Conda build

[doutriaux1]

@aashish24 @sankhesh @danlipsa @chaosphere2112 @dnadeau4

This is ready and passes on my computers. But let's carefully review baselines changes at: CDAT/uvcdat-testdata#142

[aashish24]

@doutriaux1 I will have a detailed look at it today.

[sankhesh]

Buildbot dashboard

Gitlab

CDash

[aashish24]

@sankhesh looks like conda is missing on the shell.

[sankhesh]

@aashish24 Yes, it was missing for the earlier builds.

Look at the latest one.

[sankhesh]

@doutriaux1 What is the path to conda on crunchy?

[aashish24]

thanks @sankhesh On karma right? its building now.

[doutriaux1]

@sankhesh it should be in my ${HOME}/anaconda2/bin/conda

[chaosphere2112]

@doutriaux1 Building now

[doutriaux1]

i merge master in, I'm updating the baselines for you vcsaddons.

[chaosphere2112]

@doutriaux1 @aashish24 Passes tests for me, I'm going to merge.

[stefraynaud]

I successfully installed uvcdat using conda on a supercomputer where I haven't been able to do it for years using cmake. So will the conda system be the prefered way to install uvcdat in the future?

[williams13]

Yes.

Best regards,
Dean

From: Stephane Raynaud <[email protected]mailto:[email protected]>
Reply-To: UV-CDAT/uvcdat <[email protected]mailto:[email protected]>
Date: Wednesday, June 22, 2016 at 2:31 AM
To: UV-CDAT/uvcdat <[email protected]mailto:[email protected]>
Subject: Re: [UV-CDAT/uvcdat] Conda build (#2015)

I successfully installed uvcdat using conda on a supercomputer where I haven't been able to do it for years using cmake. So will the conda system be the prefered way to install uvcdat in the future?

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHubhttps://github.com//pull/2015#issuecomment-227691435, or mute the threadhttps://github.com/notifications/unsubscribe/ACrXKT3ZrV07i3SnMpjXrmpb_tuGBeZkks5qOQDWgaJpZM4IyPZ2.

[aashish24]

I successfully installed uvcdat using conda on a supercomputer where I haven't been able to do it for years using cmake. So will the conda system be the prefered way to install uvcdat in the future?

@stefraynaud cheers 🎉 yes, conda is the preferred way for installation for the end-users. Also I would like to point out that conda is a package management system and cmake is the build system. Underneath conda, you are still using make and cmake for individual packages.

I am curious if you were able to use system MPI for your supercomputer?

[aashish24]

I am curious if you were able to use system MPI for your supercomputer?

If not, I am wondering if you can point mpi4py to use system MPI. It should be possible but may require some manual work.

[doutriaux1]

@aashish24 mpi4py states:

Install

Once you have a working MPI implementation and the mpicc compiler wrapper is on your search path, you can install this package

so for system mpi you probably will need to install mpi4py in your conda environment after the fact.

[doutriaux1]

we will be providing builds against openmpi though.

[aashish24]

@aashish24 mpi4py states:

@doutriaux1 that's what I was thinking too but this would be manual process as I stated.