We have a CI pipeline with multiple stages (see example Pipeline · PETSc / petsc · GitLab). But jobs in a stage are not really dependant on jobs in previous stages. In a regular test, failures in one stage would prevent running of following stages (to save resources). This makes sense.
Previously, when I had a failed job, my workflow was: revise my code, run a pipeline, then immediately cancelle it, then click the specific job and run it, such that only this job in question is tested.
It worked before, but now it does not because Gitlab says the job depends on jobs in the previous stage that have not been run. It is a huge waste (and time consuming) for me to run all jobs before this job during debugging.
Could you improve that? Thanks