Name "pkg" is not defined.

Related changeset: http://hg.python.org/cpython/rev/3987667bf98f#l2.89

Thanks for the report and patch.  Can you also provide a test that fails using the current code (and that passes with the patch applied)?

> Can you also provide a test that fails using the current code (and
> that passes with the patch applied)?

New patch attached with a test.

Chris, did you have a chance to review the patch I submitted?

The only usage of `iter_importers` function in stdlib is in
`iter_modules` function:

http://hg.python.org/cpython/file/cbdd6852a274/Lib/pkgutil.py#l139

Nick would be a better person to review the patch since IIRC he worked a bunch on that module (e.g. in the changeset you mentioned).  But if I get a chance I will take a look.

I've attached a new patch.

I don't think it's necessary to check for UnboundLocalError/NameError in the tests.

> I don't think it's necessary to check for UnboundLocalError/NameError
> in the tests.

Thanks for the review.

Attached a new patch.

Given the lack of proper tests for iter_importers, wouldn't adding a proper test that returns something useful, rather than testing only with a non-existent module be better (this test could be kept, maybe converted to an assertRaises)?

OK, I thought that iter_importers was at least being tested indirectly by the tests for walk_packages:
http://hg.python.org/cpython/file/default/Lib/test/test_runpy.py#l417

However, it turns out that iter_modules() only calls iter_importers() if you don't supply an explicit path, and walk_packages() always passes an explicit path down.

I suggest looking at the walk_packages test for ideas on devising a better functional test for iter_importers. Testing iter_modules directly would be good, too.

(Note: There's a series of patches from Chris to move the walk_packages test to where it belongs in test_pkgutil, but they won't be applied until after the final 3.2 maintenance release has happened. Moving the test requires some fairly significant rearranging of the test suite to get the required helper functions out of test_runpy and into a shared, discoverable, location)

Sorry for the delay and thanks for the suggestions Ezio and Nick.

> Given the lack of proper tests for iter_importers, wouldn't adding a
> proper test that returns something useful, rather than testing only
> with a non-existent module be better (this test could be kept, maybe
> converted to an assertRaises)?

I've added a test for iter_importers().

> (Note: There's a series of patches from Chris to move the walk_packages
> test to where it belongs in test_pkgutil, but they won't be applied until
> after the final 3.2 maintenance release has happened.)

Thanks for the info. These patches are in issue 15376, issue 15358,
issue 15403 and issue 15415, right?

Is there a chance to get this into 3.3.2?

New changeset c40b50d65a00 by Nick Coghlan in branch '3.3':
Close issue #16163: handle submodules in pkgutil.iter_importers
http://hg.python.org/cpython/rev/c40b50d65a00

New changeset 3bb5a8a4830e by Nick Coghlan in branch 'default':
Merge fix for #16163 from 3.3
http://hg.python.org/cpython/rev/3bb5a8a4830e