Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

conda_build result in ci doesn't run on google colab #15

Open
aymanhab opened this issue Jun 26, 2022 · 6 comments
Open

conda_build result in ci doesn't run on google colab #15

aymanhab opened this issue Jun 26, 2022 · 6 comments

Comments

@aymanhab
Copy link
Member

Comparing older versions that worked out of the box to new problematic builds shows:
new info/about.json
"conda_build_version": "3.21.9",
"conda_private": false,
"conda_version": "4.13.0"
vs. old info/about.json
"conda_build_version": "3.21.8",
"conda_private": false,
"conda_version": "4.12.0",

Maybe we can force conda version and conda build version on ci

@aymanhab
Copy link
Member Author

Possibly related
info/index.json contains
"depends": [
"libgcc-ng >=11.2.0",
"libstdcxx-ng >=11.2.0",
"numpy",
"openblas",
"python"
],
as opposed to
"depends": [
"libgcc-ng >=9.3.0",
"libstdcxx-ng >=9.3.0",
"numpy",
"openblas",
"python"
],

@aymanhab
Copy link
Member Author

@AlbertoCasasOrtiz thanks for the deep investigation into this. My sense is that we control the environment which we use to create conda builds, if we can't do it on ci we can do it in a local environment for now until this gets resolved rather than burden clients with complicated maneuvers especially considering target audience which uses conda exclusively to avoid these low level complications.

@aymanhab
Copy link
Member Author

aymanhab commented Jul 1, 2022

@moorepants Sorry to pull you into this but I'm having hard time understanding how conda resolves dependencies or particularly why it decides to upgrade libstdcxx-ng Any pointers to help troubleshoot would be greatly appreciated. Thank you

@aymanhab
Copy link
Member Author

aymanhab commented Jul 1, 2022

@hmok do you or your collaborators have experience troubleshooting this issue (latest build of opensim 4.4 pulls python38 package, auto updates dependencies then fails due to missing dependency) we have a workaround but it feels cumbersome to users. Is there a reliable setting to make sure conda packages load without error into jupyter? Thank you

@hmok
Copy link

hmok commented Jul 3, 2022

Thanks, @aymanhab for the question. I remember we had similar issues and had to stick to "python 3.7.7" for building OpenSim v4.3.

Below is the note from @ESJiang. Arthur, I wonder if you have any solution for Ayman's questions here, please. Thank you.

"_Note: It is advisable to create a virtual environment that supports python 3.7.7 before moving to the steps below. Although OpenSim 4.3 now supports Python 3.8 in default, there are many unexpected issues. You can run [python -V] to check its version. If it is not 3.7.7, you can run [conda install python=3.7.7] and recheck it._"

@moorepants
Copy link

@moorepants Sorry to pull you into this but I'm having hard time understanding how conda resolves dependencies or particularly why it decides to upgrade libstdcxx-ng Any pointers to help troubleshoot would be greatly appreciated. Thank you

conda's dependency solver looks for the most recent versions of all software in the environment that give a compatible set of versions. The conda tree tool or mamba's query tool can show the dependency chains for investigative work.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants