clusterByAlignment fails in computeSimilarityMatrix on RawMessages

[skleber]

Clustering and aligning fails with a SIGSEV for manually provided RawMessages as can be tested by the following example:

from netzob.all import *
m1=RawMessage("123")
m2=RawMessage("345")
messages = [m1, m2]
Format.clusterByAlignment(messages)

I tracked it down to the C-implemented ScoreComputation-library, to be exact the following line 192 in ClusterByAlignment.py:

        (listScores) = _libScoreComputation.computeSimilarityMatrix(
            self.internalSlick, self._cb_executionStatus, self._isFinish,
            debug, wrapper)

At this point I got stuck with debugging C inside of python for now.

To verify, other message contents failed also:

m1=RawMessage("\x41\x42\xab\xac\x09\x70\x95\xcc\xef")
m2=RawMessage("\x41\x42\x0a\x70\x03\x8f\x23\x5f")

or

m1=RawMessage("FGHJERTZ")
m2=RawMessage("CsdJERsd")

[Sygus]

Hi Stephan, this should work when using bytes for RawMessage:

i.e.

m1=RawMessage(b"123")
m2=RawMessage(b"345")

instead of

m1=RawMessage("123")
m2=RawMessage("345")

This is due to the fact that bytes and Unicode strings are handled with different C objects in the C library. And the C library expects only bytes.

But, a SIGSEV is not a very nice behavior... :D Maybe adding some type checking when building RawMessage by hand would be nice.