FOR_NETCON

[nrnhines]

Framework in place along with comment describing the data structures.

This interacts with BlueBrain/mod2c#43 and neuronsimulator/nrn#695 and they need to be built together.

[pramodk]

@nrnhines : the CI failures are caused by this line :

CoreNeuron/coreneuron/nrnoc/md1redef.h

Line 62 in f2e0f01

#define type _type

[nrnhines]

This has been tested with a mod file with a printf statement that prints all the external events arriving at the NET_RECEIVE block and processing the stdout with

$ cat test/coreneuron/test.sh
#!/bin/bash

nrniv -python test_fornetcon.py | sed -n '/zzz/s/^.*zzz //p'|uniq|sed '1d'> temp
sed -n '1,/INITIAL/p' < temp > temp1
sed '1,/INITIAL/d' < temp > temp2
sortspike temp1 temp
mv temp temp1
sortspike temp2 temp
mv temp temp2
diff temp1 temp2

(The uniq part of the filter converts the 10 lines from the INITIAL block from printf("zzz INITIAL\n") to just one line in order to conveniently separate the output of NEURON from the output of CoreNEURON).

Anyway, a better test would be for CoreNEURON to send back the weights to NEURON at the end of the run (just like the spikes) and then it will be easy to integrate the test into nrn/test/coreneuron for automatic testing.

Til then, I think it would be best to go ahead and merge this into the master as CoreNEURON is undergoing major refactoring that causes conflict with this branch.

[alexsavulescu]

LGTM, just a few suggestions.

[alexsavulescu]

@pramodk we should see about stopping running jenkins jobs when PR gets updated (save some time)

[pramodk]

LGTM

[pramodk]

@pramodk we should see about stopping running jenkins jobs when PR gets updated (save some time)

thats true! Seems like there are some solutions.