Raise import error on module import failure to avoid segfault

[overstre]

A dependency of confluent-kafka, enum34, published a broken version 1.1.8 that caused a segfault in confluent-kafka python library when accessing the module confluent_kafka.admin. For example, the list_topics call will try to import this module in function cfl_PyObject_lookup.

This results in module pointer being nil in the return of this call:
PyObject *module = PyImport_ImportModule(modulename);

and then it gets accessed here
obj = PyObject_GetAttrString(module, typename);
which results in a segfault:

Thread 28 "python" received signal SIGSEGV, Segmentation fault.
--
[Switching to Thread 0x7fff96ffd700 (LWP 8778)]
0x00000000004efa6b in PyObject_GetAttrString ()
(gdb) backtrace
#0  0x00000000004efa6b in PyObject_GetAttrString ()
#1  0x00007ffff433dae2 in cfl_PyObject_lookup (modulename=modulename@entry=0x7ffff434a067 "confluent_kafka.admin",
typename=typename@entry=0x7ffff434a057 "ClusterMetadata") at confluent_kafka/src/confluent_kafka.c:1962
#2  0x00007ffff43418a0 in c_metadata_to_py (self=0x7ffff0127440, metadata=0x7fff84002570) at confluent_kafka/src/Metadata.c:267
#3  list_topics (self=0x7ffff0127440, args=<optimized out>, kwargs=<optimized out>) at confluent_kafka/src/Metadata.c:382

This pull request is to check the value of module to ensure that it's not null otherwise return a runtime exception. This will make debugging this issue in the future.

Without patch

$ python ~/list.py
Segmentation fault

with patch:

$ python ~/list.py
Traceback (most recent call last):
  File "/home/brian/list.py", line 5, in <module>
    print producer.list_topics(topic=topic).topics.values()
RuntimeError: Module confluent_kafka.admin import failed.

[ghost]

It looks like @overstre hasn't signed our Contributor License Agreement, yet.

The purpose of a CLA is to ensure that the guardian of a project's outputs has the necessary ownership or grants of rights over all contributions to allow them to distribute under the chosen licence.
Wikipedia

You can read and sign our full Contributor License Agreement here.

Once you've signed reply with [clabot:check] to prove it.

Appreciation of efforts,

clabot

[overstre]

[clabot:check]

[ghost]

@confluentinc It looks like @overstre just signed our Contributor License Agreement. 👍

Always at your service,

clabot

[rnpridgeon]

What are your thoughts on removing the enum34 dependency all together. I'm actually not a huge fan of this dependency and we can easily achieve the desired functionality with constants as we do with logical offsets.

[overstre]

I'm in favor of reducing dependencies and keeping the style consistent.

[edenhill]

Thanks for finding this, but I think the initial intent was for the if-statement above to handle this, could you change that one instead to check for module rather than modulename, and remove the first %s and modulename from the argument list.

Also, change it to PyExc_ImportError.

[overstre]

Thanks for the review. I've updated the code.

brian@dev-brian-1:~/code/confluent-kafka-python$ python ~/list.py
Traceback (most recent call last):
  File "/home/brian/list.py", line 5, in <module>
    print producer.list_topics(topic=topic).topics.values()
ImportError: Module not found when looking up confluent_kafka.admin.ClusterMetadata