jenkinsci
fix: don't trigger poll unless target ref matches
The existing webhook handling code would trigger the GitSCM poll for the default branch regardless of whether or not the target ref of the webhook push event matched the branch specifier of the Job definition. This is problematic and often leads to duplicate builds for the same branch+revision if pushes to other branches happened to occur whilst build(s) were already in progress for that branch. This is because the GitSCM trigger just compares the HEAD revision of any branch matching the branch specifier against the last completed build of that branch (ignoring in-progress builds) and schedules a new build for it.
The code for this PR is mostly a rebased version of the original changes proposed by @aserrallerios under https://github.com/jenkinsci/github-plugin/pull/44 with some small refinements.
- [x] Make sure you are opening from a topic/feature/bugfix branch (right side) and not your main branch!
- [x] Ensure that the pull request title represents the desired changelog entry
- [x] Please describe what you did
- [x] Link to relevant issues in GitHub or Jira
- [x] Link to relevant pull requests, esp. upstream and downstream changes
- [x] Ensure you have provided tests - that demonstrates feature works or fixes the issue
Note: there is likely still discussion and changes needed here, but I wanted to open a PR to re-open discussions as the old one had long been abandoned and closed.
@KostyaSha would you be willing to take a look over this PR? It does fix the problem and has been running in our CI for a while now
@oleg-nenashev have you taken over as a maintainer on github-plugin now, please could you take a look at this PR?
@KostyaSha would you be willing to take a look over this PR? It does fix the problem and has been running in our CI for a while now
That looks like out of scope of current functionality, current functionality is only kick poll. For native branch triggering there is https://github.com/KostyaSha/github-integration-plugin which deals with exact branches and doesn't use gitscm.
As per the original description from 2 years ago, the intention of this PR is not to provide native branch triggering, the intention is to prevent pushes to unrelated branches triggering a build.
i.e., if the Jenkins build is configured to build the default branch of the repo (e.g., normally master or main) then the build should only be triggered when a GitHub push happens to that branch, not when one happens to other branches within the repository.
I don't know why this PR is stuck but this functionality is definitely needed. Recently we started seeing infinite build loops in our website builds due to GitHub SCM Polling triggering both when master is pushed and when gh-pages is pushed by the job itself, resulting in an infinite loop.
@KostyaSha this definitely is an essential functionality, and it does seem to me like it used to work, since this infinite loop behavior is something I only started observing in our setup a few months ago. I don't know what changed but this PR certainly appears to be fixing it. Please reconsider, saying that this is "out of scope" is a bit silly to me, and it should to you if you think about it a bit.
If I'm configuring my branch to fetch master branch and enable "GitHub hook trigger for GITScm polling" I obviously intend the builds to run for that branch, and not any branch.
This issue appears to be related:
- https://github.com/KostyaSha/github-integration-plugin/issues/388
@jakubgs that other plugin doesn’t use this plugin, so that linked issue isn’t related
But yes, this PR would prevent your gh-pages pushed from triggering a new master branch build (causing endless loop). We’ve run the .hpi from the build of this PR for more than 2 years now (to fix the bug) without issue and the change itself is very small and is fixing a bug not adding a feature, so I’m not sure why it couldnt be reviewed and merged
Thanks for the contribution man. I will try it, because just yesterday Jenkins committed 6000 times to our gh-pages branch for one website. I don't understand what changed since this used to work, but it makes no sense. Git poll log states:
> git ls-remote -h -- [email protected]:status-im/infra-bi.git # timeout=10
Found 15 remote heads on [email protected]:status-im/infra-bi.git
[poll] Latest remote head revision on refs/heads/master is: d88fecd7ef67995f14d7f2163a01dc5decabe5ae - already built by 615
So the build was already performed for given commit, and then it just runs again for no good reason. I don't understand it.
Will try your fix. Thanks.
Though I'm not fan of running custom builds as they will not receive fixes for vulnerabilities and other bugs.
But it is better than running infinite website build loops...
I actually built the HPI from this project using mvn hpi:hpi, but now the "GitHub hook trigger for GITScm polling" button is gone:
Is there some special flag that needs to be used at build time that I don't know about?
I wasn't really on dnwe:only-trigger-if-branchspec-matches-ref branch, it just defaulted to master, despite that branch being the default for the fork. But I built dnwe:only-trigger-if-branchspec-matches-ref and installed it through "advanced" page in plugins section, but the "GitHub hook trigger for GITScm polling" button is still missing, any ideas?
@jakubgs sounds like your compile didn't work — just grab this .hpi file from the linked Jenkins jobs on the PR.
Direct Link: github-1.38.1-rc825.6efa_b_a_0b_7b_f8.hpi
It doesn't seem to work, builds are still triggered by other branches:
This job is configured to build from develop, and yet is being triggered by deploy-develop. This issue is infuriating.
@jakubgs make sure your branch specifier is absolute with the refs/heads prefix
^ here's one of mine (for example), make yours refs/heads/develop if that's the only branch you want to build off
Thanks for the help, but it doesn't seem this branch fixes the loop. My website builds still get stuck in an endless build loop.
Need to debug this more.