add apptools

[grlee77]

Add apptools as required by #285.

This PR will require both #391 and #394 to be merged first

[conda-forge-admin]

Hi! This is the friendly conda-forge-admin automated user.

I just wanted to let you know that I linted all conda-recipes in your PR (recipes/apptools) and found it was in an excellent condition.

[jakirkham]

Restarting the CIs here now that the dependencies are available.

[jakirkham]

Does this support Python 3 or is Python 2 only?

[patricksnape]

It should support Python 3 as well - the problem seems to be it isn't picking up the new pyface 3.x builds yet.

[jakirkham]

Good point. The Windows packages are not released yet. So, let's wait for that to happen. Will check back later and restart AppVeyor.

[grlee77]

Is it desirable to run the full package test suite rather than just the import tests done here? The tests themselves only take about a second to complete.

If so, I can update the test section to something like the following:

test:
  requires:
    - nose
    - numpy
    - pandas
    - pytables
  commands:
    - python -m nose.core -v apptools

(for this package, the tests have a number of additional requirements that are available through the default channel).

I could do the same for the traits-feedstock, etc. that were already merged

[jakirkham]

So, it is generally good to run tests. However, I have tried to use the following rule of thumb. If it is compiled, try to run the tests (or at least some subset that reasonably exercises the code). Particularly, try to do so for anything where numeric accuracy is important (e.g FFTW). If it is Python only and they do CI, I am far less concerned about testing and generally skip it. Importing a bunch of things tends to be just as effective at catching most of the errors you will see (missing dependencies or syntax issues).

[grlee77]

okay, let's leave this one as is then. Can someone restart the CI?

[jakirkham]

I think I had, but I don't think everything had been released yet. I have restarted the CIs again. Let's see how this goes.

[jakirkham]

Looks like there are some requirements issues that need to be addressed here. That is the cause of the CI failure.

[grlee77]

The Appveyor failure seems to be a time out, but the CirceCI one I need help with.
It looks like maybe it is trying to fetch the traitsui dependency from the default channel instead of conda-forge?

excerpts from the log:

The following looks okay, with conda-forge listed as the source for pyface & traitsui

The following NEW packages will be INSTALLED:

    apptools:   4.4.0-py27_0  file:///opt/conda/conda-bld/linux-64/
    configobj:  5.0.6-py27_0  defaults       
    .
    .
    . 
    pyface:     5.0.0-py27_0  conda-forge
    traitsui:   5.0.0-py27_0  conda-forge                             

but then this error occurs

conda.resolve.Unsatisfiable: The following specifications were found to be in conflict:
  - python 3.4*
  - traitsui -> pyface 4.3.0|4.4.0|4.5.2
  - traitsui -> python 2.7*

The default channel only constains a traitsui for Python 2.7, and that seems to be what it is finding. Is there a way to work around this?

The conda-forge packages all exist here:
https://ci.appveyor.com/project/conda-forge/traitsui-feedstock

[grlee77]

Okay, I think I see the problem is that traitsui is listed as failed at:
https://conda-forge.github.io/feedstocks.html

so I have to fix that one first. I thought it had passed before we merged it

[jakirkham]

Yeah, I think some of the pyface packages weren't released when the traitsui Linux build started. If you just restarted the traitsui build on Linux, you should be good to go.

[grlee77]

Okay, I fixed the Linux build for traitsui, so this is ready to try again.

also, both pyface and traitsui released v5.1.0 earlier today. I will wait a bit for them to show up on pypi before updating the feedstocks.

[jakirkham]

Fair enough. Restarted the CIs here to make sure everything now works.

[jakirkham]

Weird. Still seeing a failure on Windows, but only on Python 3.4 64-bit.

[grlee77]

I don't see any obvious error messages during the build, but it does say:

writing list of installed files to 'record.txt'
number of files: 0

which would explain why it fails to import

Any idea if the following might be related to the problem?

The succeeding cases seem to be building into C:\Miniconda-x64\envs\_build\Lib

but in the failing Python 3.4 64-bit build things are building in C:\Miniconda-x64\Lib\ (why no _build environment?)

[grlee77]

The only other thing that looks like an error occurs higher up in the log, but seems related to visual studio, which we aren't using in this recipe:

C:\Miniconda-x64\conda-bld\work\apptools-4.4.0>call "C:\Program Files\Microsoft SDKs\Windows\v7.1\Bin\SetEnv.cmd" /x64 /Release 
Setting SDK environment relative to C:\Program Files\Microsoft SDKs\Windows\v7.1\.
Targeting Windows 7 x64 Release

The specified configuration type is missing.  The tools for the
configuration might not be installed.

[jakirkham]

Yeah the fact that it is installing to C:\Miniconda-x64\Lib\ is very strange indeed. @patricksnape and @msarahan, I am totally out of my depth here. Is there any reason that on the Python 3.4 64-bit build the wrong install location (C:\Miniconda-x64\Lib\) is being used?

[jakirkham]

Workarounds in place for the Python 3.4 64-bit Windows issue. AppVeyor has been restarted.

[grlee77]

thanks. ready to merge?
once this one is available it will be possible to finish #396

[jakirkham]

Absolutely! Thanks for being so patient.

[jakirkham]

Well, we are not exactly sure why, but this is not getting converted to a feedstock correctly. See this issue ( #458 ).