Files
fusion/.changeset/fix-phantom-finalize-diagnostic-reset.md
gsxdsm a2a5db8151 fix(engine): surface phantom-finalize cause and reset HEAD when branch is authoritative
commitOrAmendMergeWithFixes used to swallow every unexpected throw as
`reason: "unknown-phantom"` and the two callers re-threw a bare
"verification fix finalize failed (unknown phantom)" with no operator-
actionable signal. FN-5422-class wedges (preAttemptHeadSha == currentHead
but branchTip ahead with task-trailered commits) couldn't be diagnosed
without re-running.

The catch now records the original error and probes whether the branch
ref itself is authoritative for the task (tip carries Fusion-Task-Id
trailer, base..branch is foreign-contamination-free). When it is — i.e.
the work isn't lost, the integration worktree just didn't advance — the
catch resets rootDir to preAttemptHeadSha so the next merge attempt
starts from a known baseline instead of inheriting partial squash state,
and returns `reason: "branch-ref-ahead-reset"`. The two callers fold
`originalError=` and `branchAuthority=` into the thrown message so the
real cause (diff-volume gate, file-scope, transient git, etc.) is
visible in the run log.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-21 19:10:24 -07:00

1.3 KiB

@runfusion/fusion
@runfusion/fusion
patch

Engine reliability: better diagnostics + clean-baseline reset on phantom finalize.

  • commitOrAmendMergeWithFixes previously swallowed all unexpected errors as reason: "unknown-phantom" and the two callers re-threw a verification fix finalize failed (unknown phantom) error with no surface area beyond the SHAs. FN-5422-class tasks wedged in review with no actionable signal in the failure message.
  • The catch now captures the original error message and runs an isBranchAuthoritativeForTask probe (existing branch ref carries this task's Fusion-Task-Id trailer + foreign-contamination check against base). When the branch ref is authoritative — meaning the AI's work is safely stored on fusion/<id> and only the in-merge attempt's integration worktree drifted — the catch resets rootDir to preAttemptHeadSha and returns reason: "branch-ref-ahead-reset". The next merge attempt then starts from a known-good baseline instead of inheriting half-built squash state.
  • Verification-fix and build-verification-fix callers now include the original error and the branch-authority probe outcome in the thrown error, so operators see the actual failure cause (e.g. diff-volume regression, file-scope violation, transient git error) rather than unknown phantom.