Normally the pipeline is triggered by push. There is the possibility to configure an 'on pull request' webhook, but how to connect it with a pipeline and disable or reconfigure the default 'Pipelines' webhook which triggers on push?
This has since been implemented as a feature configurable in bitbucket-pipelines.yml: https://confluence.atlassian.com/bitbucket/configure-bitbucket-pipelines-yml-792298910.html#Configurebitbucket-pipelines.yml-ci_pull-requests
No; it will be triggered when a PR is created or updated (the source branch of the PR is pushed to).
When it is merged, the relevant pipelines on the target branch will run, if they are configured.
For example, I configured my pipelines to run tests on pull requests, and on the master branch. So when I merge a pull request to master, the tests run on master because of the configuration of master, not because of the configuration of the pull requests.
Hi Brian, there's any way to have the source and destination branch?
In branch' triggered pull requests I have the BITBUCKET_BRANCH that allow me to execute some script in order to run some code, but not in PR triggered pipelines.
Edit: with this line you can have the origin branch
export BRANCH=$(git branch | grep \* | cut -d ' ' -f2)
Thank you. This is _very_ important for me. I run a team of 3 developers and contractors that float in-and-out. We manage a dozen repos and quality-checking is a big issue for us. In addition to manual code-reviews, making sure all tests run as the PR is created or updated is one less step I need to take.
@Brian Maissy In the documentation it reads "When it's triggered, we'll merge the destination branch into your working branch before it runs. If the merge fails we will stop the pipeline."
So is my understanding (below) correct:
If this is indeed the case, how do I execute steps on merge failure?
Is after-script applicable here?
@Yonel Meza Avila Unfortunately the configuration in bitbucket-pipelines.yml only supports defining pipelines to run on pull requests based on the source branch.
If you want to run based on the destination branch, you could run on all branches (by specifying '**') and then check the BITBUCKET_PR_DESTINATION_BRANCH variable. If it isn't the destination branch you are interested in, then just finish the run immediately without running anything else. It's not great, since your list of pipelines runs will be polluted by a lot of empty runs, but it's a possible workaround.
Yeah, I would love this feature. The ideal workflow in my mind is to have the build pass as a requirement before merging the Pull Request. This keeps your deployment branches clean. After the PR is merged, the current feature set is nice to watch master/development branches and deploy them as changes are made (ie PRs are merged).
Actually, never mind that - a search reveals no current issue open on the subject, so I created one:
please vote it if you'd like this feature.
So the idea here is that instead of running your pipeline on every push you run it on every PR creation?
Is it not more convenient to build and test your code ASAP?
Update: I have funnelled the feedback about the expense of up to date deploys back to the Bitbucket Pipelines team.
Exactly @Jan Steinke. Developers on any team should be encouraged to push their branches to remote often, even before raising a PR, without worrying about consuming build minutes on unnecessary builds. They can add [skip ci] to commit messages, but will often forget to do so.
Build everything = $$. Plus there may be times where I only want to perform certain actions on a PR but not a typical feature branch push. For example, when Joe is checking in code to remote feature branch for the weekend, I don't necessarily want to burn build time running static code analysis on incomplete code, but on a PR, I absolutely do.
At the moment it is not possible to trigger a pipeline on a pull request being created. However, pipelines run on the source branch will be displayed along with the pull request. If this is something you would like in the product please create a feature request at https://bitbucket.org/site/master/ telling us what you'd like in this feature.
Is it possible to rerun the pipeline of the merge when the pull-request is updated? We want to require at least two successful builds without failed builds. One for the last commit itself, and one for the potential merge.
What I've observed is that a change in the pull request will reset the "successful build counter" and will initiate a new build on the new commit (we have it setup to build on each commit), but will not trigger a new build on the potential merge. Is there currently a way to achieve that?
Nevermind, it seems that I mislabeled the branch that I want this to affect in my yml.
It now works. I can't seem to make it work for a specific branch. Is there a special format?
And none of these work.
I needed to trigger bitbucket pipeline for every created pull requests. As this is not available by default, I've managed to do it with a simple zapier zap, here is the tutorial :
Hello, based on documentation:
pull-requests - A special pipeline that only runs on pull requests initiated from within your repo. It merges the destination branch into your working branch before it runs. Pull requests from a forked repository don't trigger the pipeline. If the merge fails, the pipeline stops.
Is there any way so I can override merging my working branch with the destination branch ? I know it might be odd but I want to merge my working branch with a different branch than the destination.
I have a Jenkins pipeline to increase the version number of my artifacts automatically, what it does is:
My question is: Is there some way to indicate BitBucket if a Pull Request comes from a particular user (in this case Jenkins) to auto approve immediately or not request any approver to merge?
It would be, but when we start having to pay for minutes to build, I think I will want this feature too. That way I could setup pipelines to only run tests on the master branch, and if anyone creates a pull request to the master branch, those also get run through pipelines. This might help save some minutes.