I am using a block step to get user confirmation before triggering the next step. I initially did not add any dependency for the block step but it turns out that a failed or cancelled previous step will disable/gray out the block button.
So I instead explicitly declare the dependency in the block step and also sets allow_dependency_failure to true. This works if the previous step fails (as in I can unblock and trigger the next step). But when the previous step was cancelled, the block button is still grayed out.
If the first step is cancelled, the block step, which depends on the first step, will not run because the dependency cannot be fulfilled.
I see you’re using allow_dependency_failure. While this option allows the block step to run if the dependent step fails, it doesn’t help if the job is cancelled. In cases of failure, allow_dependency_failure ensures you can still access the block step, but it does not cover cancellations.
If you are interested in a cleanup, i can suggest you use a pre-exit hook, which does that perfectly.
Hi @stephanie.atte , thanks for the pointer to documentation! From there I found that I actually need an input field instead, which does not create any implicit dependencies.