loops in CircleCI

[djarecka]

Summary

I'm not sure if for loops in config.yml change anything, I think it does not retry if the command fails. I mostly noticed this when I tried to use the same logic in the Nipype Tutorial repo, where datalad command fails quite often.

A simple example showing that CircleCI tries only once is here

[oesteban]

Seems like they added the -e flag to all the commands. Yep that particular loop is now a bit pointless.

[djarecka]

we have a few similar loops. do you know what is a proper way to implement retry logic in CircleCI? Since Docker doesn't have this mechanism, at least it would be nice to retry with CircleCI.

[effigies]

We could write a script to do something similar to travis_retry.