BLD: Pinned cftime 1.0.1

[cpelley]

#3239, #3215, #3225

[rcomer]

My user code is failing with cftime version 1.0.0b1 so I am not sure if <1.0.2.1 will solve all problems (although getting Travis checks to pass is obviously desirable!) What version would Travis be picking up here? I'm sure that info must be in the output somewhere but I can't see it!

[cpelley]

Travis is using cftime 1.0.1 (conda-forge)

[pp-mo]

cftime version 1.0.0b1 ... What version would Travis be picking up here? I'm sure that info must be in the output somewhere but I can't see it!

You can find this from a Travis log :
To view a log, first click on the tests icon by the commit in Github (green tick or red cross!),
navigate to the "continuous-integration/travis-ci/pr ... Details".
On the test page, hit another tick (or cross) for the log of one testing "phase"
(e.g. python 2.7, TEST_TARGET=default, TEST_MINIMAL=True).

In the log, look for the line "$ conda list -n $ENV_NAME ....",
the initial wedge ">" opens the collapsed command output ...
which is the list of installed packages !!

In this case, it has "cftime 1.0.1"

[rcomer]

Thanks both. I hadn't even twigged that there was collapsed output in the Travis log!

[bjlittle]

@cpelley Why not pin directly to 1.0.1, just to make this explicit and clear.

We still need to fix iris to deal with 0 based dates, which are now illegal in the latest cftime, and hence the root cause of current master failures.

So this pin would only be a temporary pin...

[cpelley]

...Why not pin directly...

Sure.

We still need to fix iris to deal with 0 based dates...

Agreed. Some relating discussion in #3239

[bjlittle]

@cpelley Apologies. Sorry to be a pain...

But could you please target the 2.2.x branch instead of master, and then I'll merge back the 2.2.x branch to master...

[cpelley]

NP :)

[bjlittle]

@cpelley Surely cftime==1.0.1 is what you mean, right? Doesn't that work instead of >=1.0.1,<=1.0.1...

[cpelley]

👍

[bjlittle]

Thanks @cpelley 👍

Ping @pp-mo