Conda channel nightly updates not up-to-date

[durack1]

It seems that the nightly conda channel is not automagically updating.. It'd be great if this channel was available for testing the master branch changes before official releases are uploaded

[durack1]

@doutriaux1 just wondering whether this conda experience is what you'd expect:

(uvcdat)durack1@oceanonly:[LeyangFeng_CEDSEmissionsSO2]:[1251]> conda update uvcdat
Using Anaconda Cloud api site https://api.anaconda.org
Fetching package metadata: ....
# All requested packages already installed.
# packages in environment at /export/durack1/anaconda2/envs/uvcdat:
#
uvcdat                    2.4.1                    UVCDAT    uvcdat
(uvcdat)durack1@oceanonly:[LeyangFeng_CEDSEmissionsSO2]:[1251]> conda update uvcdat -c uvcdat
Using Anaconda Cloud api site https://api.anaconda.org
Fetching package metadata: ......
Solving package specifications: .........
Warning: 2 possible package resolutions (only showing differing packages):
  - hdf5-1.8.16-0.tar.bz2
  - hdf5-1.8.16-UVCDAT.tar.bz2

Package plan for installation in environment /export/durack1/anaconda2/envs/uvcdat:

The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    pyspharm-1.0.8             |           UVCDAT         695 KB
    libcdms-2.6                |           UVCDAT         1.9 MB
    uvcdat-2.6                 |           UVCDAT           2 KB
    ------------------------------------------------------------
                                           Total:         2.6 MB

The following NEW packages will be INSTALLED:

    basemap:  1.0.7-np19py27_0
    cycler:   0.10.0-py27_0
    geos:     3.3.3-0
    pyspharm: 1.0.8-UVCDAT

The following packages will be UPDATED:

    uvcdat:   2.4.1-UVCDAT     --> 2.6-UVCDAT

The following packages will be DOWNGRADED:

    libcdms:  2016.5.24-UVCDAT --> 2.6-UVCDAT

Proceed ([y]/n)?

[doutriaux1]

@durack1 you should use environments. The nightly are purposely versioned with the date so that they superseed any existing version. "upgrading" to a stable version hence is not possible.

[durack1]

@doutriaux1 I believe this is still not functioning - the latest master PR merge is dated 5 days ago, whereas the UV-CDAT conda channel nightly is showing 2016.6.27 as latest

[durack1]

@zshaheen @doutriaux1 @dnadeau4 I'm just trying to update my uvcdatNightly conda environment and it seems that the conda nightly UV-CDAT channel no longer exists..

https://anaconda.org/uvcdat/uvcdat/labels - only main

Am I doing something wrong here, or have you changed things?

[dnadeau4]

It seems that uvcdat/uvcdat is not updated since it has no build.sh. Should we create a nightly uvcdat package so that conda install uvcdat -c uvcdat/label/nightly update to the latest?

[zshaheen]

@durack1 @dnadeau4 I don't think the nightly building from Travis that's in PMP is implemented in UV-CDAT. I'm looking here.

[durack1]

@zshaheen that's what I thought.. But the weird thing is the last time I installed UV-CDAT I was directed to a nightly channel that has since been deleted..

[doutriaux1]

the nightly channel exists for each individual channels, not for uvcdat. You pick up each individual packages nightlies by passing the nighlty channel . You do not need a uvcdat nightly
conda install uvcdat -c uvcdat/label/nightly is bad by the way @dnadeau4 we are trying hard to have user use environments exclusively
to get nighlties you need:
conda create -n NAME -c uvcdat/label/nightly -c uvcdat uvcdat

[doutriaux1]

with that said we need nightlies for pmp , vcs and cdms2 do not seem to happen automagically. @zshaheen do you mind taking a look?

[dnadeau4]

@doutriaux1 I though @durack1 wanted to update a package into an existing environment. In that case you need install not create.

[doutriaux1]

@dnadeau4 I see in this case let's post:
conda install -n MYENV package -c channel

[dnadeau4]

@doutriaux1 sounds good!

[zshaheen]

@doutriaux1 I though pmp was working? Did you want to get vcs and cdms2 working as well?

[doutriaux1]

i got cdms2 to work for sure will push vcs soon.

[durack1]

@doutriaux1 @zshaheen it seems that uvcdat nightlies are having problems:

public1:/home/p1d/> conda create -n UVCDATNightly -c conda-forge -c uvcdat/label/nightly uvcdat
Fetching package metadata .............

PackageNotFoundError: Package not found: Conda could not find '

This is on a redhat6 machine:

> uname -a
Linux public1 2.6.32-642.6.1.el6.x86_64 #1 SMP Thu Aug 25 12:42:19 EDT 2016 x86_64 x86_64 x86_64 GNU/Linux
> more /etc/redhat-release
Red Hat Enterprise Linux Workstation release 6.8 (Santiago)

[doutriaux1]

@durack1 I think it's the lab that doesn't let you out. Did you see: https://github.com/UV-CDAT/uvcdat/wiki/Install-using-Anaconda and also did you click on the lab banner before going out to the internet?

[zshaheen]

FYI, this may be an issue on Anaconda's side, since we're getting a truncated error message. If I remember correctly, the PackageNotFoundError was more verbose.

conda install uses conda search to find the packages. conda search currently doesn't work for me.

[zshaheen]

I also get the same error message as @durack1 on aims2, which is running RHEL 7.

[durack1]

@doutriaux1 this is happening at GFDL, so nothing to do with the LLNL cert/authentication stuff

[durack1]

@doutriaux1 @zshaheen if you want to test things out you can use:

conda create -n UVCDATNightly --verbose --verbose --verbose -c conda-forge -c uvcdat/label/nightly uvcdat

You should get a similar error to me..

[durack1]

@doutriaux1 @zshaheen, my question then, is this an anaconda issue, or are one of the UV-CDAT dependencies not there, and so things fail

[doutriaux1]

ok you forgot a -c uvcdat so it cant find some package that are on the uvcdat channel only (like uvcdat for instance)

You can pester conda team about having a better message

bottom line you want:

conda create -n UVCDATNightly --verbose --verbose --verbose -c conda-forge -c uvcdat/label/nightly -c uvcdat uvcdat

[durack1]

@doutriaux1 can you label this one "user error" please? :)