CondaValueError: pip returned an error while creating the conda environment

[paugier]

Details

• Project URL: fluidsim
• Build URL (if applicable): https://readthedocs.org/projects/fluidsim/builds/6740246/
• Read the Docs username (if applicable): paugier

Expected Result

Building the doc with conda (as described in the files .readthedocs.yml and doc/environment.yml).

Actual Result

The build fails at the first command (creating the conda environment) with the very short error log:

  Running setup.py bdist_wheel for fluidfft: started

CondaValueError: pip returned an error.

See the failing build and a very similar succeeding build (Completed Nov. 22, 2017). Note that pip has to install exactly the same packages for the two builds (with the same pip version).

Of course I can create the environment at home with the commands:

hg clone https://bitbucket.org/fluiddyn/fluidsim
conda env create --name latest --file fluidsim/doc/environment.yml

[Tikam02]

Try running as default environment 2.7 or 3.4, without anaconda

[paugier]

Thank you for the advice, but I don't see how the documentation could be built without conda. The dependencies:

  - numpy
  - cython
  - h5py
  - sphinx
  - numpydoc
  - jupyter
  - matplotlib
  - fftw
  - pyfftw
  - mako

FFTW on the readthedocs server without conda?

For scientific code, it is a very nice feature of readthedocs to be able to work with conda. I don't see how I could use readthedocs without conda (or a similar distribution) for such type of codes.

[stsewd]

I tested on my local instance, seems like a problem with your repo visibility, I tried git clone https://bitbucket.org/fluiddyn/fluidsim on my terminal and also fails.

[paugier]

I think the repository has no problem. It is a Mercurial repository and Git can not handle Mercurial repositories (while it is possible to clone a Git repository with Mercurial).

Anyway, can you please try with

hg clone https://bitbucket.org/fluiddyn/fluidsim

[stsewd]

@paugier, sorry, my bad. I was able to pass the first step on my local instance, but increasing the memory and time limits. So I think that's the problem here (memory limit exceeded).

[paugier]

Thank you for the hint. Do you know what are the memory and time limits for the "production" instance?

[stsewd]

According to https://docs.readthedocs.io/en/latest/builds.html#build-process 1GB and 15 minutes.

[paugier]

Thanks a lot @stsewd for your time and your useful hint. I manage to decrease the memory usage during the build and the new version of the documentation is now available!

Of course, it could be nice to get an explicit warning when the memory or time limits are exceeded.

[RichardLitt]

Of course, it could be nice to get an explicit warning when the memory or time limits are exceeded.

This would be good, I think. Any chance you'd like to work on a PR for this? If not, let's flag open a new ticket for this.

[stsewd]

@RichardLitt I create a new issue for this (#3613).
@paugier glad I was helpful here :)

[mrconway]

I've been getting this error for the past couple of months without any change in the build procedure:

https://readthedocs.org/projects/alphapy/builds/7177424/

Could be a memory limit issue as well... the process just stops in the middle of downloading the Python packages.

[RichardLitt]

@mrconway Could you open a new issue and reference this one for that? This should have been resolved, so we should look into your error separately.