[TIP] [tox] Stale pyc files causing false positive test successes

holger krekel holger at merlinux.eu
Mon Oct 15 23:49:21 PDT 2012

Hey Laurens,

On Sun, Oct 14, 2012 at 17:39 +0200, Laurens Van Houtven wrote:
> Hi guys,
> I'm generally a pretty happy tox user, but recently I ran into something
> that appears to have generated a false positive. Apparently I misunderstand
> how tox works, because I figured this wouldn't be possible...
> Anyway, I removed a package (both the directory, the __init__.py and the
> module inside of it), re-ran the tests, 

Did you re-run tox or tests directly in your own venv?
Tox would actually sdist-package your project and i assume this
should usually fail if a package is missing.  Or did you also remove
it from setup.py?

> and it worked fine. Later, when
> actually trying to run the code in the venv, it broke -- 

you invoked some .tox/*/bin/... i assume.

> there was still
> something relying on importing that package but the tests didn't catch it.

If you used tox above then that is indeed strange.

> I honestly have no idea whatsoever how this happened. It's the first time
> I've seen this behavior, despite it not being the first package I pruned.
> I guess the way to fix this is by recreating the environment? That seems to
> have fixed the issue. I guess maybe in doing so I've also nuked any
> possibility of further diagnosing the problem; hopefully the above
> recollection is enough to ring a bell somewhere.

--recreate is a good idea for a "does it really work" check 
when deleting files / dirs i think because without it we rely on 
distribute/distutils mechanics to do the proper thing when 
re-installing a package including pyc file handling.  For my quick
experiment it worked fine, however.


More information about the testing-in-python mailing list