Crash if no pipeline is run

[smarkets]

Crash if no pipeline is run

[aschmolck]

This looks like a bug. Note that --use-merge-strategy is experimental (and marked as such), and generally not super well behaved, we accepted an MR to add minimal support, but it's not very good yet and I just accepted another one (belatedly) that fixed a bad function call (#102, also see #72).

Meta question: Irrespective of its current state of implementation, are you sure you want to --use-merge-strategy in the first place? Note that you can still can have your source branches merged into master (rather than rebased) by just leaving the default MR strategy in gitlab active (which does that). You only need this options if it's important to you that updating the source branch to the latest commit in the target branch happens via merge rather than rebase (and I'm still not sure why one would want that).

[Jellby]

I guess the crash happens when a pipeline was run for the branch, but not for its latest commit (assert current_pipeline.sha == commit_sha), which can happen if pipelines are run manually or if the CI configuration changes. And the problem is having an assert with no exception handling.

Regarding the merge strategy, I may change my mind, but I don't think rewriting (i.e., rebasing) commits that are already pushed is a good idea. I rebase commits in my local branches, but once it's pushed, it's better not to touch it. (By the way, doesn't rebasing break any signature the commits may have?) I also have to deal with a community of developers who are not git masters and have already been trained to use merge. What marge-bot does should not affect them much, but if there is a conflict and when they are going to fix it the remote repository has rebased the commits that they may have further developed on in their branches... I'm sure it will cause confusion.