PYTHONPATH anomaly

PYTHONPATH anomaly

[Hartmut Goebel]

Hi,

Introduction:

today I kept struggling with obscure failing tests with a python
package: some of the test-files kept failing to import the very package,
while others succedes. It figured out that one of the test-files did
manipulate sys.path, so the test-files loaded earlier did raise the
import error and those loaded afterwards passed.

Trying to reproduce this in a venv, did not succeed. Say: the test pass.

As reason for this difference I discovered that in guix the current path
is not part of sys.path, while in the venv it is.

Questions:

1) Is it intended that the current dir is not part of sys.path?

2) Where is this patched into?

3) Why does this problem not occur when running "setup.py test"

4) Related: Shouldn't "build/lib" (resp. the related one for extension
packages) be preppended to sys.path? As this is where the build python
files are.


-- 
Regards
Hartmut Goebel

| Hartmut Goebel          | h.goebel@crazy-compilers.com               |
| www.crazy-compilers.com | compilers which you thought are impossible |

Re: PYTHONPATH anomaly

[Ludovic Courtès]

Hello!

Hartmut Goebel <h.goebel@crazy-compilers.com> skribis:

> Questions:
>
> 1) Is it intended that the current dir is not part of sys.path?

I don’t know about the specifics of Python here, but it’s always a
security problem to put “.” in a search path: users could end up
arbitrary code that just happens to be in $PWD.

So it’s probably intended.

Ludo’.