matlab(executable='/home/bbeltr1/bin/mat')
----> 3 lab.start()
      4

/home/bbeltr1/Developer/python-matlab-bridge/venv-pymatbridge/lib/python3.4/site-packages/pymatbridge/pymatbridge.py in start(self)
    226         # Test if connection is established
    227         if not self.is_connected():
--> 228             raise ValueError("%s failed to start" % self._program_name())
    229
    230         self.logger.info("%s started and connected!", self._program_name())

ValueError: MATLAB failed to start

This error is due to MATLAB shipping with an ancient version of libc, and is simply resolved by symlinking the file $matlab_root/sys/os/glnxa64/libstdc++.so to the system's libstdc++.so.6 file.

However, this is strangely not necessary if I use one of your old messenger.mex files.

I am guessing this was worked around before by building the messenger statically?

Sorry for the quadruple-comment, been running around a lot today. So I guess after getting back to my laptop it seems like the update works after rebasing, then again, that's mostly because my version of the messenger has overwritten yours after rebasing, so Travis is still failing. Any pointers on how to build the messenger so that Travis doesn't get so upset?