fix: remove error-entry exclusion from countFeedbackRounds

[adalton]

Summary

• Fix cost label oscillation observed in production on osac-operator PR #338: labels cycled between "Feedback (15)" and "Feedback (14) retry 1" because countFeedbackRounds excluded error entries from the count
• Remove !strings.Contains(e.Label, "error") filter — error entries with round numbers (e.g., "Feedback (15) (error)") should count as rounds since feedbackLabel always includes the round number
• Follow-up to fix: prevent feedback loop from silent reply failures and stale session data #44 which introduced the error exclusion based on a code review concern about bare "Feedback (error)" labels, which can't occur in practice

Test plan

• Updated TestCountFeedbackRounds — "error entries count as rounds" replaces "error entries do not count as rounds"
• All executor tests pass with -race
• Lint clean

Assisted-by: Claude noreply@anthropic.com

• Affected package: executor/
• Execution pipeline: Updates feedback handling so error feedback entries count toward distinct feedback rounds, preventing oscillating cost labels across retries.
• Tests: Adjusts round-counting expectations for error feedback entries.
• Infrastructure/configuration: No changes to container management, workspace lifecycle, crash recovery, configuration, or deployment.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Enterprise

Run ID: 33fb4e9f-cfa6-4f50-8fd7-26125f9ab329

📥 Commits

Reviewing files that changed from the base of the PR and between aba05f6 and 327d347.

📒 Files selected for processing (2)

• executor/costcomment.go
• executor/costcomment_test.go

---

Walkthrough

countFeedbackRounds now counts every Feedback label that is not a retry, including error-labelled entries. The associated test updates the expected count from 2 to 3.

Changes

Feedback round counting

Layer / File(s)	Summary
Count non-retry feedback rounds executor/costcomment.go, executor/costcomment_test.go	countFeedbackRounds no longer excludes labels containing error, and TestCountFeedbackRounds verifies that error-labelled feedback counts as an additional round.

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested labels: executor

🚥 Pre-merge checks | ✅ 13

✅ Passed checks (13 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main change: removing error-entry exclusion from countFeedbackRounds.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
No-Hardcoded-Secrets	✅ Passed	No hardcoded secrets, credentialed URLs, or long base64 literals were introduced in the changed files.
No-Weak-Crypto	✅ Passed	No weak crypto, custom crypto, or secret comparisons were introduced; the PR only changes feedback-label counting logic in costcomment.go/tests.
No-Injection-Vectors	✅ Passed	The patch only changes feedback-round counting and tests; no SQL/shell/eval/yaml/exec injection sinks or user-controlled execution paths were added.
Container-Privileges	✅ Passed	PASS: Only executor Go logic/tests changed; no container/K8s manifests or privileged settings (privileged, hostPID/Network/IPC, allowPrivilegeEscalation, root) were added.
No-Sensitive-Data-In-Logs	✅ Passed	PASS: no logging changes; the patch only adjusts feedback-round counting and tests, so it introduces no new sensitive-data exposure risk.
Resource-Leaks	✅ Passed	No new file, HTTP, DB, or network opens/goroutines were added; the patch only changes label-counting logic and tests.
Unchecked-Errors	✅ Passed	Touched files only change round counting/tests; no _ = err or silent error discard was introduced, and existing error paths are logged.
Ai-Attribution	✅ Passed	HEAD commit includes an Assisted-by trailer for Claude, and no Co-Authored-By AI trailer was found.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}