gsd-pi 2.81.0 → 2.82.0-dev.2841a1e44

package/src/resources/extensions/gsd/tests/clean-root-preflight.test.ts

@@ -8,7 +8,7 @@
 
 import test from "node:test";
 import assert from "node:assert/strict";
-import { mkdtempSync, mkdirSync, writeFileSync, rmSync, readFileSync, realpathSync } from "node:fs";
+import { existsSync, mkdtempSync, mkdirSync, writeFileSync, rmSync, readFileSync, realpathSync } from "node:fs";
 import { join } from "node:path";
 import { tmpdir } from "node:os";
 import { execSync } from "node:child_process";
@@ -208,8 +208,8 @@ test("postflightPopStash conflict warning names the exact stash ref", () => {
     assert.match(postflight.message, /failed after merge of milestone M005C/);
 
     const warning = notifications.find((n) => n.level === "warning")?.msg ?? "";
-    assert.match(warning, /git stash pop stash@\{\d+\}/);
     assert.match(warning, /git stash apply stash@\{\d+\}/);
+    assert.match(warning, /git stash drop stash@\{\d+\}/);
   } finally {
     try { rmSync(repo, { recursive: true, force: true, maxRetries: 3, retryDelay: 100 }); } catch { /* ignore */ }
   }
@@ -279,3 +279,108 @@ test("postflightPopStash falls back to milestone marker prefix when exact marker
     try { rmSync(repo, { recursive: true, force: true, maxRetries: 3, retryDelay: 100 }); } catch { /* ignore */ }
   }
 });
+
+test("postflightPopStash preserves stash when untracked collision differs from merged file", () => {
+  const repo = createTempRepo();
+  try {
+    writeFileSync(join(repo, "tests.txt"), "local preflight test\n");
+    const preflight = preflightCleanRoot(repo, "M009", () => {});
+    assert.equal(preflight.stashPushed, true, "preflight must stash untracked file");
+
+    writeFileSync(join(repo, "tests.txt"), "merged milestone test\n");
+    run("git add tests.txt", repo);
+    run('git commit -m "feat: add merged test"', repo);
+
+    const notifications: Array<{ msg: string; level: string }> = [];
+    const postflight = postflightPopStash(repo, "M009", preflight.stashMarker, (msg, level) => {
+      notifications.push({ msg, level });
+    });
+
+    assert.equal(postflight.needsManualRecovery, false, "different already-present untracked files must not stop auto-mode");
+    assert.equal(postflight.resolution, "already-present-preserved");
+    assert.deepEqual(postflight.collidedPaths, ["tests.txt"]);
+    assert.equal(readFileSync(join(repo, "tests.txt"), "utf-8"), "merged milestone test\n");
+    assert.equal(run("git status --porcelain", repo), "", "merged file must stay clean");
+
+    const stashList = run("git stash list", repo);
+    assert.ok(preflight.stashMarker && stashList.includes(preflight.stashMarker), "stash backup must be preserved");
+    assert.ok(
+      notifications.some((n) => n.level === "warning" && n.msg.includes("preserving")),
+      "user must be warned that the stash was preserved as backup",
+    );
+  } finally {
+    try { rmSync(repo, { recursive: true, force: true, maxRetries: 3, retryDelay: 100 }); } catch { /* ignore */ }
+  }
+});
+
+test("postflightPopStash drops stash when untracked collision is identical to merged file", () => {
+  const repo = createTempRepo();
+  try {
+    writeFileSync(join(repo, "tests.txt"), "same test content\n");
+    const preflight = preflightCleanRoot(repo, "M010", () => {});
+    assert.equal(preflight.stashPushed, true, "preflight must stash untracked file");
+
+    writeFileSync(join(repo, "tests.txt"), "same test content\n");
+    run("git add tests.txt", repo);
+    run('git commit -m "feat: add same test"', repo);
+
+    const postflight = postflightPopStash(repo, "M010", preflight.stashMarker, () => {});
+
+    assert.equal(postflight.needsManualRecovery, false, "identical already-present files must not stop auto-mode");
+    assert.equal(postflight.resolution, "already-present-dropped");
+    assert.equal(readFileSync(join(repo, "tests.txt"), "utf-8"), "same test content\n");
+
+    const stashList = run("git stash list", repo);
+    assert.ok(!stashList.includes(preflight.stashMarker ?? ""), "identical stash must be dropped");
+  } finally {
+    try { rmSync(repo, { recursive: true, force: true, maxRetries: 3, retryDelay: 100 }); } catch { /* ignore */ }
+  }
+});
+
+test("postflightPopStash requires manual recovery for mixed tracked changes and untracked collision", () => {
+  const repo = createTempRepo();
+  try {
+    writeFileSync(join(repo, "README.md"), "# local tracked work\n");
+    writeFileSync(join(repo, "tests.txt"), "local untracked work\n");
+    const preflight = preflightCleanRoot(repo, "M011", () => {});
+    assert.equal(preflight.stashPushed, true, "preflight must stash mixed changes");
+
+    writeFileSync(join(repo, "tests.txt"), "merged milestone test\n");
+    run("git add tests.txt", repo);
+    run('git commit -m "feat: add merged test"', repo);
+
+    const postflight = postflightPopStash(repo, "M011", preflight.stashMarker, () => {});
+
+    assert.equal(postflight.needsManualRecovery, true, "tracked stash payload must still require manual recovery");
+    assert.equal(postflight.resolution, "manual-recovery");
+    const stashList = run("git stash list", repo);
+    assert.ok(preflight.stashMarker && stashList.includes(preflight.stashMarker), "mixed stash must be preserved");
+  } finally {
+    try { rmSync(repo, { recursive: true, force: true, maxRetries: 3, retryDelay: 100 }); } catch { /* ignore */ }
+  }
+});
+
+test("postflightPopStash requires manual recovery when an untracked stash path is missing after collision", () => {
+  const repo = createTempRepo();
+  try {
+    writeFileSync(join(repo, "tests.txt"), "local untracked work\n");
+    writeFileSync(join(repo, "other-tests.txt"), "other local untracked work\n");
+    const preflight = preflightCleanRoot(repo, "M012", () => {});
+    assert.equal(preflight.stashPushed, true, "preflight must stash untracked files");
+
+    writeFileSync(join(repo, "tests.txt"), "merged milestone test\n");
+    run("git add tests.txt", repo);
+    run('git commit -m "feat: add one merged test"', repo);
+
+    const postflight = postflightPopStash(repo, "M012", preflight.stashMarker, () => {});
+
+    assert.equal(postflight.needsManualRecovery, true, "partial untracked restores must still require manual recovery");
+    assert.equal(postflight.resolution, "manual-recovery");
+    assert.equal(existsSync(join(repo, "other-tests.txt")), true, "git may partially restore the non-colliding path");
+    assert.match(run("git status --porcelain", repo), /\?\? other-tests\.txt/, "partial restore must leave manual recovery visible");
+    const stashList = run("git stash list", repo);
+    assert.ok(preflight.stashMarker && stashList.includes(preflight.stashMarker), "stash must remain for manual recovery");
+  } finally {
+    try { rmSync(repo, { recursive: true, force: true, maxRetries: 3, retryDelay: 100 }); } catch { /* ignore */ }
+  }
+});

package/src/resources/extensions/gsd/tests/closeout-git-deferral.test.ts

@@ -0,0 +1,16 @@
+// Project/App: GSD-2
+// File Purpose: Tests closeout git action deferral policy for auto-mode units.
+
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { shouldDeferCloseoutGitAction } from "../auto-post-unit.ts";
+
+test("execute-task defers closeout git action until verification passes", () => {
+  assert.equal(shouldDeferCloseoutGitAction("execute-task"), true);
+});
+
+test("non execute-task units keep pre-verification closeout git action", () => {
+  assert.equal(shouldDeferCloseoutGitAction("plan-slice"), false);
+  assert.equal(shouldDeferCloseoutGitAction("complete-slice"), false);
+});

package/src/resources/extensions/gsd/tests/context-store-decisions-from-memories.test.ts

@@ -0,0 +1,312 @@
+// ADR-013 Phase 6 cutover (Stage 1) — queryDecisionsFromMemories parity test.
+//
+// Verifies that reading active decisions from the `memories` table returns
+// the same Decision[] shape and content as the legacy `queryDecisions` read
+// from the `decisions` table, once Phase 5 dual-write has populated both
+// surfaces. Lock-in regression for the prompt-inline read path which was
+// switched to the memories source in auto-prompts.ts:inlineDecisionsFromDb.
+//
+// Scope of parity: ACTIVE decisions only. Superseded rows are intentionally
+// skipped by the existing backfill, so this test does not assert parity for
+// the supersedes-chain — that gap is acknowledged in
+// queryDecisionsFromMemories' contract and tracked for Stage 2/3.
+
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, mkdirSync, rmSync } from "node:fs";
+import { join } from "node:path";
+import { tmpdir } from "node:os";
+
+import {
+  closeDatabase,
+  insertDecision,
+  openDatabase,
+} from "../gsd-db.ts";
+import { saveDecisionToDb } from "../db-writer.ts";
+import {
+  queryDecisions,
+  queryDecisionsFromMemories,
+} from "../context-store.ts";
+
+function makeTmpBase(): string {
+  const base = mkdtempSync(join(tmpdir(), "gsd-decisions-memories-"));
+  mkdirSync(join(base, ".gsd"), { recursive: true });
+  openDatabase(join(base, ".gsd", "gsd.db"));
+  return base;
+}
+
+function cleanup(base: string): void {
+  try {
+    closeDatabase();
+  } catch {
+    /* noop */
+  }
+  try {
+    rmSync(base, { recursive: true, force: true });
+  } catch {
+    /* noop */
+  }
+}
+
+async function seedDecision(
+  base: string,
+  fields: {
+    when_context: string;
+    scope: string;
+    decision: string;
+    choice: string;
+    rationale: string;
+    revisable?: string;
+    made_by?: "human" | "agent" | "collaborative";
+  },
+): Promise<string> {
+  // saveDecisionToDb writes ONLY to memories post-Stage-3. For parity tests
+  // comparing the legacy `queryDecisions` against `queryDecisionsFromMemories`,
+  // mirror the same row into the legacy decisions table directly so both
+  // surfaces hold the same data and the parity assertion is well-defined.
+  const result = await saveDecisionToDb(
+    {
+      when_context: fields.when_context,
+      scope: fields.scope,
+      decision: fields.decision,
+      choice: fields.choice,
+      rationale: fields.rationale,
+      revisable: fields.revisable ?? "Yes",
+      made_by: fields.made_by ?? "agent",
+    },
+    base,
+  );
+  insertDecision({
+    id: result.id,
+    when_context: fields.when_context,
+    scope: fields.scope,
+    decision: fields.decision,
+    choice: fields.choice,
+    rationale: fields.rationale,
+    revisable: fields.revisable ?? "Yes",
+    made_by: fields.made_by ?? "agent",
+    superseded_by: null,
+  });
+  return result.id;
+}
+
+test("queryDecisionsFromMemories returns empty when no decisions exist", () => {
+  const base = makeTmpBase();
+  try {
+    assert.deepEqual(queryDecisionsFromMemories(), []);
+    assert.deepEqual(queryDecisionsFromMemories({ milestoneId: "M001" }), []);
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("queryDecisionsFromMemories matches queryDecisions for a single active decision", async () => {
+  const base = makeTmpBase();
+  try {
+    await seedDecision(base, {
+      when_context: "M001 discuss phase",
+      scope: "M001",
+      decision: "Adopt SQLite for persistence",
+      choice: "better-sqlite3",
+      rationale: "Native, synchronous, well-supported",
+    });
+
+    const fromDecisions = queryDecisions();
+    const fromMemories = queryDecisionsFromMemories();
+
+    assert.equal(fromDecisions.length, 1);
+    assert.equal(fromMemories.length, 1);
+
+    // Compare the user-visible Decision fields. seq differs across tables
+    // (memories.seq vs decisions.seq) so it's intentionally excluded.
+    const { seq: _seq1, ...d1 } = fromDecisions[0]!;
+    const { seq: _seq2, ...d2 } = fromMemories[0]!;
+    assert.deepEqual(d1, d2);
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("queryDecisionsFromMemories preserves decision order across multiple writes", async () => {
+  const base = makeTmpBase();
+  try {
+    const id1 = await seedDecision(base, {
+      when_context: "M001 discuss",
+      scope: "M001",
+      decision: "First decision",
+      choice: "A",
+      rationale: "first",
+    });
+    const id2 = await seedDecision(base, {
+      when_context: "M001 plan",
+      scope: "M001",
+      decision: "Second decision",
+      choice: "B",
+      rationale: "second",
+    });
+    const id3 = await seedDecision(base, {
+      when_context: "M002 discuss",
+      scope: "M002",
+      decision: "Third decision",
+      choice: "C",
+      rationale: "third",
+    });
+
+    const fromMemories = queryDecisionsFromMemories();
+    assert.equal(fromMemories.length, 3);
+    assert.deepEqual(
+      fromMemories.map((d) => d.id),
+      [id1, id2, id3],
+    );
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("queryDecisionsFromMemories filters by milestoneId (substring match on when_context)", async () => {
+  const base = makeTmpBase();
+  try {
+    await seedDecision(base, {
+      when_context: "M001 discuss",
+      scope: "M001",
+      decision: "M001 decision",
+      choice: "A",
+      rationale: "x",
+    });
+    await seedDecision(base, {
+      when_context: "M002 plan",
+      scope: "M002",
+      decision: "M002 decision",
+      choice: "B",
+      rationale: "y",
+    });
+    await seedDecision(base, {
+      when_context: "M001 execute",
+      scope: "M001-S01",
+      decision: "M001 follow-up",
+      choice: "C",
+      rationale: "z",
+    });
+    await seedDecision(base, {
+      when_context: "M003 plan",
+      scope: "M003",
+      decision: "Use M001 as precedent",
+      choice: "D",
+      rationale: "Mentions M001 outside when_context",
+    });
+
+    const m001 = queryDecisionsFromMemories({ milestoneId: "M001" });
+    assert.equal(m001.length, 2, "two decisions reference M001 in when_context");
+    assert.ok(m001.every((d) => d.when_context.includes("M001")));
+
+    const m002 = queryDecisionsFromMemories({ milestoneId: "M002" });
+    assert.equal(m002.length, 1);
+    assert.equal(m002[0]?.decision, "M002 decision");
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("queryDecisionsFromMemories filters by scope (exact match, no prefix collisions)", async () => {
+  const base = makeTmpBase();
+  try {
+    await seedDecision(base, {
+      when_context: "M001 discuss",
+      scope: "M001",
+      decision: "Milestone-level",
+      choice: "A",
+      rationale: "x",
+    });
+    await seedDecision(base, {
+      when_context: "M001 plan",
+      scope: "M001-S01",
+      decision: "Slice-level",
+      choice: "B",
+      rationale: "y",
+    });
+    await seedDecision(base, {
+      when_context: "M001 plan",
+      scope: "M001-S02",
+      decision: "Different slice",
+      choice: "C",
+      rationale: "z",
+    });
+
+    // Exact-scope filter must not match prefix-similar values.
+    const milestoneScope = queryDecisionsFromMemories({ scope: "M001" });
+    assert.equal(milestoneScope.length, 1, "scope=M001 must not match M001-S01 / M001-S02");
+    assert.equal(milestoneScope[0]?.scope, "M001");
+
+    const sliceScope = queryDecisionsFromMemories({ scope: "M001-S01" });
+    assert.equal(sliceScope.length, 1);
+    assert.equal(sliceScope[0]?.scope, "M001-S01");
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("queryDecisionsFromMemories matches queryDecisions for combined milestoneId + scope filters", async () => {
+  const base = makeTmpBase();
+  try {
+    await seedDecision(base, {
+      when_context: "M001 discuss",
+      scope: "M001",
+      decision: "A",
+      choice: "1",
+      rationale: "x",
+    });
+    await seedDecision(base, {
+      when_context: "M001 plan",
+      scope: "M001-S01",
+      decision: "B",
+      choice: "2",
+      rationale: "y",
+    });
+    await seedDecision(base, {
+      when_context: "M002 discuss",
+      scope: "M002",
+      decision: "C",
+      choice: "3",
+      rationale: "z",
+    });
+
+    const opts = { milestoneId: "M001", scope: "M001-S01" };
+    const fromDecisions = queryDecisions(opts);
+    const fromMemories = queryDecisionsFromMemories(opts);
+
+    assert.equal(fromDecisions.length, fromMemories.length);
+    assert.equal(fromMemories.length, 1);
+    assert.equal(fromMemories[0]?.id, fromDecisions[0]?.id);
+    assert.equal(fromMemories[0]?.scope, "M001-S01");
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("queryDecisionsFromMemories ignores memories without a sourceDecisionId marker", async () => {
+  const base = makeTmpBase();
+  try {
+    // Insert a user-authored memory (no sourceDecisionId) — must not appear in
+    // the decisions-from-memories projection.
+    const { createMemory } = await import("../memory-store.ts");
+    createMemory({
+      category: "architecture",
+      content: "User-authored architecture note, not derived from a decision",
+      scope: "project",
+    });
+
+    await seedDecision(base, {
+      when_context: "M001 discuss",
+      scope: "M001",
+      decision: "Real decision",
+      choice: "A",
+      rationale: "x",
+    });
+
+    const fromMemories = queryDecisionsFromMemories();
+    assert.equal(fromMemories.length, 1, "user-authored memory must not appear as a decision");
+    assert.equal(fromMemories[0]?.decision, "Real decision");
+  } finally {
+    cleanup(base);
+  }
+});

package/src/resources/extensions/gsd/tests/custom-engine-loop-integration.test.ts

@@ -182,7 +182,6 @@ function makeMockDeps(overrides?: Partial<LoopDeps>): LoopDeps & { callLog: stri
     pruneQueueOrder: () => {},
     isInAutoWorktree: () => false,
     shouldUseWorktreeIsolation: () => false,
-    mergeMilestoneToMain: () => ({ pushed: false, codeFilesChanged: false }),
     teardownAutoWorktree: () => {},
     createAutoWorktree: () => "/tmp/wt",
     captureIntegrationBranch: () => {},
@@ -420,6 +419,10 @@ describe("Custom engine loop integration", () => {
     await autoLoop(ctx, pi, s, deps);
 
     assert.deepEqual(turnResults, [{ status: "completed", failureClass: "none", error: undefined }]);
+    assert.ok(
+      deps.callLog.includes("journal:iteration-end"),
+      `complete workflow should emit iteration-end; log=${deps.callLog.join(",")}`,
+    );
     assert.ok(
       deps.callLog.indexOf("turnResult:completed") < deps.callLog.indexOf("stopAuto:Workflow complete"),
       `turn should finalize before stopAuto; log=${deps.callLog.join(",")}`,
@@ -472,6 +475,10 @@ describe("Custom engine loop integration", () => {
     assert.equal(turnResults[0].status, "stopped");
     assert.equal(turnResults[0].failureClass, "manual-attention");
     assert.match(turnResults[0].error ?? "", /custom-engine-dispatch-stop/);
+    assert.ok(
+      deps.callLog.includes("journal:iteration-end"),
+      `blocked workflow should emit iteration-end; log=${deps.callLog.join(",")}`,
+    );
     assert.equal(s.currentTraceId, null);
     assert.equal(s.currentTurnId, null);
     assert.equal(pi.calls.length, 0, "blocked workflow should not dispatch a custom step");
@@ -664,11 +671,16 @@ describe("Custom engine loop integration", () => {
       activeRunDir: runDir,
       basePath: runDir,
     });
+    const journalEvents: Array<{ eventType: string; data?: any }> = [];
     const deps = makeMockDeps({
       stopAuto: async (_ctx, _pi, reason) => {
         deps.callLog.push(`stopAuto:${reason ?? "no-reason"}`);
         s.active = false;
       },
+      emitJournalEvent: (entry: any) => {
+        journalEvents.push(entry);
+        deps.callLog.push(`journal:${entry.eventType}`);
+      },
     });
 
     const resolver = setInterval(() => {
@@ -700,6 +712,21 @@ describe("Custom engine loop integration", () => {
     assert.match(stopEntry ?? "", /requested retry 4 times without passing/);
     const finalGraph = readGraph(runDir);
     assert.equal(finalGraph.steps[0]?.status, "active", "failed verification must not reconcile the step complete");
+
+    const unitEndIndexes = journalEvents
+      .map((entry, index) => entry.eventType === "unit-end" ? index : -1)
+      .filter((index) => index >= 0);
+    const iterationEndIndexes = journalEvents
+      .map((entry, index) => entry.eventType === "iteration-end" ? index : -1)
+      .filter((index) => index >= 0);
+    assert.equal(unitEndIndexes.length, 4, "each custom verification retry/stop attempt must emit unit-end");
+    assert.equal(iterationEndIndexes.length, 4, "each custom verification retry/stop iteration must close after unit-end");
+    for (const [i, unitEndIndex] of unitEndIndexes.entries()) {
+      assert.ok(
+        iterationEndIndexes[i]! > unitEndIndex,
+        `custom verification attempt ${i + 1} should emit iteration-end after unit-end`,
+      );
+    }
   });
 
   it("persists custom verification retry budget across a session restart", async () => {

package/src/resources/extensions/gsd/tests/db-writer.test.ts

@@ -27,6 +27,7 @@ import {
   saveArtifactToDb,
   extractDeferredSliceRef,
 } from '../db-writer.ts';
+import { getAllDecisionsFromMemories } from '../context-store.ts';
 import type { Decision, Requirement } from '../types.ts';
 
 // ═══════════════════════════════════════════════════════════════════════════
@@ -326,11 +327,14 @@ describe('db-writer', () => {
 
       assert.deepStrictEqual(result.id, 'D001', 'saveDecisionToDb returns D001 as first ID');
 
-      // Verify DB state
-      const dbDecision = getDecisionById('D001');
-      assert.ok(!!dbDecision, 'decision exists in DB after save');
-      assert.deepStrictEqual(dbDecision?.scope, 'arch', 'DB decision has correct scope');
-      assert.deepStrictEqual(dbDecision?.choice, 'Option A', 'DB decision has correct choice');
+      // ADR-013 Stage 3: decisions land in memories, not the legacy table.
+      const memoryDecisions = getAllDecisionsFromMemories();
+      assert.equal(memoryDecisions.length, 1, 'one memory row exists after save');
+      const memDecision = memoryDecisions[0];
+      assert.ok(memDecision, 'memory decision exists after save');
+      assert.equal(memDecision.id, 'D001');
+      assert.equal(memDecision.scope, 'arch', 'memory decision has correct scope');
+      assert.equal(memDecision.choice, 'Option A', 'memory decision has correct choice');
 
       // Verify markdown file was written
       const mdPath = path.join(tmpDir, '.gsd', 'DECISIONS.md');
@@ -394,10 +398,11 @@ describe('db-writer', () => {
         assert.match(id, /^D\d{3}$/, `ID ${id} should match D### pattern`);
       }
 
-      // Verify all 5 exist in DB
+      // ADR-013 Stage 3: verify all 5 exist in the memories table (decisions
+      // table receives no writes from saveDecisionToDb post-cutover).
+      const memoryIds = new Set(getAllDecisionsFromMemories().map((d) => d.id));
       for (const id of ids) {
-        const row = getDecisionById(id);
-        assert.ok(row, `Decision ${id} should exist in DB`);
+        assert.ok(memoryIds.has(id), `Decision ${id} should exist in memories`);
       }
     } finally {
       closeDatabase();