DO NOT MERGE — canary(META-321): calibration of the CLI Greptile policy

[qmarcelle]

Disposable canary. Do not merge. Closed without merging once calibration completes.

This is the positive control for the repo-owned policy that landed in #22 (f11d430). It exists to answer one question: do the CLI-specific rules actually fire on the defects they name? A rule that has not been shown to fail for its own defect is not eligible to become a hard gate.

Injected defects — head 0db1a26

Three deliberate semantic defects, each mapped to exactly one repo-owned rule so caught/not-caught is attributable per rule.

#	File	Rule under test	Defect
D1	packages/cli/src/producer/generate.ts	history-refresh-refusal-observable	historyRefresh is emitted only when mining produced a block, so an explicit mineHistory: true that refuses returns a successful-looking result carrying the previous revision's counts, with no refusal reported. This is the PR #20 P1 reintroduced verbatim.
D2	packages/cli/src/producer/generate.ts	ordinary-generation-never-mines	The opt-in default is inverted to mineHistory !== false, so ordinary generation reads the commit graph on every run.
D3	packages/mining-core/src/project.ts	canonical-utf8-endpoint-order	canonicalPair drops compareUtf8 for bare <=, substituting UTF-16 code unit order for the UTF-8 byte order two independent producers are compared on.

Expected outcome

CI is expected red on this head — that is part of the control, confirming the defects are real rather than cosmetic. Measured locally before pushing:

• @workspacejson/mining-core — 1 failed / 96 passed (97): project.test.ts > uses UTF-8 BYTE order, not UTF-16 code unit order (D3)
• @workspacejson/cli — 24 failed, concentrated in generate.integration.test.ts and producer-conformance.test.ts (D1/D2)
• pnpm typecheck — clean on all three packages. None of these defects is a type error.

That last line is the point of the exercise. Type checking cannot see any of them, and D1 in particular is the class of defect that ships a plausible artifact — it is only visible to a reviewer that knows what the field is for.

What is being calibrated

1. Branch-local .greptile configuration is read, and the review check is tied to the current head.
2. Each rule catches its own positive control — recorded per rule, caught or not.
3. A reverted head does not repeat the finding.
4. A new push retriggers review.
5. Unresolved actionable findings affect merge eligibility as claimed.

Results are recorded on META-321 as observed counts and states.

Scope

No production change lands from this branch. Nothing here is merged, no producer behavior on main is modified, and no release or publish path is touched.

Summary by Sourcery

Adjust CLI workspace generation to mine history by default and simplify history refresh reporting, and change mining-core project file-pair canonicalization to use standard string ordering instead of UTF-8 byte-order comparison.

Enhancements:

• Change CLI history mining behavior to default to mining unless explicitly disabled.
• Simplify CLI history refresh outcome reporting to only emit when a history block is produced.
• Update mining-core canonical file pair ordering to rely on default string comparison rather than UTF-8 byte-order comparison.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[sourcery-ai]

Reviewer's guide (collapsed on small PRs)

Reviewer's Guide

Disposable canary PR that deliberately injects three semantic defects into CLI history mining and mining-core canonicalization to calibrate repo-owned Greptile policy rules and confirm they trigger on real, non-type-check-visible issues.

Sequence diagram for generateWorkspaceJson history mining and refresh outcome (defects D1/D2)

sequenceDiagram
  actor User
  participant Cli as cli_generate
  participant Generator as generateWorkspaceJson
  participant Miner as mineHistoryBlock

  User->>Cli: run workspace JSON generation
  Cli->>Generator: generateWorkspaceJson(options)
  alt options.mineHistory !== false
    Generator->>Miner: mineHistoryBlock(resolvedRoot, refreshDiagnostics)
    Miner-->>Generator: minedHistory
  else options.mineHistory === false
    Generator-->>Generator: minedHistory = undefined
  end

  Generator-->>Generator: preservedHistory = carryForwardHistory(existing)
  Generator-->>Generator: history = minedHistory ?? preservedHistory.history

  alt options.mineHistory === true && minedHistory !== undefined
    Generator-->>Generator: historyRefresh = { requested: true, mined: true, preserved: false }
  else [no mined history or not explicitly requested]
    Generator-->>Generator: historyRefresh = undefined
  end

  Generator-->>Cli: WorkspaceJsonV4 (history, historyRefresh)

Loading

File-Level Changes

Change	Details	Files
Altered CLI history mining defaults and refresh reporting to reintroduce two semantic defects related to mining behavior and refresh observability.	Changed history mining opt-in default so history is mined whenever mineHistory is not explicitly set to false, causing ordinary generation to mine history on every run. Simplified historyRefresh to only report when mining ran and produced a block, dropping refusal reporting and preserved-history signaling when mining is requested but produces no block.	packages/cli/src/producer/generate.ts
Modified mining-core canonical pair ordering to drop UTF-8 byte-order comparison in favor of naive string comparison, reintroducing a defect in endpoint canonicalization.	Removed compareUtf8 import from mining-core project module, leaving only type imports. Updated canonicalPair to use direct string <= comparison instead of compareUtf8, switching from UTF-8 byte-order to UTF-16 code unit ordering when canonicalizing file pairs.	packages/mining-core/src/project.ts

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[sourcery-ai]

Hey - I've reviewed your changes and they look great!

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[greptile-apps]

Greptile Summary

The canary’s three deliberate semantic defects have been reverted. This head only adds explanatory comments documenting refusal diagnostics and canonical UTF-8 endpoint ordering.

• Clarifies why history-refresh diagnostics must be passed into the mining operation.
• Documents why artifact-bound endpoint ordering must use compareUtf8.

Confidence Score: 5/5

The current head appears safe to merge because no blocking failure remains.

No blocking failure remains.

Important Files Changed

Filename	Overview
packages/cli/src/producer/generate.ts	Adds a comment explaining the existing refusal-diagnostics propagation; executable behavior remains correct.
packages/mining-core/src/project.ts	Adds a comment documenting the existing UTF-8 comparator requirement; canonicalization still uses compareUtf8.

_{Reviews (2): Last reviewed commit: "canary(META-321): non-violating head tou..." | Re-trigger Greptile}

[qmarcelle]

Calibration complete. Closing without merging — this branch existed only as a positive control.

Results across three heads

Head	Diff	Greptile check	Findings
0db1a26 3 defects	non-empty	failure (run 94133982853)	3, one per injected defect
2eac40d revert	nets to empty vs main	no check run posted at all	none
7fbb87c non-violating, same two files	non-empty	success (run 94135738644)	none — total stayed 3

Caught / not-caught, per rule

• D1 → history-refresh-refusal-observable — caught, P1, rule cited by name
• D2 → ordinary-generation-never-mines — caught, P1, rule cited by name
• D3 → canonical-utf8-endpoint-order — caught, P1, rule cited by name

3/3. Each finding cited Rule Used: … ([source](.greptile)), so branch-local repo-owned configuration was read, not a default profile. pnpm typecheck was clean on all three defects — none is visible to the type system.

No repeat on a non-violating head

The revert head was deliberately not accepted as proof: with a net-empty diff nothing was reviewed, so "no repeat" there would have been absence of evidence. 7fbb87c restored a real diff on the same two files with comment-only changes; Greptile reviewed it, concluded success, and added no findings. The three original threads went isOutdated: true.

Reconciliation

All three findings were replied to individually on their own threads, each confirming the defect, its rule, and its revert. No bulk resolution was used to produce that state.

Nothing from this branch lands. The branch is deleted.