We’re evaluating CircleCI to replace our current build system but I’ve run into a few oddities. I have a workflow, so that I can set a global context with an env var that’s required for “docker.auth”. This works fine when triggered from the workflow, but if I rerun a job that failed it immediately bombs out because the context env vars are now missing.
I would expect rerunning an individual job to retain whatever metadata/context it started with, not require me to go and find the workflow and rerun that elsewhere. At least give a warning, or block reruns of jobs triggered by workflows, as this took me a long time to figure out.
Can you just re-run the job from the workflow screen? I believe one gets the choice of re-running everything, or just a failed step in the workflow? I think that’s what I do, and it works for me.
Update: ah, I see you’re doing that. Fair enough! Perhaps you could add a step in the job to fail early if an env var is not present, so at least it fails quickly.
Update again: I see why you take your view, but running a job outside of the context of a workflow could be expected by some users not to inherit the properties of that workflow. I wonder if this behaviour should be a config file option, or an options panel tick?