iframe-proxy

oliverpool · 2026-10-01T20:29:26Z

fixes #234

Description

This changes the behavior introduced in #231 (aborting early if the branch exists on the remote) by @BtbN (tagging you to get your feedback :)

Instead of early-aborting if the same branch exists on the remote, this PR performs the cherry-pick and the compares the local and remote:

if the remote has some original work (someone else pushed a commit), abort
otherwise force-push with lease (equivalent of rebasing the remote branch, if some new commit occured on the target branch)
Afterwards the PR creation is always attempted: if a 409 reply is received (a PR already exists), log the message and succeed (only for github/forgejo).

This would be valuable for Forgejo on codeberg, where sometimes the branch creation succeeds, but the PR creation fails: currently the branch is left there forever (needs a manual PR opening).
With this fix, a re-run of the job would update the branch and retry PR creation (nicer than a manual PR opening).

How Has This Been Tested?

Unit tests have been added at the git-cli level, as well as at the runner level.
I performed a manual test on a local forgejo instance:

setup a repo with main and lts branches
create and merge a PR targeting main
run git-backporting: backport PR is opened (unchanged)
push some commits to lts
re-run git-backporting: the existing branch will be rebased (force push) - PR will be detected as present
close the PR (but keep the branch)
re-run git-backporting: the existing branch will be left unchanged and a new PR created
push some commits to the backport branch
re-run git-backporting: failure with message Remote branch contains original work, aborting

Best reviewed commit-by-commit.

Checklist

Tests added if applicable.
Documentation updated if applicable.

Merge criteria:

The commits and have meaningful messages; the author will squash them after approval or will ask to merge with squash.
Testing instructions have been added in the PR body (for PRs involving changes that are not immediately obvious).
The developer has manually tested the changes and verified that the changes work

Note: dist/cli/index.js and dist/gha/index.js are automatically generated by git hooks and gh workflows.

github-actions · 2026-10-01T20:30:32Z

This reverts commit c59a6f7.

…nal work

the content will have been updated by the force-push

lampajr · 2026-10-17T18:04:15Z

Thanks again @oliverpool , I really appreciate the effort you're putting in improving this tool ❤️

Overall the idea looks good but I'd like to take some time to ensure it is working as expected.

if the remote has some original work (someone else pushed a commit), abort
otherwise force-push with lease (equivalent of rebasing the remote branch, if some new commit occured on the target branch)

This is the tricky part: I haven't checked the code yet, but I just want to make sure we don't run the risk of doing a force push to a branch we shouldn't, since that could be irreversible.

Afterwards the PR creation is always attempted: if a 409 reply is received (a PR already exists), log the message and succeed (only for github/forgejo).

This is very good improvement, +100 🚀

oliverpool

@lampajr thanks for the kind words!

To ease your review, here are the key additions of this PR.
Let me know what you think, once you have got time to review them!

oliverpool · 2026-10-19T19:29:19Z

-    const output = await simpleGit().raw(["ls-remote", "--heads", this.remoteWithAuth(remote), branch]);
-    return output.trim().length > 0;
+  async _isUpstreamMissingWork(cwd: string, upstream: string, head: string): Promise<boolean> {
+    const stdout = await this.git(cwd).raw(["cherry", upstream, head]);


See https://git-scm.com/docs/git-cherry

oliverpool · 2026-10-19T19:30:01Z

@@ -187,7 +227,15 @@ export default class GitCLIService {
    if (force) {
      options.push("--force-with-lease");


Note that I was able to keep the force-with-lease (since we fetch the upstream branch, to check for original work)

oliverpool · 2026-10-19T19:31:18Z

+          await git.gitCli.hasRemoteOriginalWork(configs.folder, backportPR.head, target_remote)
+        ) {
+          logger.info("Remote branch contains original work, aborting");
+          // throw original error: we cannot push in non-fast-forwarding way
+          throw err.cause;
+        }
+        if (
+          await git.gitCli.hasLocalOriginalWork(configs.folder, backportPR.head, target_remote)
+        ) {
+          logger.info("Force-pushing rebased branch");
+          await git.gitCli.push(configs.folder, backportPR.head, target_remote, true);


Here is the core of the logic:

fetch the remote branch

if it contains original work, abort

otherwise if we have original work (ie rebase), force-push-with-lease

I tried this with original work and it works pretty well:

[INFO ] [prod] Remote branch contains original work, aborting [ERROR] [prod] Something went wrong backporting to prod: Error: To https://github.com/lampajr/backporting-example.git ! refs/heads/bp-prod-6353aff:refs/heads/bp-prod-6353aff [rejected] (non-fast-forward) Done Pushing to https://github.com/lampajr/backporting-example.git error: failed to push some refs to 'https://github.com/lampajr/backporting-example.git' hint: Updates were rejected because the tip of your current branch is behind hint: its remote counterpart. If you want to integrate the remote changes, hint: use 'git pull' before pushing again. hint: See the 'Note about fast-forwards' in 'git push --help' for details. [ERROR] Error: Failure occurred during one of the backports: [Error: To https://github.com/lampajr/backporting-example.git ! refs/heads/bp-prod-6353aff:refs/heads/bp-prod-6353aff [rejected] (non-fast-forward) Done Pushing to https://github.com/lampajr/backporting-example.git error: failed to push some refs to 'https://github.com/lampajr/backporting-example.git' hint: Updates were rejected because the tip of your current branch is behind hint: its remote counterpart. If you want to integrate the remote changes, hint: use 'git pull' before pushing again. hint: See the 'Note about fast-forwards' in 'git push --help' for details. ] [INFO ] Process failed

We could probably strip some error msg here as it might be confusing but the important bit is [INFO ] [prod] Remote branch contains original work, aborting - can be improved later.

oliverpool · 2026-10-19T19:33:30Z

+        // Forgejo returns unstructured message:
+        // pull request already exists for these targets [id: .., issue_id: .., head_repo_id: .., base_repo_id: .., head_branch: bp-..., base_branch: ...]
+        this.logger.warn(err.message);
+        return "";


This is a bit hacky: return an empty URL if a pull request already exists for this base+head.

Ideally we should return the URL of the existing PR, but Forgejo does not expose this in a structured way (yet). I haven't tested github.

As a temporary solution might, even though I don't think it is working as expected with github:

POST /repos/lampajr/backporting-example/pulls - 422 with id 9BB4:1EB9F9:EF86DE:1631085:6ADBBFF9 in 750ms [ERROR] [prod] Something went wrong backporting to prod: HttpError: Validation Failed: {"resource":"PullRequest","code":"custom","message":"A pull request already exists for lampajr:bp-prod-6353aff."} - https://docs.github.com/rest/pulls/pulls#create-a-pull-request [ERROR] Error: Failure occurred during one of the backports: [HttpError: Validation Failed: {"resource":"PullRequest","code":"custom","message":"A pull request already exists for lampajr:bp-prod-6353aff."} - https://docs.github.com/rest/pulls/pulls#create-a-pull-request] [INFO ] Process failed

The process still fails in case of existing PR

lampajr

Hey @oliverpool , I tested this out a little bit and it looks good to me!

There is only one small issue that I'd like to cover before merging, looks like Forgejo returns different http code compared to github in case of existing PR - 409 (forgejo) vs 422 (github). Do you see any problem in implementing this as well?

feat: dedicated error type for git.push fast-forward failure

f04a0bf

oliverpool added 4 commits October 2, 2026 10:10

feat: git.hasRemoteOriginalWork git.hasLocalOriginalWork

19fa077

Revert "fix: don't backport if branch already exists (kiegroup#231)"

c131d24

This reverts commit c59a6f7.

feat: force-push-with lease if the remote branch exists without origi…

fd6cf0b

…nal work

feat: ignore existing pull request

e94d039

the content will have been updated by the force-push

oliverpool force-pushed the force-with-lease branch from fdbbe5f to e94d039 Compare October 2, 2026 08:11

oliverpool changed the title ~~Handle existing remote branch or pull request~~ Gracefully handle existing remote branch and/or pull request Oct 2, 2026

oliverpool commented Oct 19, 2026

View reviewed changes

lampajr reviewed Oct 23, 2026

View reviewed changes

Sunbelt Computer Software

PL/B Language Development and Support

Navigation Menu

Repository navigation

Gracefully handle existing remote branch and/or pull request - #239

Gracefully handle existing remote branch and/or pull request#239
oliverpool wants to merge 5 commits into
kiegroup:mainfrom
oliverpool:force-with-lease

oliverpool commented Oct 1, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 1, 2026 •

edited

Loading

Uh oh!

lampajr commented Oct 17, 2026

Uh oh!

oliverpool left a comment

Uh oh!

oliverpool Oct 19, 2026

Uh oh!

oliverpool Oct 19, 2026

Uh oh!

oliverpool Oct 19, 2026

Uh oh!

lampajr Oct 23, 2026

Uh oh!

oliverpool Oct 19, 2026

Uh oh!

lampajr Oct 23, 2026

Uh oh!

lampajr left a comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		@@ -187,7 +227,15 @@ export default class GitCLIService {
		if (force) {
		options.push("--force-with-lease");

Sunbelt Computer Software

PL/B Language Development and Support

Conversation

oliverpool commented Oct 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

How Has This Been Tested?

Checklist

Merge criteria:

Uh oh!

github-actions Bot commented Oct 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Coverage report

Test suite run success

Uh oh!

lampajr commented Oct 17, 2026

Uh oh!

oliverpool left a comment

Choose a reason for hiding this comment

Uh oh!

oliverpool Oct 19, 2026

Choose a reason for hiding this comment

Uh oh!

oliverpool Oct 19, 2026

Choose a reason for hiding this comment

Uh oh!

oliverpool Oct 19, 2026

Choose a reason for hiding this comment

Uh oh!

lampajr Oct 23, 2026

Choose a reason for hiding this comment

Uh oh!

oliverpool Oct 19, 2026

Choose a reason for hiding this comment

Uh oh!

lampajr Oct 23, 2026

Choose a reason for hiding this comment

Uh oh!

lampajr left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

oliverpool commented Oct 1, 2026 •

edited

Loading

github-actions Bot commented Oct 1, 2026 •

edited

Loading