Issue using conda environment

[tritemio]

Summary

I tried the using conda environment with fretbursts to see if I can get rid of the mocking layer.

However, building the docs fails on RTD unless I reintroduce the mocking for scipy.

Details

• Project: http://readthedocs.org/projects/fretbursts/
• Build #:
  • Not Working: http://readthedocs.org/projects/fretbursts/builds/3813468/
  • Working: http://readthedocs.org/projects/fretbursts/builds/3813474/

Commit between "Working" and "Not Working":
tritemio/FRETBursts@d9e6421

[ericholscher]

This looks like an error on a C module, libgfortran.so.1, not on anything we can install with Python or Conda. We likely don't have a Fortran compiler on our build servers, and aren't planning to include one.

[tritemio]

@ericholscher, this is plain scipy. Are you saying that scipy cannot be installed on RTD? Conda packages are precompiled and should include all these binaries AFAIK.

[ericholscher]

Well, the install of the conda scipy seemed to work fine. I'm not sure why this particular library is missing. This seems more like a conda issue, or something in the way we're using it, which I'm not really able to debug.

[tritemio]

@ericholscher, yes seems a missing library on RTD that is required by scipy's conda package.

[Korijn]

Related: conda/conda#444

[jakirkham]

Yeah, this is definitely a conda issue. I believe they relinked some stuff to libgfortran.so.3 (no idea how they are doing versioning with this stuff), but some packages are still linked to libgfortran.so.1. This issue is related. ( conda/conda#2177 ) Though it couldn't hurt to open a new one with conda.

[agjohnson]

Closing the issue here, as this doesn't seem to be an issue with RTD. Feel free to reopen if there is anything actionable here.

[Korijn]

Something "actionable": is the pre-installed conda pinned to a specific version on RTD or do you always run conda update conda? Conda 4.x came out recently and it fixed a large number of important bugs related to package resolution.

[agjohnson]

Is this an issue of an out-dated conda? This doesn't seem to be an issue of a wrong conda version, as the conda ticket above notes that it was a packaging issue.

[jakirkham]

I don't think so. That is an orthogonal issue I believe. I have placed it in this issue ( #2076 ) and cc'd the two of you on the ticket.