Temporarily pin conda-build to 2.0.*

[jakirkham]

Encountering some strange test failures particularly on Python 2.7 since conda-build version 2.1.0 was released. This pins the conda-build 2 tests to 2.0.* to avoid these issues until we can properly address them. At present, these issues seem to only occur on the CIs and are not reproducible locally.

[jakirkham]

@nicoddemus , seems pinning conda-build to 2.0.* "fixes" the problem. Not exactly what I would want to do. Sadly the error message is so terse I'm still not really sure what the error is. Not to mention I have been unable to reproduce this locally.

[nicoddemus]

That's a bummer. Well, thanks for the info though. Let's see what we can discover in the next few days.

[isuruf]

Note that py36 selector support is added in conda-build 2.1, so we might want to backport that if conda-build 2.0.* is pinned

[ocefpaf]

What's is the failure? I am sure @msarahan would like to know; ;-)

…

On Jan 5, 2017 1:45 AM, "Isuru Fernando" ***@***.***> wrote: Note that py36 selector support is added in conda-build 2.1, so we might want to backport that if conda-build 2.0.* is pinned — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#410 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA6BL7naGaZeouJbmDlzOO0HDNmEmq7nks5rPHVwgaJpZM4La56A> .

[jakirkham]

Have added a null change commit above so the error can be inspected.

Unfortunately there is not a clear traceback leading to conda-build. It was only noticed by inspecting to recent builds. This was meant as a test of that. So I wasn't going to bring him in until we had more information.

TBH the error looked like an issue with imports between Python 2/3. The fact that it is confined to Python 2.7 makes it more strange. This suggests to me that it may not even be a conda-build issue, but a conda-smithy issue that we have uncovered in the conda-build upgrade. So I had tried PR ( #410 ) to use absolute imports everywhere. That appears to have changed where the error comes from, but not the fact that there is one.

Have been unable to reproduce this locally either on my Mac or Linux. In any event, I would not worry about impact to conda-forge services as they all use Python 3.5, which doesn't experience this problem. I'm not even sure this pinning should be used outside of this CI until it can actually be confirmed as a elsewhere.

[jakirkham]

After playing around with a few PRs, it does seem to be a conda-build issue after all. In particular, it is not installing enum34 on Python 2.7.

xref: conda/conda-build#1623

[jakirkham]

Would rather just install enum34 on Python 2 CI builds near term until it gets sorted. So have added PR ( #414 ).