bradygaster · bradygaster · Aug 21, 2026 · Aug 21, 2026
@@ -0,0 +1,256 @@
+# Windows Test Baseline
+
+**Author:** EECOM (Core Dev)
+**Measured:** 2026-08-21
+**Baseline commit:** `369bba8f` (`dev` tip at time of measurement)
+**Host:** Windows, Node via vitest 4.1.10, non-Windows-Terminal console
+**Checkout topology:** git **worktree** (`C:\src\copilot-worktrees\squad\...`, main checkout `C:\src\squad`)
+
+> **Why this document exists.** Three times this workstream has been bitten by
+> *reports nothing → mistaken for passing*. `check-changeset-drift.test.ts` ran
+> **0 of 8 tests** for a month while everyone read it as environment noise. This
+> baseline removes the ambiguity: it states exactly what red is expected, and
+> separates suites that **failed assertions** from suites that **never ran**.
+
+---
+
+## 1. The number
+
+**On Windows at commit `369bba8f`, in a git worktree, on a non-Windows-Terminal console:**
+
+```
+Test Files   9 failed | 264 passed | 1 skipped (274)
+Tests       15 failed | 7431 passed | 42 skipped | 47 todo
+```
+
+**Anything beyond that is new.**
+
+Linux CI on the same tree is **fully green — 274/274 files passed** (run `32461944794`).
+Every failure below is therefore Windows-and/or-topology specific *by construction*.
+
+### The number moves for three reasons — know which one you hit
+
+The single-number baseline is a lie unless you also state the environment. Three
+independent modifiers change the count *without any code changing*:
+
+| Modifier | Effect | Verified |
+| --- | --- | --- |
+| Running in **Windows Terminal** (`WT_SESSION` set) | `repl-ux.test.ts` **passes** → **8 failed**, 14 tests failed | ✅ measured both ways |
+| Running in a **plain clone** instead of a worktree | `acceptance.test.ts` init failure disappears → one fewer | ⚠️ mechanism proven, not re-run in a plain clone |
+| **#1790 merged AND working tree renormalized** | 3 suites go green → **6 failed** | ✅ measured (see §4) |
+
+If you see a count other than 9, **check these three before declaring a regression.**
+
+---
+
+## 2. Load failures vs assertion failures — read this first
+
+These two look nearly identical in vitest output and are completely different in consequence.
+
+| Class | What it means | Coverage |
+| --- | --- | --- |
+| **Assertion failure** | The suite loaded, ran its tests, some assertions were false | Real |
+| **LOAD FAILURE** | The suite **never ran a single test**. `Tests: no tests` | **Silently zero** |
+
+**At `369bba8f` there are 2 load failures and 1 partial-load failure — 3 suites in
+the #1788 family, not 1.** Two of them were previously unknown.
+
+| Suite | Tests it *should* run | Tests it actually ran |
+| --- | --- | --- |
+| `test/scripts/check-changeset-drift.test.ts` | 8 | **0** |
+| `test/promote-insider-tag.test.ts` | 15 | **0** |
+| `test/cli/patch-esm-imports.test.ts` | 6 | 6 loaded, but **3 die with the same `SyntaxError`** |
+
+That is **23 tests of enforcement with zero effective coverage on Windows**, plus 5
+more failing from the same root cause.
+
+---
+
+## 3. Triage table
+
+| # | Test file | Failure mode | Pre-existing? | Windows-only? | Mechanism | Verdict |
+| --- | --- | --- | --- | --- | --- | --- |
+| 1 | `test/scripts/check-changeset-drift.test.ts` | **LOAD FAILURE — 0 of 8 tests**, `SyntaxError: Invalid or unexpected token` | Yes | Yes | CRLF shebang in imported `scripts/check-changeset-drift.mjs`; Vite's shebang stripping doesn't survive `\r`, leaving a bare `#` as first token | **real-defect** — filed as #1788, fix open as #1790 |
+| 2 | `test/promote-insider-tag.test.ts` | **LOAD FAILURE — 0 of 15 tests**, same `SyntaxError` | Yes | Yes | Identical to #1, via `scripts/promote-insider-tag.mjs` | **real-defect** — *previously unknown*; covered by #1790's fix + guard test |
+| 3 | `test/cli/patch-esm-imports.test.ts` | 5 of 6 fail; **3 with `SyntaxError`**, 2 with `expected false to be true` | Yes | Yes | Identical to #1, via `packages/squad-cli/scripts/patch-esm-imports.mjs` | **real-defect** — *previously unknown, and previously mislabeled as "ESM patching noise"* |
+| 4 | `test/scheduler.test.ts` | 4 assertion failures (`success` false, `code` undefined, `stderr` empty) | Yes | Manifestation only | `LocalPollingProvider.execute` does `entry.task.ref.trim().split(/\s+/)` — **no quote handling**. Test ref embeds `process.execPath` = `C:\Program Files\nodejs\node.exe`; the space splits the command, spawn ENOENTs | **real-defect** — genuine cross-platform argv bug, *not* "scheduler timing" |
+| 5 | `test/cli-packaging-smoke.test.ts` | 1 assertion: `squad --version` returns `''` with exit 0 | Yes | Yes | 2000 ms `execFileSync` timeout exceeded on Windows cold start. Harness catch-block maps `err.killed`/`SIGTERM` to **`exitCode: 0`**, so a hang is indistinguishable from success | **real-defect** — harness converts hangs into passes |
+| 6 | `test/plugin-extensibility.test.ts` | `EPERM: operation not permitted, symlink` — throws in test *setup* | Yes | Yes | Windows requires Developer Mode/elevation to create symlinks | **expected-noise** — but see caveat below |
+| 7 | `test/repl-ux.test.ts` | 1 assertion: expected `──────────`, got ASCII `------` | Yes | **Terminal-dependent, not OS-dependent** | `terminal.ts:84` — `supportsUnicode: plat !== 'win32' \|\| Boolean(process.env['WT_SESSION'])`. Production is behaving **correctly**; the test hardcodes the Unicode glyph | **expected-noise** (test-side platform assumption) |
+| 8 | `test/cli/watch-capabilities.test.ts` | 1 assertion: expected `StringContaining ".squad/ralph-instructions.md"`, got `"\some\repo\.squad\ralph-instructions.md"` | Yes | Yes | Production builds the path with `path.join` (correct on Windows); the test asserts a POSIX-separator substring | **expected-noise** (test-side) |
+| 9a | `test/acceptance/acceptance.test.ts` › *Init in existing project* | Expected `Squad initialized`, got `⚠ Git worktree detected / Main checkout: C:\src\squad` | Yes | **NO — worktree-only** | `init.ts:208` worktree guard fires because `.git` is a **file** (worktree) and the main checkout already has `.squad/`. Git topology, not OS | **expected-noise in a worktree** — would not reproduce in a plain clone |
+| 9b | `test/acceptance/acceptance.test.ts` › *Consult blocked in squadified project* | Expected `No personal squad found`, got `This project already has a .squad/ directory…` | Yes | **Unknown** | `setupConsultMode` throws a different error than the feature file expects. Ruled out: `~/.squad` does **not** exist here, so the obvious "developer has a personal squad" explanation is **wrong**. Ordering difference between `SquadifiedProjectError` and `PersonalSquadNotFoundError` not isolated | **UNKNOWN** — see §6 |
+
+### Caveat on #6 (`plugin-extensibility`)
+
+The mechanism is a genuine OS limitation, so the verdict is *expected-noise*. But
+the consequence is not benign: the test is named **"rejects symlinked plugin source
+files before install writes state"** — a **security guard** — and it throws during
+*setup*, so the guard is **never exercised on Windows**. Same "silently zero
+coverage" family as #1788, different cause. It should `skip` explicitly with a
+stated reason rather than error, so it is visibly-unverified instead of
+indistinguishable-from-broken.
+
+---
+
+## 4. #1790 will not fix your working tree — proven
+
+`.gitattributes` governs **checkout**, not files already on disk. Merging #1790
+adds `*.mjs text eol=lf`, but git does **not** rewrite existing working-tree files
+whose index content is unchanged.
+
+Measured here by temporarily applying #1790's `.gitattributes` line:
+
+```
+i/lf    w/crlf  attr/text eol=lf      packages/squad-cli/scripts/patch-esm-imports.mjs
+i/lf    w/crlf  attr/text eol=lf      scripts/promote-insider-tag.mjs
+
+Test Files  1 failed (1)
+     Tests  no tests            <-- still zero. Attribute applied, disk still CRLF.
+```
+
+`core.autocrlf = true`, index is already LF, so the attribute alone changes nothing
+on disk. #1790 renormalizes only the 9 `.mjs` files that were stored **CRLF in the
+index**; the 3 suites above are not among them.
+
+**After pulling #1790, force a re-checkout of the `.mjs` files** or the three suites
+stay red and it will look like the fix failed:
+
+```powershell
+# targeted
+Remove-Item scripts\promote-insider-tag.mjs, scripts\check-changeset-drift.mjs, packages\squad-cli\scripts\patch-esm-imports.mjs -Force
+git checkout -- scripts/promote-insider-tag.mjs scripts/check-changeset-drift.mjs packages/squad-cli/scripts/patch-esm-imports.mjs
+
+# or repo-wide
+git rm --cached -r . ; git reset --hard
+```
+
+Verified result:
+
+```
+i/lf    w/lf    attr/text eol=lf      packages/squad-cli/scripts/patch-esm-imports.mjs
+i/lf    w/lf    attr/text eol=lf      scripts/promote-insider-tag.mjs
+
+Test Files  2 passed (2)
+     Tests  21 passed (21)
+```
+
+**Verify with `git ls-files --eol '*.mjs'` — you want `w/lf`, not `w/crlf`.**
+
+#1790's guard test (`test/scripts/mjs-shebang-loadable.test.ts`) scans **every**
+tracked shebanged `.mjs` and reads raw bytes, so it will cover #2 and #3 as well as
+#1. That design is correct and needs no change.
+
+---
+
+## 5. Controlled proofs
+
+Every mechanism above marked *real-defect* was proven by flipping one variable and
+re-running — not by reading code.
+
+| Claim | Experiment | Result |
+| --- | --- | --- |
+| #1/#2 are CRLF, not "environment" | Convert both `.mjs` to LF, re-run | `0 tests` → **23 passed (2 files)** |
+| #3 is the *same* bug, not "ESM patching" | Convert `patch-esm-imports.mjs` to LF, re-run | **5 failed → 6 passed** |
+| #7 is terminal-, not OS-dependent | Set `WT_SESSION`, re-run | **110 passed (0 failed)** |
+| #1790 alone doesn't repair a checkout | Apply the attribute, re-run without re-checkout | still **`no tests`** |
+| Re-checkout is the missing step | Delete + `git checkout --`, re-run | **21 passed** |
+
+Raw full-run output archived at
+`~/.copilot/session-state/a70ee969-.../files/full-test-run.txt`.
+
+---
+
+## 6. Unknowns — stated plainly
+
+An unknown reported as expected-noise is exactly how #1788 survived a month.
+**One item is unresolved:**
+
+**9b — `acceptance.test.ts` › "Consult blocked in squadified project".**
+The feature file expects `No personal squad found`; the CLI emits the *squadified*
+error instead. Linux CI is green, so the error ordering differs by environment. The
+obvious explanation (a personal squad at `~/.squad` short-circuiting the check) is
+**disproven** — `C:\Users\bradyg\.squad` does not exist. I did not isolate the real
+cause.
+
+Recommended controlled experiment: run
+`npx vitest run test/acceptance/acceptance.test.ts` in (a) a plain clone on Windows
+and (b) this worktree, with `SQUAD_HOME`/`HOME` pinned identically, and diff which
+branch of `setupConsultMode` throws. Until then this is **unknown**, not noise.
+
+**Also not fully isolated:** 9a is classified worktree-only from mechanism
+(`init.ts:208` reads `.git`-as-file, which is platform-independent) plus green Linux
+CI. It was **not** re-run in a plain Windows clone. Confidence high, proof partial.
+
+---
+
+## 7. Working-tree side effects of `npm test`
+
+Running the suite **mutates tracked files**. Check `git status --short` after every
+run and restore before committing.
+
+| File | Mutation |
+| --- | --- |
+| `.github/agents/squad.agent.md` | Version stamp rewritten `0.0.0-source` → `0.13.0` |
+| `test/__snapshots__/parser-contracts.test.ts.snap` | CRLF/LF churn |
+
+`npm run build` additionally rewrites to `-build.N` versions:
+`package.json`, `package-lock.json`, both `packages/*/package.json`, and both
+`packages/*/templates/skills/release-process/SKILL.md`. **Never commit any of these.**
+
+---
+
+## 8. Reconciliation with Procedures' independent run
+
+Procedures independently measured **9 failing files, all pre-existing** on his #1784
+branch. My count on clean `369bba8f` is also **9**. The counts agree — **no evidence
+of flakiness at the file level**, which is itself a useful result for tomorrow.
+
+His categorization does **not** survive verification:
+
+| His category | Actual |
+| --- | --- |
+| "ESM patching" | `patch-esm-imports.test.ts` is the **#1788 CRLF bug**, third instance — not an ESM-patching quirk |
+| "scheduler timing" | Not timing. `LocalPollingProvider` splits a command string on whitespace with **no quote handling**; any path with a space is unrunnable |
+| "Windows path separators" | Accurate for `watch-capabilities` only |
+
+Two of three category labels were wrong, and both wrong ones concealed real defects.
+That is the #1788 failure shape repeating: **a plausible grouping is not a diagnosis.**
+
+---
+
+## 9. Filed issues
+
+One issue per defect — deliberately not batched.
+
+| Table row | Finding | Issue |
+| --- | --- | --- |
+| §4 | #1790 doesn't repair existing working trees; needs a forced re-checkout | **#1793** |
+| 4 | `LocalPollingProvider` splits `task.ref` on whitespace without quote handling | **#1794** |
+| 5 | Packaging-smoke harness maps timeouts to `exitCode: 0`, masking hangs | **#1795** |
+| §7 | `npm test` mutates tracked files | **#1796** |
+
+**Not filed — already owned:** rows 1, 2, 3 all resolve to #1788 / #1790. Rows 2 and 3
+were previously unknown instances, but #1790's global `*.mjs text eol=lf` rule and its
+byte-level guard test (`test/scripts/mjs-shebang-loadable.test.ts`, which scans *every*
+tracked shebanged `.mjs`) cover them without modification. Filing separate issues would
+be duplicates. **The residual that #1790 does not cover is #1793.**
+
+**Not filed — correctly classified as noise:** rows 6, 7, 8, 9a.
+Row 6 carries a recommendation (make the symlink test `skip` explicitly rather than
+error, so an unverifiable security guard is visibly-unverified) but the mechanism is a
+genuine OS limitation, not a defect in our code.
+
+**Not filed — unknown:** row 9b. See §6. It gets an issue once the cause is isolated;
+filing an issue that says "something differs by environment" would be noise.
+
+---
+
+## 10. How to use this tomorrow
+
+1. Record the commit you are testing. This baseline is only valid for `369bba8f`.
+2. Note your console (`$env:WT_SESSION`) and topology (worktree vs clone).
+3. Expect **9 failed / 264 passed / 1 skipped**, adjusted by the modifiers in §1.
+4. **Grep the output for `no tests` and `Tests  no tests` first.** A suite that
+   loads zero tests is a dead gate, not a passing gate.
+5. Any failure not in §3 is new. Treat it as a regression until proven otherwise.