Hi Lewis, All good points. The reason I initially thought it would be good to put an rc on conda-forge would be to make it easy to process the dependencies for all python versions in travis.yml. However this is not a feasible option now that you mentioned that you can only put an rc on PyPI's test site, but the conda recipe gets the source distribution from the main site.
Fortunately I have found some alternatives since I wrote my last response I have investigated some better ways to integrate conda with travis CI while building for different python versions. For now I think the approach we should use is to have two separate text files listing dependencies, one each for Python 2 and 3. Of course it would be easier if we had the full recipe due to the ability to manage dependencies with preprocessing selectors (eg, specify in the comments of the recipe to ignore dependency X if it isn't python 3 compatible), but we can always change this in the future. For python 3 compatability, the current offending packages are pydap (which you are addressing separately) and esgf-pyclient / myproxyclient. The maintainer of the latter project has been responsive as far as my previous interactions have been concerned, so I think I'll see what I can do as far as getting it working with python 3. This probably can't be done in time for 1.2.0, but we can always update the conda recipes for those three packages on conda-forge separately. So tl;dr, I think it should be feasible to resolve CLIMATE-874 (and therefore, the CI issues) by the end of the week, and update the conda recipe from 1.1.0 to 1.2.0 only when we are ready to release it. Does this sound good? Thanks, Alex On Wed, Oct 19, 2016 at 9:35 PM, lewis john mcgibbney <lewi...@apache.org> wrote: > Hi Alex, > Responses inline > > On Wed, Oct 19, 2016 at 6:31 PM, <dev-digest-h...@climate.apache.org> > wrote: > > > > > From: "Goodman, Alexander (398K)" <alexander.good...@jpl.nasa.gov> > > To: "dev@climate.apache.org" <dev@climate.apache.org> > > Cc: > > Date: Wed, 19 Oct 2016 18:25:35 -0700 > > Subject: Re: Preparatory work for 1.2.0 Release > > Also, one thing I would like to add is that the dependency issues we are > > seeing with the CI tests have been happening since this commit > > > > https://github.com/apache/climate/commit/98020c9fe7654267d03 > > f8553f35af4ca9ed55952 > > > > Good catch! > > > > > > Based on the raw logs from the CI tests, it seems like the dependencies > are > > being downloaded but setuptools can't find them. We could revert this > > commit but that would be pointless anyway since we now also want to test > > for Python 3.x. > > > Yeah. I think that having CI 'unstable' for now is a reasonable temporary > sacrifice for us actually fixing the issue. > > > > Thus, solving this issue may be tied to CLIMATE-879 (put > > the conda recipe on conda-forge) and CLIMATE -874 (Remove Easy OCW and > > replace with a pure conda installation method). > > > All have been remarked for 1.2.0. > > > > Only thing to keep in mind > > is that if we go down this route, we won't be able to have a fully > > functional CI test-suite for Python 3.x until AFTER the 1.2.0 package is > > uploaded. > > > I just woke up... can you expand a bit? Why is this the case? > > Thus, I think we will need to do before officially releasing > > 1.2.0: > > > > 1) Get 1.1.0 on conda-forge [0] > > 2) Make ocw compatible with podaacpy 1.4.0 [1] > > > > Both of the above seem fine. > > > > 3) Put 1.2.0 release candidate on PyPI AND THEN conda-forge > > > > This is a bit tricky, the reason I say is that we can put a release > candidate into test.pypi.org but not on to pypi.org. The former is fine, > the latter would constitute a release... which is not what we are trying to > do unless we are actually releasing. > > > > 4) Update CI testing to download all packages via conda and fully > > deprecating Easy-OCW > > > > ack, however could this not be done prior to releasing 1.2.0? > > > > 5) If resolved, we are ready to release > > > > [0] https://github.com/conda-forge/staged-recipes/pull/1784 > > [1] https://github.com/apache/climate/pull/411 > > > > What do you think? > > > > > -- Alex Goodman Data Scientist I Science Data Modeling and Computing (398K) Jet Propulsion Laboratory California Institute of Technology Tel: +1-818-354-6012
