Skip redundant builds for identical SHAs?



A suggestion for a default setting or advanced option to not run tests again for an identical SHA that already has a passing result? Or that has a passing result in the last x days?

We often rename/play with branches pointing to identical commits, add a missing story identifier to a branch name, etc, and it seems silly not to be able to use the test result that passed if the commit literally has not changed…

I guess if it were master or a branch with deployment action you want to be able to take actions? So I can see some room for subtlety/confusion/etc, but curious if it could be offered at least when no deployment action would occur post test run, or if you could offer to skip the test run and only perform the deploy action if an existing result is green.

On a related note though, I am experiencing a pathological case where the same sha does not necessarily behave the same on differently named branches and identical ENV vars, with and without cache rebuilds, so, it would also be nice if things were predictable/consistent on the same code! I can’t imagine how I would acheive this behavior if I were trying to.


I agree. Often we have a feature branch that goes green and merge to master where we have to wait for it to go green again before deploying. It would be handy to have circle know that the HEAD has already gone green and skip specs and just run the deploy step.