fix(regenerate): tolerate orphaned ACC baseline SHA in module detection

course-build/scripts/detect-affected-modules.mjs

@@ -11,11 +11,16 @@
 //   node detect-affected-modules.mjs --acc <acc-repo-path> --from <sha> --to <sha>
 //   node detect-affected-modules.mjs --acc <path> --to <sha>        # no --from: treat as first run
 //
-// Output (stdout, JSON): { "modules": [4,6], "min": 4, "fromModule": 4, "firstRun": false }
-//   modules    sorted unique affected module numbers (1..7)
-//   min        smallest affected module (the cascade root), or null when none
-//   fromModule generator --from value = min (module N produces start-of-module-(N+1))
-//   firstRun   true when --from was absent/empty (caller should treat as full regen)
+// Output (stdout, JSON): { "modules": [4,6], "min": 4, "fromModule": 4, "firstRun": false, "baselineMissing": false }
+//   modules         sorted unique affected module numbers (1..7)
+//   min             smallest affected module (the cascade root), or null when none
+//   fromModule      generator --from value = min (module N produces start-of-module-(N+1))
+//   firstRun        true when --from was absent/empty OR unreachable (caller: full regen)
+//   baselineMissing true when a --from was supplied but is not a reachable commit in --acc
+//
+// Robustness: if the supplied --from is not a reachable commit in the ACC repo (e.g. the
+// ACC history was rewritten/force-pushed and the recorded baseline was orphaned), we do NOT
+// crash on `git diff <bad-object>`. Instead we fall back to first-run semantics (full regen).
 
 import { execFileSync } from 'node:child_process';
 import { fileURLToPath } from 'node:url';
@@ -39,6 +44,25 @@ function git(cwd, ...args) {
   return execFileSync('git', args, { cwd, encoding: 'utf8', maxBuffer: 64 * 1024 * 1024 });
 }
 
+// True when `rev` resolves to a commit object that is actually present in the repo at `cwd`.
+// Guards against a recorded baseline SHA that was orphaned by an ACC history rewrite, which
+// would otherwise make `git diff <bad-object>` abort with "fatal: bad object".
+function commitExists(cwd, rev) {
+  try {
+    execFileSync('git', ['rev-parse', '--verify', '--quiet', `${rev}^{commit}`], { cwd, stdio: 'ignore' });
+    return true;
+  } catch (err) {
+    // `git rev-parse --verify --quiet` exits 1 (silently) for a well-formed but
+    // unknown/unreachable revision. Any other failure (bad --acc path, not a git repo,
+    // git missing) exits 128 or fails to spawn; rethrow so real errors fail fast instead
+    // of masquerading as a missing baseline and silently triggering a full regen.
+    if (err && err.status === 1) return false;
+    throw err;
+  }
+}
+
+export { commitExists };
+
 function moduleForPath(p) {
   // content/NN-*.md
   let m = p.match(/^content\/(\d{2})-.*\.md$/);
@@ -53,11 +77,24 @@ export { moduleForPath };
 
 function main() {
   const args = parseArgs(process.argv.slice(2));
-  const firstRun = !args.from || args.from.trim() === '';
+  // Normalize revisions once so the existence check and the diff always use the exact
+  // same strings (a stray newline/space, e.g. from a file, must not let the check pass
+  // while `git diff` still aborts on a bad revision).
+  const from = (args.from || '').trim();
+  const to = (args.to || '').trim();
+  const hasFrom = from !== '';
+
+  // A supplied baseline that is no longer reachable in the ACC repo (history rewrite / orphaned
+  // SHA) is treated as a missing baseline: fall back to first-run instead of crashing on git diff.
+  const baselineMissing = hasFrom && !commitExists(args.acc, from);
+  if (baselineMissing) {
+    console.error(`WARN: baseline commit ${from} is not reachable in --acc; treating as first run (full regen).`);
+  }
+  const firstRun = !hasFrom || baselineMissing;
 
   let files = [];
   if (!firstRun) {
-    const out = git(args.acc, 'diff', '--name-only', `${args.from}`, `${args.to}`);
+    const out = git(args.acc, 'diff', '--name-only', from, to);
     files = out.split('\n').map(s => s.trim()).filter(Boolean);
   }
 
@@ -74,6 +111,7 @@ function main() {
     min,
     fromModule: min,
     firstRun,
+    baselineMissing,
   }));
 }
 

course-build/scripts/selftest.mjs

@@ -44,5 +44,56 @@ check('M04=asset', bySource[4] === 'asset');
 check('M05=seed', bySource[5] === 'seed');
 check('M06=seed', bySource[6] === 'seed');
 
+// Baseline reachability + first-run fallback (integration: real temp git repo).
+// Guards the regenerate workflow against an orphaned .last-acc-sha crashing `git diff`.
+console.log('detect-affected-modules baseline fallback:');
+{
+  const { execFileSync } = await import('node:child_process');
+  const { mkdtempSync, writeFileSync, mkdirSync, rmSync } = await import('node:fs');
+  const { tmpdir } = await import('node:os');
+  const { join } = await import('node:path');
+  const { commitExists } = await import('./detect-affected-modules.mjs');
+
+  const script = resolve(__dirname, 'detect-affected-modules.mjs');
+  const repo = mkdtempSync(join(tmpdir(), 'acc-detect-'));
+  const g = (...a) => execFileSync('git', a, { cwd: repo, encoding: 'utf8', stdio: ['ignore', 'pipe', 'ignore'] });
+  const detect = (...a) => JSON.parse(execFileSync(process.execPath, [script, '--acc', repo, ...a], { encoding: 'utf8' }));
+  try {
+    g('init', '-q');
+    g('config', 'user.email', 'test@example.com');
+    g('config', 'user.name', 'test');
+    mkdirSync(join(repo, 'content'));
+    writeFileSync(join(repo, 'content', '03-x.md'), 'a\n');
+    g('add', '-A'); g('commit', '-q', '-m', 'c1');
+    const from = g('rev-parse', 'HEAD').trim();
+    writeFileSync(join(repo, 'content', '05-y.md'), 'b\n');
+    g('add', '-A'); g('commit', '-q', '-m', 'c2');
+    const to = g('rev-parse', 'HEAD').trim();
+
+    check('commitExists true for real commit', commitExists(repo, from) === true);
+    const BOGUS = 'b17669201ec145c91db7175e1fa4a1d60ba9fc01';
+    check('commitExists false for orphaned SHA', commitExists(repo, BOGUS) === false);
+
+    // A non-git directory is a real error, not a missing baseline: must rethrow, not return false.
+    const notARepo = mkdtempSync(join(tmpdir(), 'acc-norepo-'));
+    let threw = false;
+    try { commitExists(notARepo, from); } catch { threw = true; } finally { rmSync(notARepo, { recursive: true, force: true }); }
+    check('commitExists rethrows on non-git dir (fails fast)', threw === true);
+
+    const good = detect('--from', from, '--to', to);
+    check('valid baseline diffs (module 5 detected)', good.min === 5 && good.firstRun === false && good.baselineMissing === false);
+
+    // A baseline with surrounding whitespace (e.g. read from a file) must be normalized
+    // so the existence check and the diff agree; it must not fall into a spurious regen.
+    const padded = detect('--from', `\n  ${from}\n`, '--to', to);
+    check('whitespace-padded baseline is normalized (still module 5)', padded.min === 5 && padded.firstRun === false);
+
+    const missing = detect('--from', BOGUS, '--to', to);
+    check('orphaned baseline -> firstRun (no crash)', missing.firstRun === true && missing.baselineMissing === true);
+  } finally {
+    rmSync(repo, { recursive: true, force: true });
+  }
+}
+
 if (failures) { console.error(`\n${failures} check(s) failed.`); process.exit(1); }
 console.log('\nAll self-test checks passed.');