[Swig-user] swig-passed values are garbage except for weird interaction with numpy

[Angus M. <am...@gm...>]

Hi all,

I have some C++ code that implements a data messaging system between
processes, and from which I am generating a python wrapper using swig.
90% of the time this is all working fine. In two instances though I
have seen a problem where a variable passed into python contains
garbage values, rather than the expected vector of zeros. The
weirdness is that if I, in python, convert the data into a numpy array
AND print that array, then I get the values I expect. I've tried this
with only one of either using np.asarray or the print, but the "fix"
only works with both. Obviously I'd like to not have to depend on a
print statement to get my values out.

Here's a little code python code snippet to show what I mean:

################

import numpy as np

# this is just data format declarations
import RTMA_config as rc

# these are the swig-wrapped messaging interface
from PyRTMA import RTMA_Module, CMessage, copy_from_msg

def setup_RTMA(server):
    mod = RTMA_Module(rc.MID, 0)
    mod.ConnectToMMM(server)
    mod.Subscribe(rc.MT_DATA)
    mod.SendModuleReady()
    print "Connected to RTMA at", server
    return mod

def run(mod):
    while True:
        msg = CMessage()
        rcv = mod.ReadMessage(msg, 0.1)
        if rcv == 1:
            process_message(msg)

def process_message(msg):
    if msg.GetHeader().msg_type == rc.MT_DATA:
        data = rc.MDF_DATA()
        copy_from_msg(data, msg)
        #print "A", np.asarray(data.dof_vals[:])   # this is the
critical line -> put this in and the next line works
        print "B", data.dof_vals[:]                     # shows
garbage values without previous line, or 0s with it

server = 'localhost:7111'
mod = setup_RTMA(server)
run(mod)

################

copy_to_msg is defined in PyRTMA.i as:

%pythoncode %{
        from ctypes import memmove, addressof, sizeof

        def copy_from_msg(data, cmsg):
                memmove(addressof(data), cmsg.data, sizeof(data))
%}

Does anyone have any idea what might be causing my particular issue,
or how I should go about debugging it?

Thanks,

Angus.
--
AJC McMorland
Post-doctoral research fellow
Neurobiology, University of Pittsburgh