Problems using mmv with Python

[taintedkernel]

Registering this issue after discussing the problem in #pcp without any resolution.

After discovering the ability to instrument Python code with PCP, I attempted to do so with example code provided in the module - src/python/pcp/mmv.py. With a few modifications to get it to run, I had problems retrieving the data with pminfo. I also compiled the acme.c example, which worked just fine - pminfo -f mmv saw it without problems (and /var/lib/pcp/tmp/mmv/acme had data). Strace returned nothing for /var/lib/pcp/tmp/mmv on the Python code, so something appears to be broken on the binding.

This is on a RHEL6 box with latest packages (3.11.3 as of writing) built from scratch via Makepkgs and Python2.6. I can provide the full strace or example Python code if necessary.

[fche]

At least one bug evident in the python mmv bindings:

diff --git a/src/python/pcp/mmv.py b/src/python/pcp/mmv.py
-LIBPCP_MMV.mmv_lookup_value_desc.restype = pmAtomValue
+LIBPCP_MMV.mmv_lookup_value_desc.restype = POINTER(pmAtomValue)

which triggers a stack-smashing crash under glibc/ffi

[taintedkernel]

Actually see a few other issues as well, going to continue to work on this.

[natoscott]

Oops, meant to comment on close: "all should be resolved in pcp git master branch now ... lemme know if issues persist"