[pyemma] bump version 2.3.1

[marscher]

No description provided.

[marscher]

The Appveyor queue for Roberts account seems stuck. I've reported this in support forums. I hope this will be fixed soon.

[jchodera]

Looks like we're running into another conda-build API issue:

Traceback (most recent call last):
  File "/anaconda/bin/conda-build", line 6, in <module>
    sys.exit(conda_build.cli.main_build.main())
  File "/anaconda/lib/python3.5/site-packages/conda_build/cli/main_build.py", line 322, in main
    execute(sys.argv[1:])
  File "/anaconda/lib/python3.5/site-packages/conda_build/cli/main_build.py", line 313, in execute
    noverify=args.no_verify)
  File "/anaconda/lib/python3.5/site-packages/conda_build/api.py", line 97, in build
    need_source_download=need_source_download, config=config)
  File "/anaconda/lib/python3.5/site-packages/conda_build/build.py", line 1466, in build_tree
    config=config)
  File "/anaconda/lib/python3.5/site-packages/conda_build/render.py", line 157, in render_recipe
    config=config)
  File "/anaconda/lib/python3.5/site-packages/conda_build/render.py", line 81, in parse_or_try_download
    if (force_download or (not no_download_source and metadata.needs_source_for_render)):
  File "/anaconda/lib/python3.5/site-packages/conda_build/metadata.py", line 832, in needs_source_for_render
    return self.uses_vcs_in_meta or self.uses_setup_py_in_meta or self.uses_regex_in_meta
  File "/anaconda/lib/python3.5/site-packages/conda_build/metadata.py", line 851, in uses_vcs_in_meta
    metayaml = f.read()
  File "/anaconda/lib/python3.5/encodings/ascii.py", line 26, in decode
    return codecs.ascii_decode(input, self.errors)[0]
UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 1107: ordinal not in range(128)

[marscher]

mdtraj/geometry/src/_geometry.cpp(5047) : error C3861: 'roundf': identifier not found

[rmcgibbo]

1. why does compiling pyemma also require recompiling mdtraj? don't we have binary packages for mdtraj?
2. has the roundf bug(?) been fixed upstream?

[marscher]

1. It doesn't. But the win binaries for py27 are still missing.
2. I do not know. Fix geometry compilation for Windows mdtraj/mdtraj#1101 is contained in 1.8 I'd guess

[marscher]

The function is only defined for _MSC_VER < 1900, but for this compiler it should be included (vc9).
Since this error occurs in the Cython generated file, I guess the import within the pyx is not working as expected. It even failed, when I patched away the version check, so the definition should be there.

[marscher]

mdtraj dropped support for python2.7 on windows like 8 months ago.

@franknoe: How should we deal with this? I strongly suggest to require mdtraj>=1.8, because of the NaN geometry bug fixed in this version. We can unpin it for windows users and print a big fat warning, suggesting to upgrade to python3.

mdtraj/mdtraj@ed28733

[mpharrigan]

What about patching out the tng extension for window < py35

[franknoe]

We should discuss in Omnia how to generally handle this. We cannot simply stop Python 2 support without notice, so we have to find a strategy to deal with this in a way that's not destructive for users. Am 07/02/17 um 01:10 schrieb Martin K. Scherer:

…

mdtraj dropped support for python2.7 on windows like 8 months ago. @franknoe <https://github.com/franknoe>: How should we deal with this? I strongly suggest to require mdtraj>=1.8, because of the NaN geometry bug fixed in this version. We can unpin it for windows users and print a big fat warning, suggesting to upgrade to python3. ***@***.*** <mdtraj/mdtraj@ed28733> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#696 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGMeQo_BKssMykMNSTec_CpFzHxqJgBPks5rZ7Z9gaJpZM4L4ROi>.

--

---------------------------------------------- Prof. Dr. Frank Noe Head of Computational Molecular Biology group Freie Universitaet Berlin Phone: (+49) (0)30 838 75354 Web: research.franknoe.de Mail: Arnimallee 6, 14195 Berlin, Germany ----------------------------------------------

[marscher]

On 07.02.2017 01:20, Matthew Harrigan wrote: What about patching out the tng extension for window < py35

I've already fixed tng for this configuration. Now the compiler does not find roundf.

[jchodera]

We should discuss in Omnia how to generally handle this. We cannot simply stop Python 2 support without notice, so we have to find a strategy to deal with this in a way that's not destructive for users.

Agreed that we should discuss this. We tried to get a discussion of this going in July 2016 in this thread, but there was never enough momentum to actually commit to deprecating python 2.x. MDtraj decided unilaterally to deprecate python 2.x, so now we're stuck with this issue.

[jchodera]

Is there a possibility of getting the mdtraj developers to do a bugfix release for the last mdtraj release to support python 2.7? That seems like a reasonable burden to carry given the decision to drop python 2.x support.

[marscher]

I think they dropped the support because of all the hassle of the lacking/partial C99 support of the old M$ compiler. However with a few patches/workarounds this should be easy to support. I don't think mdtraj makes use of new py3k exclusive language features, do they?

[marscher]

inlining the definition works great. So now only parmed and solvationkit fails for py27, IMHO this mergable now.

[jchodera]

Woohoo! Thanks!

[mpharrigan]

@jchodera it's py27 on windows

[marscher]

On 07.02.2017 09:57, Frank Noe wrote: We should discuss in Omnia how to generally handle this. We cannot simply stop Python 2 support without notice, so we have to find a strategy to deal with this in a way that's not destructive for users.

I think we've agreed at the Omnia meeting to continue the support for 2.7 to have a smooth transition to Python3. We have to encourage users (omni-present) to upgrade soon. Since I've fixed the build problems for this version/platform in mdtraj, I could imagine, that @mpharrigan could be willing to re-include 2.7 in the testing matrix. The packages for the recent release are already online.