Skip to content

fix(#8841): fall back to fork default branch for unsynced GitLab forks - #8928

Merged
ggallen merged 3 commits into
mainfrom
agent/8841-gitlab-unsynced-fork-scaffold
Nov 11, 2026
Merged

ggallen merged 3 commits into
mainfrom
agent/8841-gitlab-unsynced-fork-scaffold

Conversation

@fullsend-ai-coder

@fullsend-ai-coder fullsend-ai-coder Bot commented Oct 30, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Scaffold delivery to a GitLab fork that is behind upstream no longer fails with 400 ... invalid reference name '<sha>'. When the fork doesn't contain the upstream HEAD commit, the scaffold branch is created from the fork's own default-branch tip, but only after validation shows that is safe. Otherwise delivery stops with an error asking the user to sync the fork.

Changes

  • internal/forge/gitlab/repo.go: CreateBranchFromSHA maps GitLab's 400 "invalid reference name" (the SHA isn't in the target project) to forge.ErrNotFound. "already exists" still maps to forge.ErrAlreadyExists, and other 400s are unchanged.
  • internal/forge/forge.go: the CreateBranchFromSHA interface doc mentions the GitLab ErrNotFound case.
  • internal/layers/commit.go: in the cross-fork path of deliverBranchAndPR, an ErrNotFound triggers checkUnsyncedForkFallback, which:
    • refuses when the delivery contains deletions;
    • resolves the fork's own default branch and requires its tip to be an ancestor of (or identical to) the upstream commit;
    • refuses when an existing requested file differs between the fork and upstream;
    • returns the validated fork tip SHA, and the branch is created from exactly that SHA via CreateBranchFromSHA.
      A warning is emitted when the fallback is used, and an extra warning when an identical existing file is requested as executable (100755), because file modes cannot be compared and an upstream-only mode change would be lost.
  • The fallback sits inside the createBranch closure, so the stale-branch delete-and-recreate path uses it too.

Testing

  • go test ./internal/layers/... ./internal/forge/... passes.
  • Tests cover the GitLab error mapping, the fallback, each refusal case (deletions, diverged/unknown fork, unresolvable or differently named default branch, files differing between fork and upstream), and the executable-mode warning.
  • pre-commit couldn't fetch its hook repos in the sandbox (TLS failure); gofmt and go vet were run directly.

Closes #8841

Post-script verification

  • Branch is not main/master (agent/8841-gitlab-unsynced-fork-scaffold)
  • Secret scan passed (gitleaks — 8b8c75bf5d3ffa22c3b20ff7d96be17835c0688a..HEAD)
  • PR body secret scan passed (gitleaks — no-git)
  • Final PR description secret scan passed (gitleaks — no-git)

@fullsend-ai-review

fullsend-ai-review Bot commented Oct 30, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 11:13 AM UTC · Completed 11:23 AM UTC

Commit: 1d85d93 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $2.60

@fullsend-ai-review fullsend-ai-review Bot added the risk/moderate PR risk: moderate label Oct 30, 2026
@fullsend-ai-review

fullsend-ai-review Bot commented Oct 30, 2026 •

Copy link
Copy Markdown

Risk Assessment: moderate (2/5)

Details

Targeted bug fix with clear issue alignment and solid test coverage, balanced against active churn and coupling across core forge and layer commit packages.

Previous run

Risk Assessment: moderate (2/5)

Details

Targeted bug fix with clear issue alignment and solid test coverage, balanced against active churn and coupling across core forge and layer commit packages.

Previous run (2)

Risk Assessment: moderate (2/5)

Details

Targeted bug fix with clear issue alignment and solid test coverage, balanced against active churn and coupling across core forge and layer commit packages.

Previous run (3)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Targeted GitLab fork fallback fix with strong test coverage and clear issue alignment, balanced against high churn and change coupling across forge and commit layer packages.

Previous run (4)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Targeted GitLab fork fallback fix with strong test coverage and clear issue alignment, balanced against high churn and change coupling across forge and commit layer packages.

Previous run (5)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Targeted bug fix with clear issue alignment and solid test coverage, balanced against active churn and coupling across core forge and layer commit packages.

Previous run (6)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Targeted bug fix with clear issue alignment and solid test coverage, balanced against active churn and coupling across core forge and layer commit packages.

@fullsend-ai-review

fullsend-ai-review Bot commented Oct 30, 2026 •

Copy link
Copy Markdown

Looks good to me

Earlier findings

  • internal/forge/forge.go:1172, backward-incompatible: resolved. The CreateBranchFromSHA doc comment no longer promises ErrNotFound for an unknown commit on every implementation. It now limits that to a missing or inaccessible repository and to the GitLab unknown-SHA case, and says other implementations such as GitHub are not guaranteed to return it.

Labels: PR fixes CLI install/convergence scaffold delivery to GitLab forks

Previous run

<!-- fullsend:review-findings-v2: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 -->

Review

Findings

Low

  • [backward-incompatible] internal/forge/forge.go:1172 — The updated CreateBranchFromSHA doc says every implementation returns forge.ErrNotFound when "the commit or the target repository is not found". GitHub's implementation only yields ErrNotFound for HTTP 404 (via APIError.Unwrap), and GitHub normally rejects an unknown SHA with a 422, so a missing commit there likely does not satisfy forge.IsNotFound. The only caller (internal/layers/commit.go) is unaffected; a future caller relying on the interface wording could be misled.
    Remediation: Narrow the wording so ErrNotFound is guaranteed only for a missing/inaccessible target repository (HTTP 404) and, on GitLab, for an unknown SHA (400 "invalid reference name").

Earlier findings

  • internal/forge/gitlab/gitlab_test.go:951, pattern-inconsistency — resolved: both CreateBranchFromSHA tests now assert the handler was invoked.
  • internal/layers/commit.go:547, error-handling-idiom — resolved: checkUnsyncedForkFallback now splits the error branches and wraps underlying errors with %w.
  • internal/forge/forge.go:1172, backward-incompatible — resolved: the interface doc was rewritten to cover the missing commit or repository case (a narrower wording issue remains, listed above).

Labels: Go bug fix in the forge/layers code paths.

Previous run (2)

<!-- fullsend:review-findings-v2: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 -->

Review

Findings

Low

  • [pattern-inconsistency] internal/forge/gitlab/gitlab_test.go:951 — The new httptest handlers in TestCreateBranchFromSHA_SHANotInRepo and TestCreateBranchFromSHA_OtherBadRequest do not assert that the handler was invoked, as docs/contributing/go-code.md requires. Risk is small, since other assertions would likely fail on an unmatched route.
    Remediation: Add a handlerCalled flag set inside each handler and assert it after the client call.
  • [error-handling-idiom] internal/layers/commit.go:547 — In checkUnsyncedForkFallback, the CompareCommits guard (and the GetRepo guard near line 533) combine the err != nil and non-ancestor/empty-branch cases in one message that always interpolates error: %v, so a diverged fork yields error: <nil>, and %v drops wrapping. forkBranch is formatted with %s instead of %q.
    Remediation: Split the API-error case (wrap with %w) from the non-ancestor case, and quote forkBranch with %q.
  • [backward-incompatible] internal/forge/forge.go:1172 — The interface doc attributes ErrNotFound from CreateBranchFromSHA to GitLab only, but the fallback keys on forge.IsNotFound(err) with no forge check. On GitHub, any HTTP 404 (e.g. a missing or inaccessible fork repo) maps to ErrNotFound and would enter the fallback; its safety checks fail closed, so the effect is mainly a different error message.
    Remediation: Document ErrNotFound as covering "commit or target repository not found" for all implementations, or add a dedicated sentinel for the missing-commit case.
Previous run (3)

<!-- fullsend:review-findings-v2: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 -->

Review

Findings

Low

  • [logic-error] internal/layers/commit.go:565 — checkUnsyncedForkFallback accepts the fallback when each existing requested file has identical content at the fork tip and upstream, but it never compares file modes. GitLab's commitFilesImpl skips a file only when both blob SHA and mode already match the fork (internal/forge/gitlab/repo.go:1152). If the fork tip has a script at 100755, upstream later changed only the bit to 100644 with identical content, and the scaffold requests 100755, the fork commit skips that file, so the MR carries no mode change. If nothing else changed, delivery reports "up to date". The trigger is narrow and the gap is documented in the function comment, but the loss is silent.
    Remediation: Compare file modes at the validated fork and upstream commits (e.g. a forge.Client method returning the tree-entry mode at a ref) and refuse the fallback when an existing requested file's mode differs, with a regression test for identical content and differing executable bits. Alternatively, emit a StepWarn when the fallback is taken for files requested as 100755.

Earlier findings

  • internal/layers/commit.go:554, logic-error — resolved_by_change: the latest commit now compares fork and upstream content for every existing requested file and refuses the fallback if they differ; TestCommitScaffoldViaPR_CrossForkUnsyncedRefusesPartialRestoration covers the former gap.
Previous run (4)

<!-- fullsend:review-findings-v2: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 -->

Review

Findings

Medium

  • [logic-error] internal/layers/commit.go:554 — The fallback still silently drops partial-file restorations. When the requested content differs from the fork content, the loop does continue without checking upstream. Example: the fork has setting=old and other=old in separate sections, upstream changes setting=new, and the scaffold requests setting=old and other=new. Validation passes, but the MR diff against the fork point changes only other, and a clean merge keeps setting=new instead of the requested setting=old. The new guard catches whole-file no-ops but not unchanged portions of otherwise modified files. This also contradicts the doc comment that a stale fork can "at worst cause merge conflicts".
    Remediation: Before permitting the fallback, check upstream for every requested path. Refuse when an existing requested file differs between fork and upstream (including mode), unless delivery can guarantee the requested result. Add a regression test with independent upstream and requested changes in separate sections of one file.

Earlier findings

  • internal/layers/commit.go:386, logic-error — resolved. The fallback now creates the branch from the validated fork tip SHA via CreateBranchFromSHA instead of re-resolving the default branch.
  • internal/layers/commit.go:508, logic-error — resolved. Validation now resolves the fork's own default branch via GetRepo instead of using upstream's branch name; tests cover a differently named default branch.

Next steps:

  • /fs-fix — agent addresses review findings automatically
  • /fs-fix <your instruction> — agent fixes with your specific guidance
  • Push commits directly — review re-runs automatically on push
  • /fs-fix-stop — disable automatic fix runs for this PR
Previous run (5)

<!-- fullsend:review-findings-v2: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 -->

Review

Findings

Medium

  • [logic-error] internal/layers/commit.go:508 — The ancestry guard validates the fork branch named by upstream's default branch, but the fallback CreateBranch resolves the fork's own default branch independently. If the two default branch names differ, an ancestral fork main passes the check while the scaffold is created from an unchecked, possibly diverged branch, so fork-only commits can enter the scaffold MR. This contradicts the function's own doc comment.
    Remediation: Resolve the fork's actual default branch, validate its tip, and create the scaffold branch from that exact SHA via CreateBranchFromSHA. Add a test where upstream and fork default branch names differ.

  • [logic-error] internal/layers/commit.go:386 — Rejecting deletions does not prevent every silently incomplete delivery from a stale ancestor. If a scaffold file is A at the fork tip, upstream changed it to B, and the delivery restores A, the branch from the fork tip already matches A. GitLab's commitFilesImpl skips files whose sha and mode already match, and the MR diff from the merge base shows no change, so delivery can report success while upstream keeps B.
    Remediation: Base the scaffold on upstream's tree, or reject the fallback when any requested file matches the fork version but differs from upstream. Add a regression test for restoring a file to the fork's existing content.

Earlier findings

  • internal/layers/commit.go:380, logic-error — resolved by change: the unconditional fallback is now gated by checkUnsyncedForkFallback (refuses deletions, requires fork tip to be an ancestor of upstream), with new tests.
  • internal/layers/commit.go:380, logic-error — resolved by change: diverged/behind forks, unknown fork commits and unresolvable fork branches are now refused, with new tests.

Next steps:

  • /fs-fix — agent addresses review findings automatically
  • /fs-fix <your instruction> — agent fixes with your specific guidance
  • Push commits directly — review re-runs automatically on push
  • /fs-fix-stop — disable automatic fix runs for this PR
Previous run (6)

<!-- fullsend:review-findings-v2:eyJ2ZXJzaW9uIjoyLCJhY3Rpb24iOiJyZXF1ZXN0LWNoYW5nZXMiLCJmaW5kaW5ncyI6W3sic2V2ZXJpdHkiOiJtZWRpdW0iLCJjYXRlZ29yeSI6ImxvZ2ljLWVycm9yIiwiZmlsZSI6ImludGVybmFsL2xheWVycy9jb21taXQuZ28iLCJpZCI6ImZfODZiZjRjNDQwY2M0MWI0OSIsImxpbmUiOjM4MCwiYWN0aW9uYWJsZSI6dHJ1ZX0seyJzZXZlcml0eSI6Im1lZGl1bSIsImNhdGVnb3J5IjoibG9naWMtZXJyb3IiLCJmaWxlIjoiaW50ZXJuYWwvbGF5ZXJzL2NvbW1pdC5nbyIsImlkIjoiZl81NjRlZmU0MWE2NGY1YTAyIiwibGluZSI6MzgwLCJhY3Rpb25hYmxlIjp0cnVlfV19 -->

Review

Findings

Medium

  • [logic-error] internal/layers/commit.go:380 — When the fork lacks the upstream SHA, the fallback branches from the stale fork's default branch, and requested deletions for files added upstream after the fork point are silently lost (uninstall and scaffold-upgrade deliveries use TreeFile{Delete:true}; GitLab's commitFilesImpl skips deletes for paths missing from the branch tree, and the path is also absent from the merge base). Previously this failed loudly with a 400; now it succeeds with an incomplete MR. New tests cover only additions.
    Remediation: For deliveries containing deletions, base the branch on the upstream commit (sync/push it into the fork), or refuse the fallback with an actionable sync-the-fork error. Add a regression test where a file exists upstream but not at the fork point.

  • [logic-error] internal/layers/commit.go:380 — A missing upstream SHA does not prove the fork's default branch is merely behind upstream; it may have diverged or use a different default branch. CreateBranch branches from the fork's own default branch with no ancestry check, so fork-only commits become part of the scaffold MR. The code comment's claim that the merge-base diff 'still contains only the scaffold changes' is incorrect for such commits.
    Remediation: Keep the upstream-based branch invariant by getting the upstream commit into the fork, or verify the fork tip is an ancestor of upstream and refuse the fallback otherwise. Correct the misleading comment and add coverage for a fork with unrelated default-branch commits.


Next steps:

  • /fs-fix — agent addresses review findings automatically
  • /fs-fix <your instruction> — agent fixes with your specific guidance
  • Push commits directly — review re-runs automatically on push
  • /fs-fix-stop — disable automatic fix runs for this PR

fullsend-ai-review[bot]

This comment was marked as outdated.

@codecov

codecov Bot commented Oct 30, 2026 •

Copy link
Copy Markdown

Codecov Report

❌ Patch coverage is 78.57143% with 15 lines in your changes missing coverage. Please review.

Files with missing lines Patch % Lines
internal/layers/commit.go 80.00% 8 Missing and 4 partials ⚠️
internal/forge/fake.go 0.00% 2 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!

@fullsend-ai-coder

fullsend-ai-coder Bot commented Oct 30, 2026 •

Copy link
Copy Markdown
Contributor Author

🤖 Finished Fix · ✅ Success · Started 11:25 AM UTC · Completed 11:37 AM UTC

Commit: 1d85d93 · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $0.41

@fullsend-ai-coder

Copy link
Copy Markdown
Contributor Author

🔧 Fix agent — iteration 1 (bot-triggered)

Addressed both review findings: the unsynced-fork fallback is now limited to additions-only deliveries where the fork default branch is verified as an ancestor of upstream; other cases fail with an actionable sync-the-fork error. Tests added. Pre-commit could not fetch hooks in the sandbox (gofmt and go vet run directly). Unrelated internal/security tests fail in the sandbox because of DNS resolution to a blocked IP.

Fixed (2):

  1. Unsynced-fork fallback silently drops requested deletions for files added upstream after the fork point (internal/layers/commit.go): The fallback now refuses with an actionable sync-the-fork error when the delivery contains any deletion. Added a regression test.
  2. Missing upstream SHA does not prove the fork default branch is merely behind upstream; fork-only commits could leak into the scaffold change request, and the code comment was misleading (internal/layers/commit.go): Added checkUnsyncedForkFallback, which requires the fork default branch tip to be an ancestor of (or identical to) the upstream commit via CompareCommits. It refuses with a sync-the-fork error on divergence, unknown commits, or an unresolvable fork branch. Corrected the comment. Added tests for diverged, behind, unknown-commit and unresolvable-branch cases.

Tests: passed

CI inspections (4):

  1. test (passing, success): Unit tests passed on the previous head.
  2. commit-lint (passing, success)
  3. codecov/patch (passing, success)
  4. behaviour (pending, in_progress): E2E behaviour job still running on the previous head.

Conflict update: skipped; strategy none; forge MERGEABLE; target main

Decision points
  • Refuse the fallback in unsafe cases rather than push upstream objects into the fork (alternatives: Sync or push the upstream commit into the fork, Refuse with a sync-the-fork error; rationale: No forge API exists to sync a fork, and refusing keeps the previous loud-failure behavior only for cases where the fallback is unsafe, while the common behind-only addition case still works.)

Next steps:

  • /fs-review — request a re-review of the changes
  • /fs-fix <your instruction> — run another fix pass with specific guidance
  • Push commits directly — review re-runs automatically on push
    Updated by fullsend fix agent

fullsend-ai-coder Bot added a commit that referenced this pull request Oct 30, 2026
Address review feedback on PR #8928. Branching a scaffold from the
fork's default branch when the fork lacks the upstream commit was
unsafe in two cases:

- Deliveries containing deletions: files added upstream after the fork
  point are absent from the fork tree, so the deletions were silently
  dropped. The fallback is now refused with a sync-the-fork error.
- Diverged forks or a differing default branch: fork-only commits
  would become part of the scaffold change request. The fallback now
  requires the fork's default branch tip to be an ancestor of (or
  identical to) the upstream commit, verified with CompareCommits.

The misleading comment claiming the diff always contains only scaffold
changes is replaced with the actual invariant. Tests cover deletions,
diverged/behind forks, an unknown fork commit, and an unresolvable
fork branch.

Note: pre-commit could not fetch remote hook repos (TLS certificate
verification failed in the sandbox); gofmt and go vet were run
directly. go test ./internal/security fails in the sandbox because
DNS resolves a .invalid host to a blocked IP; unrelated to this change.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@fullsend-ai-review

fullsend-ai-review Bot commented Oct 30, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 11:40 AM UTC · Completed 11:46 AM UTC

Commit: 9429ec7 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $1.30

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-coder

fullsend-ai-coder Bot commented Oct 30, 2026 •

Copy link
Copy Markdown
Contributor Author

🤖 Finished Fix · ✅ Success · Started 11:48 AM UTC · Completed 11:57 AM UTC

Commit: 9429ec7 · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $0.50

fullsend-ai-coder Bot added a commit that referenced this pull request Oct 30, 2026
Resolve the fork's own default branch, validate its tip as an ancestor of
upstream, and create the scaffold branch from that exact SHA. Also refuse
the fallback when a requested file already matches the fork version but
differs upstream, since the commit would be a silent no-op.

Addresses #8928

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@fullsend-ai-coder

Copy link
Copy Markdown
Contributor Author

🔧 Fix agent — iteration 2 (bot-triggered)

Addressed both findings: the unsynced-fork fallback now validates the fork's own default-branch tip and branches from that exact SHA, and refuses when a requested file matches the stale fork version but differs upstream. New tests cover both. Layers and forge package tests pass; an unrelated internal/security SSRF test fails in this sandbox because of DNS resolution behavior (not touched by this PR). pre-commit could not fetch hook repos in the sandbox, so gofmt, go vet and the interface-doc-sync lint were run directly and passed.

Fixed (2):

  1. Ancestry guard validates the fork branch named by upstream's default branch while the fallback branches from the fork's own default branch (internal/layers/commit.go): The fallback now resolves the fork's actual default branch via GetRepo, validates that tip as an ancestor of upstream, and creates the scaffold branch from that exact SHA with CreateBranchFromSHA. Added tests where fork and upstream default branch names differ (refused when diverged, accepted when ancestral and branched from the fork's own tip).
  2. Restoring a file to the fork's stale content makes the commit a no-op while upstream keeps a different version (internal/layers/commit.go): The fallback is now refused when a requested file already has the requested content at the fork tip but differs (or is missing) at the upstream commit, with a sync-the-fork error. Added regression tests for the refused case and for the harmless case where fork and upstream match. Added a per-SHA error injection field to the forge fake so tests can simulate only the upstream commit being missing.

Tests: passed

CI inspections (8):

  1. test-sandbox-darwin (passing, success)
  2. web (passing, success)
  3. commit-lint (passing, success)
  4. pr-title-lint (passing, success)
  5. DCO (passing, success)
  6. test (pending, in_progress): Still running at inspection time.
  7. playback (pending, in_progress): E2E job still running at inspection time.
  8. behaviour (pending, in_progress): E2E job still running at inspection time.

Conflict update: skipped; strategy none; forge MERGEABLE; target main

Next steps:

  • /fs-review — request a re-review of the changes
  • /fs-fix <your instruction> — run another fix pass with specific guidance
  • Push commits directly — review re-runs automatically on push
    Updated by fullsend fix agent

@fullsend-ai-review

fullsend-ai-review Bot commented Oct 30, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 12:00 PM UTC · Completed 12:06 PM UTC

Commit: dc12202 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $1.56

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-coder

fullsend-ai-coder Bot commented Oct 30, 2026 •

Copy link
Copy Markdown
Contributor Author

🤖 Finished Fix · ✅ Success · Started 12:08 PM UTC · Completed 12:17 PM UTC

Commit: dc12202 · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $0.39

@fullsend-ai-coder

Copy link
Copy Markdown
Contributor Author

🔧 Fix agent — iteration 3 (bot-triggered)

Addressed the single open review finding by refusing the stale-fork fallback whenever an existing requested file differs between the fork tip and upstream, with regression tests. Mode differences are not detected because the client interface returns content only.

Fixed (1):

  1. Unsynced-fork fallback still silently drops partially restored sections of modified files (internal/layers/commit.go): checkUnsyncedForkFallback now compares every requested file that exists at the fork tip against upstream and refuses the fallback with a sync-the-fork error unless the contents are identical (including when upstream removed the file or the read fails). This replaces the earlier check that only caught whole-file no-ops. The doc comment is updated to match. Added regression tests for independent upstream and requested changes in one file, and for a file removed upstream; the existing restore test message expectation was updated. File mode is not compared because the forge client exposes content only.

Tests: passed

CI inspections (4):

  1. test (passing, success): CI test job passed.
  2. codecov/patch (passing, success): Patch coverage gate passed.
  3. playback (pending, in_progress): E2E playback job still running at inspection time.
  4. behaviour (pending, in_progress): E2E behaviour job still running at inspection time.

Conflict update: skipped; strategy none; forge MERGEABLE; target main

Decision points
  • How to detect unsafe partial restorations without a three-way merge (alternatives: Refuse when fork and upstream content differ for any existing requested file, Base the scaffold branch on upstream by syncing the fork; rationale: Refusing is the smallest change, fails loudly with an actionable error, and needs no write access to sync the fork.)

Next steps:

  • /fs-review — request a re-review of the changes
  • /fs-fix <your instruction> — run another fix pass with specific guidance
  • Push commits directly — review re-runs automatically on push
    Updated by fullsend fix agent

fullsend-ai-coder Bot added a commit that referenced this pull request Oct 30, 2026
… from upstream

Check every requested file that exists at the fork tip against upstream and
refuse the fallback unless the contents are identical, so partially
restored sections of a modified file are not silently dropped by a clean
merge. Adds regression tests for partial restoration and files removed
upstream.

Notes: pre-commit could not fetch hook repos (TLS failure); ran gofmt and
go vet directly. internal/cli tests fail identically without this change
(sandbox environment). File modes are not compared (client exposes content
only).

Addresses #8928

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@fullsend-ai-review

fullsend-ai-review Bot commented Oct 30, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 12:19 PM UTC · Completed 12:24 PM UTC

Commit: a62b4b1 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $1.01

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-review
fullsend-ai-review Bot dismissed their stale review November 10, 2026 17:08

Superseded by updated review

@fullsend-ai-review fullsend-ai-review Bot added ready-for-merge All reviewers approved — ready to merge go Pull requests that update go code type/bug Confirmed defect in existing behavior labels Nov 10, 2026
@ggallen

ggallen commented Nov 10, 2026

Copy link
Copy Markdown
Member

/fs-fix

@fullsend-ai-coder

fullsend-ai-coder Bot commented Nov 10, 2026 •

Copy link
Copy Markdown
Contributor Author

🤖 Finished Fix · ✅ Success · Started 6:36 PM UTC · Completed 6:42 PM UTC

Commit: 00cadc6 · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $0.24

Narrow the CreateBranchFromSHA interface doc so ErrNotFound is only
guaranteed for a missing/inaccessible target repository and, on GitLab,
for an unknown commit SHA.

Note: pre-commit hooks were not run (could not fetch hook repos in the
sandbox: TLS failure); gofmt, go vet and go test were run directly.

Addresses #8928

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@fullsend-ai-coder

Copy link
Copy Markdown
Contributor Author

🔧 Fix agent — iteration 3 (human-triggered)

The latest review approved the PR with one low-severity doc-wording finding, and it is now fixed. Earlier findings were already resolved or dismissed. Tests in internal/forge and internal/layers pass; pre-commit could not fetch hook repos in the sandbox, so gofmt and go vet were run directly.

Fixed (1):

  1. CreateBranchFromSHA interface doc overstates that every implementation returns ErrNotFound for a missing commit (forge.go) (internal/forge/forge.go): Reworded the doc so ErrNotFound is guaranteed only for a missing or inaccessible target repository (HTTP 404) and, on GitLab, for an unknown commit SHA; other implementations such as GitHub are explicitly not guaranteed to return it for an unknown SHA.

Tests: passed

CI inspections (5):

  1. behaviour (transient-infra, failure): An e2e scenario failed while allocating a test repository: the GitHub tree-creation API returned a 422 invalid-blob error during scaffold commit. This is on the same-repo GitHub path, not the cross-fork GitLab fallback this PR changes, so it looks like a transient GitHub-side issue. Recommend rerunning the E2E Tests workflow; no code change made.
  2. test (passing, success): Unit tests passed.
  3. codecov/patch (passing, success): Patch coverage gate passed.
  4. semgrep (passing, success): Passed.
  5. commit-lint (passing, success): Passed.

Next steps:

  • /fs-review — request a re-review of the changes
  • /fs-fix <your instruction> — run another fix pass with specific guidance
  • Push commits directly — review re-runs automatically on push

Updated by fullsend fix agent

@fullsend-ai-review

fullsend-ai-review Bot commented Nov 10, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 6:44 PM UTC · Completed 6:50 PM UTC

Commit: 9a193a3 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $1.09

@fullsend-ai-review fullsend-ai-review Bot removed the ready-for-merge All reviewers approved — ready to merge label Nov 10, 2026
@fullsend-ai-review
fullsend-ai-review Bot dismissed their stale review November 10, 2026 18:50

Superseded by updated review

@fullsend-ai-review fullsend-ai-review Bot added ready-for-merge All reviewers approved — ready to merge component/install CLI install and app setup labels Nov 10, 2026
@rh-hemartin
rh-hemartin added this pull request to the merge queue Nov 10, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to failed status checks Nov 10, 2026
@ggallen
ggallen added this pull request to the merge queue Nov 10, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to failed status checks Nov 10, 2026
@ggallen
ggallen added this pull request to the merge queue Nov 10, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to failed status checks Nov 10, 2026
@ggallen
ggallen added this pull request to the merge queue Nov 11, 2026
Merged via the queue into main with commit 8f39f6b Nov 11, 2026
62 checks passed
@ggallen
ggallen deleted the agent/8841-gitlab-unsynced-fork-scaffold branch November 11, 2026 01:13

This branch was previously deployed

1 inactive deployment
dev — 9a193a34 Deployed Nov 10, 2026 by fullsend-ai-coder[bot] via behaviour #19543
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

component/install CLI install and app setup go Pull requests that update go code ready-for-merge All reviewers approved — ready to merge ready-for-review Triggers review agent dispatch risk/moderate PR risk: moderate type/bug Confirmed defect in existing behavior

Projects

None yet

Development

Successfully merging this pull request may close these issues.

gitlab: fullsend fails to make branch in personal fork if not up-to-date

2 participants