The `envi.SegmentationViolation()` was missing the `va` required
parameter. This has started failing now, because calling
`vw.getSegment(0x4BA190)` for the `tests/data/mimikatz.exe_` produces
different results in Python 2 and Python 3. It returns `None` in Python
3 while the output in Python 2 is:
`(4939776, 16840, '.data', 'mimikatz')`
I have reported the issue to vivisect:
https://github.com/vivisect/vivisect/issues/370
`va` can be None and this causes Python 3 to raise a TypeError
exception. This is caused by the following breaking change in Python3:
> The ordering comparison operators (<, <=, >=, >) raise a TypeError
> exception when the operands don’t have a meaningful natural ordering.
This didn't failed in the previously tried vivisect version (master from
one week ago and not the release). This may have been caused by a bug in
vivisect that has been fixed.
Make the `backend` argument required in the `get_extractor` internal
routine. Specify a backend in the scripts which call this function. Add
a CLI backend option in capa/features/freeze.py as well.
Now we have two working backends in Python3! Add an option to select
which one to use. With this code, vivisect is the default backend, but
this is really easy to change. We could do some analysis to see if smda
performances better than vivisect once the vivisect implementation.
Compare the performance of vivisect Python 2 vs 3 by counting the number
of feature of each type extracted for every binary in `tests/data`.
Render the ones that perform bad (under a threshold - 98) and the total
performance. Render also the running time per binary for both Python 2 and 3.
From this result, it seems that vivisect behaves properly with Python3.
Vivisect has moved to Python3. Allow to run vivisect with Python3 in
capa.
I am using the following version of vivisect (which includes fixes for
some bugs I have found and some open PRs in vivisect):
https://github.com/Ana06/vivisect/tree/py-3
