numpy dependency fails to installl in a python3 environment under tox

Issue #74 on hold
Former user created an issue

There was a recent issue having to do with virtualenv 1.8.3 upgrading it's bundled distribute to 0.6.30 which triggered a build failure in numpy due to some legacy code. see It manifests as an exceptin complaining of a "detached stdout"

virtualenv 1.8.4 was released, with distribute 0.6.31 which fixes the issue, and now I can create a virtualenv py3 environment and install numpy.

however using tox -e "py32" with a package that installs numpy still produces the "detach" error. I've upgraded my system distribute and virtualenv to 0.6.32 and 1.8.4 respectively, pulled the latest 1.4.2 tox from pypi, removed the .tox directory and recreated the environments and am still getting the error, which makes tox unusable for me. If I manually activate the py32 venv created by tox and use pip install numpy the setup succeeds.

Comments (4)

  1. Holger Krekel repo owner

    Could you produce a pointer to a (little if possible) failing example project with the exact directory you are in and the invocation you are using?

  2. oren kahill

    #!/usr/bin/env python
    from distutils.core import setup


    envlist = py32
    deps =
        numpy == 1.6.2
    changedir = {envdir}
    commands =
        python -c "import numpy; print numpy.__version__"


    from the shell:

    $ ls -1
    $ tox -e "py32"
       status, output = exec_command(version_cmd,use_tee=0)
        if _with_python and (0 or sys.__stdout__.fileno()==-1):
    ValueError: underlying buffer has been detached
    Command python egg_info failed with error code 1 in /home/okahill/b/.tox/py32/build/numpy
    Storing complete log in /home/okahill/.pip/pip.log
    ERROR: could not install deps [numpy == 1.6.2]
    ______________________________________________ summary _______________________________________________
    ERROR:   py32: could not install deps [numpy == 1.6.2]

    which is the signature error for the problem described above.

  3. Log in to comment