News for hackers. Stuff that matters.
Re:GitFlow considered harmful by brento on Friday June 19, 2015 @09:53AM (#9744707)
There is nothing wrong with rebasing a feature branch imho. Feature branches should be considered ephemeral. But it probably depends on your team and project size.
Re:GitFlow considered harmful by sytse on Friday June 19, 2015 @10:04AM (#9744792)
My personal opinion is that it breaks history and CI tests for all the feature branches. But at GitLab we encountered customers that insisted on having a linear history after migrating from SVN. Therefore there is a function in the UI of GitLab EE to rebase a merge request when accepting the merge request. See http://doc.gitlab.com/ee/workflow/rebase_before_merge.html
Re:GitFlow considered harmful by solutionyogi on Friday June 19, 2015 @10:09AM (#9744838)
Could you expand on what do you mean by rebase breaking CI tests? I don't see how that's possible.
Re:GitFlow considered harmful by plorkyeran on Friday June 19, 2015 @10:27AM (#9744946)
The results from previous CI runs are no longer connected to the rebased commits, and most people don't have CI set up to retest each commit individually after a rebase.

I've never found this to be an actual problem in practice.

Re:GitFlow considered harmful by tehbeard on Friday June 19, 2015 @10:27AM (#9744947)
Since a rebase changes the commit sha, I assume that's used as a reference for which version of the code the CI test was run against.

That said, shouldn't CI be triggered after the rebase anyway to confirm all still works?

Re:GitFlow considered harmful by solutionyogi on Friday June 19, 2015 @10:36AM (#9745012)
I see. But as you noted, one must do CI tests again after a rebase.
Re:GitFlow considered harmful by sytse on Friday June 19, 2015 @10:54AM (#9745166)
Indeed you need to do them again. But you also have to rerun them after a merge anyway. The problem is that you can no longer see which commits that you merged where green before the merge. For example is very useful if the merge itself breaks the tests (uncommon but it can happen).
Re:GitFlow considered harmful by Mithaldu on Friday June 19, 2015 @11:04AM (#9745240)
Edit: There's a well-written solution for that here: https://news.ycombinator.com/item?id=9745367
Re:GitFlow considered harmful by sytse on Friday June 19, 2015 @12:56PM (#9746093)
Parent was deleted.
Re:GitFlow considered harmful by Mithaldu on Friday June 19, 2015 @01:35PM (#9746375)
Ugh dammit. Short version:

Instead of taking the branch <feature> and rebasing it, make a branch called <feature>.1 on top of <feature>, then rebase that, leaving <feature> in place.

That way all your references remain intact and you can compare CI results.

Increment the number as needed.

Re:GitFlow considered harmful by sytse on Monday June 22, 2015 @07:32AM (#9757553)
It solves the issue of the CI references. But it also seems a bit clumsy to me.
Re:GitFlow considered harmful by Mithaldu on Monday June 22, 2015 @04:02PM (#9760825)
It's all trade-offs. The trade-off being that you maintain history by having copies of a branch around, while the dude trying to fix a bug doesn't have to break out the vodka bottle upon arriving back home. :)
Re:GitFlow considered harmful by tokenizerrr on Friday June 19, 2015 @10:22AM (#9744927)
A CI test result is tied to a particular commit hash. If you change this commit hash the CI test result will no longer be tied to that one commit.