Compare commits

...

15 Commits

Author SHA1 Message Date
felixm 91a3fb29c2 Add Windows 11 support behind evidence.Source and enforce.Guard ports
Polling active-window sensor and window-minimize guard as //go:build windows
adapters over a shared internal/winapi Win32 binding. Pure logic (class
normalization, emit-on-change tracking) is TDD-tested on Linux; syscall code is
cross-compile verified. No consumer changes.
2026-06-02 12:41:25 -04:00
felixm 9546999acb Clarify null-HWND doc note in ForegroundWindow 2026-06-02 12:39:32 -04:00
felixm d45a01eca5 Verify Windows cross-compile and Linux regression for Windows support
- GOOS=windows GOARCH=amd64 go build ./...  (whole module links)
- GOOS=darwin  GOARCH=amd64 go build ./...  (!linux && !windows fallback intact)
- go build ./... && go test ./...           (Linux host green, incl. new winapi/evidence tests)
- go vet ./... on linux and windows         (clean)
- cmd/antidriftd cross-compiles to a real PE32+ x86-64 .exe
2026-06-02 12:37:24 -04:00
felixm b6c7933f16 Align Windows guard ctx param and doc with X11 guard 2026-06-02 12:36:12 -04:00
felixm f37e295166 Add Windows window-minimize guard 2026-06-02 12:34:14 -04:00
felixm d149736946 Share unavailable() so the Windows sensor logs unavailable transitions 2026-06-02 12:33:23 -04:00
felixm 0daf8c2849 Add Windows polling active-window sensor 2026-06-02 12:30:16 -04:00
felixm 0f47d2dbf0 Document title/path truncation behavior in Win32 binding 2026-06-02 12:29:07 -04:00
felixm 0ac1511aac Add Win32 binding for foreground window and minimize 2026-06-02 12:24:55 -04:00
felixm 7d62af0750 Document foregroundTracker unavailable-state contract
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-06-02 12:23:54 -04:00
felixm e1194be3c2 Add emit-on-change tracker for the Windows polling sensor 2026-06-02 12:21:58 -04:00
felixm f79c149039 Clarify .exe suffix handling in class normalization 2026-06-02 12:21:12 -04:00
felixm 4ea8aed04b Add pure process-path to class normalization for Windows 2026-06-02 12:18:59 -04:00
felixm 885b0f9224 Plan Windows 11 support implementation
Six TDD/cross-compile tasks: pure class normalization and emit-on-change
tracker (unit-tested on Linux), Win32 binding, the two windows-tagged port
adapters, build-tag narrowing, and a whole-module cross-compile gate.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-02 12:08:01 -04:00
felixm e0ea1eca83 Spec Windows 11 support via windows-tagged port adapters
Full-parity design: evidence.Source (polling) and enforce.Guard
(ShowWindow) behind the existing X11/no-op port boundaries, pure Go,
verified compile-only.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-02 11:35:39 -04:00
14 changed files with 1051 additions and 8 deletions
@@ -0,0 +1,548 @@
# Windows 11 Support Implementation Plan
> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
**Goal:** Make AntiDrift function end-to-end on Windows 11 by adding `//go:build windows` adapters behind the existing `evidence.Source` and `enforce.Guard` ports — active-window sensing (polling) and window-minimize enforcement — with no consumer code changes.
**Architecture:** Two new windows-tagged adapters call a small shared Win32 binding (`internal/winapi`). The syscall-free logic (process-path → class normalization, emit-on-change tracking) is extracted into untagged files so it is unit-tested on Linux via TDD. The syscall-bound code is verified by cross-compilation only (no Windows machine is available). The current `//go:build !linux` no-op fallbacks are narrowed to `//go:build !linux && !windows` so Windows links the real adapters while macOS/other stay no-ops.
**Tech Stack:** Go 1.26, `golang.org/x/sys/windows` (already an indirect dep; promoted to direct), pure Go (no cgo). Win32 calls: `GetForegroundWindow`, `GetWindowThreadProcessId`, `OpenProcess`/`QueryFullProcessImageName`/`CloseHandle` (all typed in `x/sys/windows`), plus `GetWindowTextW` and `ShowWindow` via `windows.NewLazySystemDLL("user32.dll")`.
**Reference spec:** `docs/superpowers/specs/2026-06-02-windows-support-design.md`
---
## File structure
- Create `internal/winapi/class.go` — untagged — `ClassFromImagePath` (pure: image path → class string). Builds on all platforms; this is the only file that makes package `winapi` compile on Linux.
- Create `internal/winapi/class_test.go` — untagged — table test for `ClassFromImagePath` (runs on Linux).
- Create `internal/winapi/winapi.go``//go:build windows``ForegroundWindow()` and `MinimizeForeground()` over the Win32 calls.
- Create `internal/evidence/foreground_tracker.go` — untagged — `foregroundTracker.changed` (pure emit-on-change predicate). Runs on Linux.
- Create `internal/evidence/foreground_tracker_test.go` — untagged — tests for the tracker.
- Create `internal/evidence/windows.go``//go:build windows` — polling `Source` (`NewSource`, `windowsSource.Watch`).
- Create `internal/enforce/windows.go``//go:build windows``Guard` (`NewGuard`, `windowsGuard.MinimizeActive`).
- Modify `internal/evidence/source_other.go:1` — build tag `//go:build !linux``//go:build !linux && !windows`.
- Modify `internal/enforce/guard_other.go:1` — build tag `//go:build !linux``//go:build !linux && !windows`.
- Modify `go.mod``go mod tidy` promotes `golang.org/x/sys` from indirect to direct.
No changes to `session`, `web`, `domain`, or `cmd` — they already consume the ports and degrade on `Health`.
---
## Task 1: Pure class-name normalization (TDD on Linux)
`ClassFromImagePath` turns a Windows process image path into the on-task class
identity: base file name, trailing `.exe` removed (case-insensitive),
lowercased. It must be correct when the test runs on Linux, so it parses
separators itself rather than using OS-dependent `path/filepath`.
**Files:**
- Create: `internal/winapi/class.go`
- Test: `internal/winapi/class_test.go`
- [ ] **Step 1: Write the failing test**
Create `internal/winapi/class_test.go`:
```go
package winapi
import "testing"
func TestClassFromImagePath(t *testing.T) {
cases := []struct{ in, want string }{
{`C:\Program Files\Microsoft VS Code\Code.exe`, "code"},
{`C:\Windows\explorer.exe`, "explorer"},
{`chrome.exe`, "chrome"},
{`C:\x\FOO.EXE`, "foo"},
{`C:\x\My.App.exe`, "my.app"},
{`C:/forward/slash/Code.exe`, "code"},
{`firefox`, "firefox"},
{``, ""},
}
for _, c := range cases {
if got := ClassFromImagePath(c.in); got != c.want {
t.Errorf("ClassFromImagePath(%q) = %q, want %q", c.in, got, c.want)
}
}
}
```
- [ ] **Step 2: Run test to verify it fails**
Run: `go test ./internal/winapi/`
Expected: FAIL — build error, `undefined: ClassFromImagePath`.
- [ ] **Step 3: Write minimal implementation**
Create `internal/winapi/class.go`:
```go
// Package winapi is the Windows Win32 binding layer for AntiDrift's OS ports.
// The syscall-bound code lives in windows-tagged files; this untagged file
// holds the pure path logic so it builds and is tested on every platform.
package winapi
import "strings"
// ClassFromImagePath derives the on-task class identity from a process image
// path: the base file name with any trailing ".exe" removed (case-insensitive),
// lowercased. It is the Windows analog of an X11 WM_CLASS. It parses both `\`
// and `/` separators so it is correct regardless of the host OS running the
// test. Returns "" for an empty path.
func ClassFromImagePath(p string) string {
if i := strings.LastIndexAny(p, `\/`); i >= 0 {
p = p[i+1:]
}
if len(p) >= 4 && strings.EqualFold(p[len(p)-4:], ".exe") {
p = p[:len(p)-4]
}
return strings.ToLower(p)
}
```
- [ ] **Step 4: Run test to verify it passes**
Run: `go test ./internal/winapi/`
Expected: PASS (`ok antidrift/internal/winapi`).
- [ ] **Step 5: Commit**
```bash
git add internal/winapi/class.go internal/winapi/class_test.go
git commit -m "Add pure process-path to class normalization for Windows"
```
---
## Task 2: Emit-on-change tracker (TDD on Linux)
The X11 source is event-driven; the Windows source polls. `foregroundTracker`
reproduces "emit only on change" for the poll loop: it remembers the last
observation and reports whether the new one differs. It tracks availability too,
so a steady "no foreground window" run does not re-emit every tick.
**Files:**
- Create: `internal/evidence/foreground_tracker.go`
- Test: `internal/evidence/foreground_tracker_test.go`
- [ ] **Step 1: Write the failing test**
Create `internal/evidence/foreground_tracker_test.go`:
```go
package evidence
import "testing"
func TestForegroundTrackerChanged(t *testing.T) {
var tr foregroundTracker
if !tr.changed(true, 100, "A") {
t.Fatal("first observation should always report changed")
}
if tr.changed(true, 100, "A") {
t.Error("identical observation should not report changed")
}
if !tr.changed(true, 100, "B") {
t.Error("title change on same hwnd should report changed")
}
if !tr.changed(true, 200, "B") {
t.Error("hwnd change should report changed")
}
if !tr.changed(false, 0, "") {
t.Error("transition to unavailable should report changed")
}
if tr.changed(false, 0, "") {
t.Error("repeated unavailable should not report changed")
}
if !tr.changed(true, 200, "B") {
t.Error("transition back to available should report changed")
}
}
```
- [ ] **Step 2: Run test to verify it fails**
Run: `go test ./internal/evidence/ -run TestForegroundTracker`
Expected: FAIL — build error, `undefined: foregroundTracker`.
- [ ] **Step 3: Write minimal implementation**
Create `internal/evidence/foreground_tracker.go`:
```go
package evidence
// foregroundTracker remembers the last observed foreground window so the
// Windows polling Source emits only on change. hwnd is held as uintptr so this
// file stays platform-neutral (it must build and test on Linux).
type foregroundTracker struct {
primed bool
available bool
hwnd uintptr
title string
}
// changed reports whether (available, hwnd, title) differs from the last
// observation and records the new values. The first call always returns true.
func (t *foregroundTracker) changed(available bool, hwnd uintptr, title string) bool {
if t.primed && available == t.available && hwnd == t.hwnd && title == t.title {
return false
}
t.primed, t.available, t.hwnd, t.title = true, available, hwnd, title
return true
}
```
- [ ] **Step 4: Run test to verify it passes**
Run: `go test ./internal/evidence/ -run TestForegroundTracker`
Expected: PASS.
- [ ] **Step 5: Commit**
```bash
git add internal/evidence/foreground_tracker.go internal/evidence/foreground_tracker_test.go
git commit -m "Add emit-on-change tracker for the Windows polling sensor"
```
---
## Task 3: Win32 binding layer (cross-compile verified)
The syscall surface. Cannot be unit-tested without Windows; verified by
cross-compilation. All referenced `x/sys/windows` symbols were confirmed present
via `GOOS=windows go doc` while writing the spec.
**Files:**
- Create: `internal/winapi/winapi.go`
- Modify: `go.mod` (via `go mod tidy`)
- [ ] **Step 1: Write the implementation**
Create `internal/winapi/winapi.go`:
```go
//go:build windows
package winapi
import (
"unsafe"
"golang.org/x/sys/windows"
)
var (
user32 = windows.NewLazySystemDLL("user32.dll")
procGetWindowTextW = user32.NewProc("GetWindowTextW")
procShowWindow = user32.NewProc("ShowWindow")
)
// ForegroundWindow returns the current foreground window's handle (as a uintptr
// so platform-neutral callers need not import windows), its title, and its
// on-task class (process exe base name; see ClassFromImagePath). ok is false
// when there is no foreground window (e.g. secure desktop / lock screen).
func ForegroundWindow() (hwnd uintptr, title, class string, ok bool) {
h := windows.GetForegroundWindow()
if h == 0 {
return 0, "", "", false
}
return uintptr(h), windowTitle(h), windowClass(h), true
}
func windowTitle(h windows.HWND) string {
const max = 512
buf := make([]uint16, max)
n, _, _ := procGetWindowTextW.Call(
uintptr(h),
uintptr(unsafe.Pointer(&buf[0])),
uintptr(max),
)
return windows.UTF16ToString(buf[:n])
}
func windowClass(h windows.HWND) string {
var pid uint32
if _, err := windows.GetWindowThreadProcessId(h, &pid); err != nil || pid == 0 {
return ""
}
proc, err := windows.OpenProcess(windows.PROCESS_QUERY_LIMITED_INFORMATION, false, pid)
if err != nil {
return ""
}
defer windows.CloseHandle(proc)
buf := make([]uint16, windows.MAX_PATH)
size := uint32(len(buf))
if err := windows.QueryFullProcessImageName(proc, 0, &buf[0], &size); err != nil {
return ""
}
return ClassFromImagePath(windows.UTF16ToString(buf[:size]))
}
// MinimizeForeground minimizes the current foreground window. It returns nil
// when nothing is focused. ShowWindow's BOOL return reports prior visibility,
// not success, so there is nothing to error-check; minimize is best-effort by
// the Guard contract.
func MinimizeForeground() error {
h := windows.GetForegroundWindow()
if h == 0 {
return nil
}
procShowWindow.Call(uintptr(h), uintptr(windows.SW_MINIMIZE))
return nil
}
```
- [ ] **Step 2: Promote the dependency and verify the package cross-compiles**
Run:
```bash
go mod tidy
GOOS=windows GOARCH=amd64 go build ./internal/winapi/
```
Expected: both succeed with no output. `go.mod` now lists `golang.org/x/sys` in the direct (non-`// indirect`) require block.
- [ ] **Step 3: Verify the package still builds on Linux (pure file only)**
Run: `go test ./internal/winapi/`
Expected: PASS — on Linux the package is just `class.go` + its test; `winapi.go` is excluded by its build tag.
- [ ] **Step 4: Commit**
```bash
git add internal/winapi/winapi.go go.mod go.sum
git commit -m "Add Win32 binding for foreground window and minimize"
```
---
## Task 4: Windows evidence.Source adapter (cross-compile verified)
The polling sensor. Emits one snapshot immediately, then on every 750ms tick
emits only when the foreground window or title changed (via `foregroundTracker`
from Task 2 and `winapi.ForegroundWindow` from Task 3). Also narrows the
non-Linux fallback so it no longer claims Windows.
**Files:**
- Create: `internal/evidence/windows.go`
- Modify: `internal/evidence/source_other.go:1`
- [ ] **Step 1: Narrow the fallback build tag**
In `internal/evidence/source_other.go`, change the first line:
```go
//go:build !linux && !windows
```
(from `//go:build !linux`). Leave the rest of the file unchanged.
- [ ] **Step 2: Write the Windows source**
Create `internal/evidence/windows.go`:
```go
//go:build windows
package evidence
import (
"context"
"time"
"antidrift/internal/winapi"
)
// pollInterval is how often the Windows sensor samples the foreground window.
// ~1s latency on a window switch is immaterial for a focus tracker, and polling
// avoids the message-loop/callback machinery a SetWinEventHook source needs.
const pollInterval = 750 * time.Millisecond
// NewSource returns the Windows active-window sensor (polling).
func NewSource() Source { return windowsSource{} }
type windowsSource struct{}
// Watch emits the current window immediately, then samples every pollInterval,
// emitting only when the foreground window or its title changes. A read with no
// foreground window yields an Unavailable snapshot (once, until it recovers). It
// runs until ctx is cancelled. It never panics the daemon.
func (windowsSource) Watch(ctx context.Context, onChange func(WindowSnapshot)) {
var tr foregroundTracker
poll := func() {
hwnd, title, class, ok := winapi.ForegroundWindow()
if !tr.changed(ok, hwnd, title) {
return
}
if !ok {
onChange(WindowSnapshot{Health: EvidenceHealth{Available: false, Reason: "no foreground window"}})
return
}
onChange(WindowSnapshot{Title: title, Class: class, Health: EvidenceHealth{Available: true}})
}
poll() // immediate current window
ticker := time.NewTicker(pollInterval)
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return
case <-ticker.C:
poll()
}
}
}
```
- [ ] **Step 3: Verify Windows build and Linux tests**
Run:
```bash
GOOS=windows GOARCH=amd64 go build ./internal/evidence/
go test ./internal/evidence/
```
Expected: the Windows build succeeds with no output; the Linux test run passes (tracker test included, X11/other files unaffected).
- [ ] **Step 4: Verify the fallback world is intact (macOS still no-op)**
Run: `GOOS=darwin GOARCH=amd64 go build ./internal/evidence/`
Expected: succeeds — `source_other.go` (`!linux && !windows`) still provides `NewSource` on darwin.
- [ ] **Step 5: Commit**
```bash
git add internal/evidence/windows.go internal/evidence/source_other.go
git commit -m "Add Windows polling active-window sensor"
```
---
## Task 5: Windows enforce.Guard adapter (cross-compile verified)
The minimize guard. Thin wrapper over `winapi.MinimizeForeground`. Narrows the
non-Linux fallback the same way as Task 4.
**Files:**
- Create: `internal/enforce/windows.go`
- Modify: `internal/enforce/guard_other.go:1`
- [ ] **Step 1: Narrow the fallback build tag**
In `internal/enforce/guard_other.go`, change the first line:
```go
//go:build !linux && !windows
```
(from `//go:build !linux`). Leave the rest of the file unchanged.
- [ ] **Step 2: Write the Windows guard**
Create `internal/enforce/windows.go`:
```go
//go:build windows
package enforce
import (
"context"
"antidrift/internal/winapi"
)
// NewGuard returns the Windows window-minimize guard.
func NewGuard() Guard { return windowsGuard{} }
type windowsGuard struct{}
// MinimizeActive minimizes the current foreground window. It is best-effort:
// with nothing focused it does nothing and returns nil. Stateless and per-call,
// mirroring the short-lived X11 connection model.
func (windowsGuard) MinimizeActive(context.Context) error {
return winapi.MinimizeForeground()
}
```
- [ ] **Step 3: Verify Windows build and the fallback world**
Run:
```bash
GOOS=windows GOARCH=amd64 go build ./internal/enforce/
GOOS=darwin GOARCH=amd64 go build ./internal/enforce/
go build ./internal/enforce/
```
Expected: all three succeed with no output (Windows uses `windows.go`; darwin and Linux-host build paths still resolve `NewGuard`).
- [ ] **Step 4: Commit**
```bash
git add internal/enforce/windows.go internal/enforce/guard_other.go
git commit -m "Add Windows window-minimize guard"
```
---
## Task 6: Full cross-compilation gate and regression check
The primary automated guarantee for the syscall-bound code: the whole module
cross-compiles for Windows, the fallback world is intact for macOS, and the
Linux host build and full test suite are still green.
**Files:** none (verification only).
- [ ] **Step 1: Whole-module Windows cross-compile**
Run: `GOOS=windows GOARCH=amd64 go build ./...`
Expected: succeeds with no output. (This is the headline check: `cmd/antidriftd`, `evidence`, `enforce`, `winapi`, and every consumer link for Windows.)
- [ ] **Step 2: Whole-module macOS cross-compile (fallback intact)**
Run: `GOOS=darwin GOARCH=amd64 go build ./...`
Expected: succeeds with no output — proves the `!linux && !windows` tag edits did not orphan `NewSource`/`NewGuard` on other platforms.
- [ ] **Step 3: Linux host build and full test suite**
Run:
```bash
go build ./...
go test ./...
```
Expected: build succeeds; all tests pass, including the new `winapi` and `evidence` pure-logic tests, with the existing Linux X11 integration tests unaffected.
- [ ] **Step 4: Confirm the Windows binary actually links (optional sanity)**
Run: `GOOS=windows GOARCH=amd64 go build -o /tmp/antidriftd.exe ./cmd/antidriftd && ls -l /tmp/antidriftd.exe`
Expected: a `.exe` is produced.
- [ ] **Step 5: Commit (if `go mod tidy` left any tidy changes)**
```bash
git add -A
git commit -m "Verify Windows cross-compile and Linux regression for Windows support" --allow-empty
```
---
## Deferred manual verification (when a Windows 11 machine is available)
Not part of this plan's automated gates — recorded for whoever runs it on real
hardware later (from the spec):
- Run `antidriftd`; confirm the browser opens and the live view shows the
current window title and class.
- Switch windows and change a browser tab; confirm snapshots update and health
reads available.
- Start a commitment with "Enforce focus" armed; focus an off-task window;
confirm the drift judge fires and the window minimizes.
---
## Self-review notes
- **Spec coverage:** both ports (Task 4 sensor, Task 5 guard); `winapi` binding (Task 3); build-tag narrowing (Tasks 45); `go mod tidy` promotion (Task 3); pure-logic TDD for class + change-detection (Tasks 12); cross-compile + darwin-fallback + Linux-regression gates (Task 6); deferred manual steps recorded. No fabricated syscall mocks. The `Class` = exe-base-name decision is implemented in Task 1.
- **Type consistency:** `ClassFromImagePath` (Tasks 1, 3); `foregroundTracker.changed(available, hwnd, title)` with `hwnd uintptr` (Tasks 2, 4); `winapi.ForegroundWindow() (uintptr, string, string, bool)` and `winapi.MinimizeForeground() error` (Tasks 3, 4, 5). Module path `antidrift` confirmed against `go.mod`.
@@ -0,0 +1,246 @@
# Windows 11 Support — Design
**Date:** 2026-06-02
**Status:** Approved, ready for implementation planning
**Scope:** Full feature parity on Windows 11 — both OS ports (active-window
sensing and window-minimize enforcement) — implemented as `//go:build windows`
adapters behind the existing interfaces. No consumer code changes.
## Problem
AntiDrift's value is making drift visible by watching the active window and, on
confirmed off-task drift, minimizing it. Both capabilities are X11-only, gated
behind `//go:build linux`. On any non-Linux target the build links the
`//go:build !linux` no-op fallbacks, so the daemon starts and serves its UI but
the entire focus loop is dead: no evidence, no drift detection, no nudges, no
enforcement.
Today's Windows behavior, verified:
- `GOOS=windows GOARCH=amd64 go build ./cmd/antidriftd` compiles cleanly (no
cgo).
- The daemon would launch and open a browser (`cmd/antidriftd/main.go` already
handles the `windows` case via `rundll32`).
- `evidence.NewSource()` returns a no-op reporting "no active-window sensor on
this platform" (`internal/evidence/source_other.go`).
- `enforce.NewGuard()` returns a no-op whose `MinimizeActive` does nothing
(`internal/enforce/guard_other.go`).
This design fills both ports on Windows so AntiDrift functions end-to-end.
## Goals
- A real `evidence.Source` on Windows: emit `WindowSnapshot{Title, Class,
Health}` on foreground-window and title changes.
- A real `enforce.Guard` on Windows: minimize the foreground window on demand.
- Pure Go, no cgo. Cross-compilation from Linux stays clean.
- No changes to `session`, `web`, `domain`, or `cmd` — the work lives entirely
behind the two existing ports.
## Non-goals (YAGNI)
- Event-driven sensing via `SetWinEventHook` (see Approach B, rejected below).
The `Source` interface hides the polling-vs-event choice; B can replace A
later with zero consumer impact if latency ever matters.
- Windows ARM64.
- Installer, system tray, packaging, autostart.
- Any session-policy or web/UI changes.
## Constraints that shaped this design
- **Verification is compile-only for now.** Development is on Manjaro Linux with
no Windows 11 machine available. The design therefore minimizes untested
syscall surface: no callbacks, no Win32 message loop, no OS-thread affinity.
Live runtime verification is deferred (see Testing).
## Port contracts being satisfied
From `internal/evidence/evidence.go`:
```go
type WindowSnapshot struct {
Title string // full window title
Class string // app identity, matched case-folded against allowed classes
Health EvidenceHealth
}
type Source interface {
// Watch runs until ctx is cancelled, invoking onChange on every
// active-window change, and once immediately with the current window.
Watch(ctx context.Context, onChange func(WindowSnapshot))
}
```
From `internal/enforce/enforce.go`:
```go
type Guard interface {
// MinimizeActive minimizes the currently-focused window. Idempotent and
// best-effort: returns an error for diagnostics; callers never block on it.
MinimizeActive(ctx context.Context) error
}
```
The `Class` field is the authoritative on-task signal. `MatchesAllowed`
(`internal/evidence/context.go`) compares it **case-folded, exact match**
against the session's allowed window classes (or matches a title substring).
## Architecture
All new code is behind `//go:build windows`. No consumer changes — `session`,
`web`, and `cmd/antidriftd` already call `evidence.NewSource()` /
`enforce.NewGuard()` and degrade on the health flags.
### New files
- `internal/winapi/winapi.go` (`//go:build windows`) — a small shared binding
layer over the Win32 calls, so the two adapters don't each redeclare the same
procs.
- `internal/evidence/windows.go` (`//go:build windows`) — the polling `Source`.
- `internal/enforce/windows.go` (`//go:build windows`) — the `ShowWindow`
`Guard`.
### Build-tag edit (correctness-critical)
The fallbacks are currently tagged `//go:build !linux`, which is what compiles
on Windows today. Once `windows.go` files exist in those packages, the fallbacks
must exclude Windows too, or the build gets duplicate `NewSource`/`NewGuard`
symbols:
- `internal/evidence/source_other.go`: `//go:build !linux` →
`//go:build !linux && !windows`
- `internal/enforce/guard_other.go`: `//go:build !linux` →
`//go:build !linux && !windows`
This yields three mutually exclusive worlds per port: `linux` (X11), `windows`
(new), everything-else (no-op). macOS and any other GOOS remain no-ops, exactly
as today.
### Dependencies
`golang.org/x/sys` is already in `go.mod` (indirect, v0.41.0). `go mod tidy`
promotes it to a direct require. No new module, no cgo.
## Component design
### `internal/winapi` — shared Win32 binding
Most of the needed calls are already typed wrappers in `x/sys/windows` (verified
under `GOOS=windows go doc`), so the hand-written binding is deliberately tiny.
Provided by `golang.org/x/sys/windows`, used directly — no LazyDLL needed:
- `GetForegroundWindow() HWND`
- `GetWindowThreadProcessId(hwnd HWND, *uint32) (tid uint32, err error)` → owning PID
- `OpenProcess`, `QueryFullProcessImageName`, `CloseHandle` → process image path
- the `windows.HWND` type and the `windows.SW_MINIMIZE` (= 6) constant
Not wrapped by `x/sys/windows`; loaded once via
`windows.NewLazySystemDLL("user32.dll")` and called through `proc.Call`:
- `GetWindowTextW(hwnd, *uint16, max int32) int32` → title
- `ShowWindow(hwnd, SW_MINIMIZE) bool`
Public surface (two helpers the adapters consume):
- `ForegroundWindow() (hwnd uintptr, title, class string, ok bool)` — resolves
the foreground window's title and process-exe-base `class` in one call. `ok`
is false when there is no foreground window (null hwnd) — e.g. secure desktop,
UAC prompt, lock screen.
- `MinimizeForeground() error` — minimizes the current foreground window;
returns nil when nothing is focused.
`class` is derived from the owning process image path
(`QueryFullProcessImageName`) as the **base name minus the `.exe` extension,
lowercased**: `C:\Program Files\Microsoft VS Code\Code.exe` → `code`. This is
the closest analog to X11 `WM_CLASS` conventions and keeps allowed-class lists
readable. Path parsing is a pure function (see Testing).
### `evidence.Source` (Windows) — polling
`Watch(ctx, onChange)`:
1. Emit one snapshot immediately (honors the "once immediately" contract).
2. `time.NewTicker(750 * time.Millisecond)`.
3. On each tick, call `winapi.ForegroundWindow()` and track the last
`(hwnd, title)`. Invoke `onChange` **only when `hwnd` or `title` changed** —
a steady window stays silent, and a same-window title change (e.g. a browser
tab switch) still fires. This reproduces the fidelity the X11 source gets
from `_NET_ACTIVE_WINDOW` plus name property notifies.
4. Map a successful read to
`WindowSnapshot{Title: title, Class: class, Health: {Available: true}}`.
5. Map `ok == false` to
`WindowSnapshot{Health: {Available: false, Reason: "no foreground window"}}`.
Recovers automatically on the next tick.
6. On `ctx.Done()`, stop the ticker and return. Nothing else to tear down.
Rationale for polling over `SetWinEventHook`: under compile-only verification,
the responsible choice is the design with no callback trampoline
(`syscall.NewCallback`), no `GetMessage`/`DispatchMessage` loop, and no
`runtime.LockOSThread` — three failure modes that are hard to validate without a
Windows machine. ~1s switch latency is irrelevant for a focus tracker, and
polling also catches in-window title changes for free.
### `enforce.Guard` (Windows)
`MinimizeActive(ctx)`:
1. Read the foreground window. If null, return nil (nothing focused — same as
the X11 `active == 0` case).
2. Otherwise `ShowWindow(hwnd, SW_MINIMIZE)`; on failure return a wrapped error.
Per-call and stateless, mirroring the short-lived X11 connection model. Errors
are for the caller to log; the caller never blocks on enforcement.
## Error handling and degradation
The existing contracts are unchanged; this design simply honors them on Windows.
- Sensor can't read a foreground window → `Available: false` snapshot with a
reason. The daemon already treats this as "no evidence this tick" and surfaces
the reason in the live view. Auto-recovers on the next tick.
- `ShowWindow` fails → error returned and logged upstream; enforcement is
best-effort, treated as "did nothing this time." No panic, no block.
- No new failure types reach `session`/`web`: they still see only a
`WindowSnapshot` and a possible `MinimizeActive` error, exactly as with X11.
## Known cross-platform nuance (not a bug)
The same application can produce a different `Class` on Linux vs Windows (X11
`WM_CLASS` vs Windows exe base name; e.g. Chrome may be `google-chrome` on Linux
and `chrome` on Windows). Allowed-class lists are per-session and case-folded,
so this only affects portability of class names *across machines*, never
matching within a single session. Documented, not engineered around.
## Testing and verification
Bounded honestly by the compile-only constraint.
1. **Cross-compilation gate (primary automated guarantee).**
`GOOS=windows GOARCH=amd64 go build ./...` must pass. Also build
`GOOS=darwin` to prove the `!linux && !windows` tag edit still selects the
no-op fallback (i.e. the everything-else world is intact).
2. **Pure-logic unit tests, no build tag (run on Linux).** The syscall-free
logic is extracted into pure functions and tested directly:
- exe-path → class normalization: `Code.exe` → `code`, strip `.exe`,
lowercase, handle no-extension names and UNC/odd separators.
- change-detection predicate: emit on `(hwnd, title)` change, stay silent
otherwise.
3. **No fabricated syscall mocks.** We will not write a fake that pretends to
exercise `user32` — that produces false confidence. Live verification is
deferred until a Windows 11 machine is available.
4. **`x11_integration_test.go` files stay `//go:build linux`**, untouched.
### Deferred manual verification (when a Windows 11 machine is available)
- Run `antidriftd`; confirm the browser opens and the live view shows the
current window title/class.
- Switch windows and change a browser tab; confirm snapshots update and the
health reads available.
- Start a commitment with "Enforce focus" armed; focus an off-task window;
confirm the drift judge fires and the window minimizes.
## Out of scope (restated)
`SetWinEventHook`, Windows ARM64, tray/installer/packaging/autostart, and any
session/web/domain changes. This is pure adapter work behind two existing ports.
+1 -1
View File
@@ -7,6 +7,7 @@ require (
github.com/google/uuid v1.6.0
github.com/jezek/xgb v1.3.1
github.com/jezek/xgbutil v0.0.0-20260124183602-9fd151d6a51a
golang.org/x/sys v0.41.0
)
require (
@@ -36,7 +37,6 @@ require (
golang.org/x/arch v0.22.0 // indirect
golang.org/x/crypto v0.48.0 // indirect
golang.org/x/net v0.51.0 // indirect
golang.org/x/sys v0.41.0 // indirect
golang.org/x/text v0.34.0 // indirect
google.golang.org/protobuf v1.36.10 // indirect
)
+1 -1
View File
@@ -1,4 +1,4 @@
//go:build !linux
//go:build !linux && !windows
package enforce
+22
View File
@@ -0,0 +1,22 @@
//go:build windows
package enforce
import (
"context"
"antidrift/internal/winapi"
)
// NewGuard returns the Windows window-minimize guard.
func NewGuard() Guard { return windowsGuard{} }
type windowsGuard struct{}
// MinimizeActive minimizes the current foreground window. It is best-effort:
// with nothing focused it does nothing and returns nil. Stateless and per-call,
// with no connection or shared state to manage. ShowWindow does not report a
// minimize failure, so this always returns nil today.
func (windowsGuard) MinimizeActive(_ context.Context) error {
return winapi.MinimizeForeground()
}
+23
View File
@@ -0,0 +1,23 @@
package evidence
// foregroundTracker remembers the last observed foreground window so the
// Windows polling Source emits only on change. hwnd is held as uintptr so this
// file stays platform-neutral (it must build and test on Linux).
type foregroundTracker struct {
primed bool
available bool
hwnd uintptr
title string
}
// changed reports whether (available, hwnd, title) differs from the last
// observation and records the new values. The first call always returns true.
// Callers pass the zero values (hwnd 0, title "") when available is false, so a
// steady "no foreground window" run does not re-emit.
func (t *foregroundTracker) changed(available bool, hwnd uintptr, title string) bool {
if t.primed && available == t.available && hwnd == t.hwnd && title == t.title {
return false
}
t.primed, t.available, t.hwnd, t.title = true, available, hwnd, title
return true
}
@@ -0,0 +1,29 @@
package evidence
import "testing"
func TestForegroundTrackerChanged(t *testing.T) {
var tr foregroundTracker
if !tr.changed(true, 100, "A") {
t.Fatal("first observation should always report changed")
}
if tr.changed(true, 100, "A") {
t.Error("identical observation should not report changed")
}
if !tr.changed(true, 100, "B") {
t.Error("title change on same hwnd should report changed")
}
if !tr.changed(true, 200, "B") {
t.Error("hwnd change should report changed")
}
if !tr.changed(false, 0, "") {
t.Error("transition to unavailable should report changed")
}
if tr.changed(false, 0, "") {
t.Error("repeated unavailable should not report changed")
}
if !tr.changed(true, 200, "B") {
t.Error("transition back to available should report changed")
}
}
+1 -1
View File
@@ -1,4 +1,4 @@
//go:build !linux
//go:build !linux && !windows
package evidence
+10
View File
@@ -0,0 +1,10 @@
package evidence
import "log"
// unavailable builds an Unavailable snapshot and logs the reason. It is shared
// by every platform sensor so an unobservable window is recorded consistently.
func unavailable(reason string) WindowSnapshot {
log.Printf("evidence: %s", reason)
return WindowSnapshot{Health: EvidenceHealth{Available: false, Reason: reason}}
}
+51
View File
@@ -0,0 +1,51 @@
//go:build windows
package evidence
import (
"context"
"time"
"antidrift/internal/winapi"
)
// pollInterval is how often the Windows sensor samples the foreground window.
// ~1s latency on a window switch is immaterial for a focus tracker, and polling
// avoids the message-loop/callback machinery a SetWinEventHook source needs.
const pollInterval = 750 * time.Millisecond
// NewSource returns the Windows active-window sensor (polling).
func NewSource() Source { return windowsSource{} }
type windowsSource struct{}
// Watch emits the current window immediately, then samples every pollInterval,
// emitting only when the foreground window or its title changes. A read with no
// foreground window yields an Unavailable snapshot (once, until it recovers). It
// runs until ctx is cancelled. It never panics the daemon.
func (windowsSource) Watch(ctx context.Context, onChange func(WindowSnapshot)) {
var tr foregroundTracker
poll := func() {
hwnd, title, class, ok := winapi.ForegroundWindow()
if !tr.changed(ok, hwnd, title) {
return
}
if !ok {
onChange(unavailable("no foreground window"))
return
}
onChange(WindowSnapshot{Title: title, Class: class, Health: EvidenceHealth{Available: true}})
}
poll() // immediate current window
ticker := time.NewTicker(pollInterval)
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return
case <-ticker.C:
poll()
}
}
}
-5
View File
@@ -4,7 +4,6 @@ package evidence
import (
"context"
"log"
"github.com/jezek/xgb/xproto"
"github.com/jezek/xgbutil"
@@ -88,7 +87,3 @@ func snapshot(X *xgbutil.XUtil) WindowSnapshot {
}
}
func unavailable(reason string) WindowSnapshot {
log.Printf("evidence: %s", reason)
return WindowSnapshot{Health: EvidenceHealth{Available: false, Reason: reason}}
}
+22
View File
@@ -0,0 +1,22 @@
// Package winapi is the Windows Win32 binding layer for AntiDrift's OS ports.
// The syscall-bound code lives in windows-tagged files; this untagged file
// holds the pure path logic so it builds and is tested on every platform.
package winapi
import "strings"
// ClassFromImagePath derives the on-task class identity from a process image
// path: the base file name with any trailing ".exe" removed (case-insensitive),
// lowercased. It is the Windows analog of an X11 WM_CLASS. It parses both `\`
// and `/` separators so it is correct regardless of the host OS running the
// test. Returns "" for an empty path.
func ClassFromImagePath(p string) string {
if i := strings.LastIndexAny(p, `\/`); i >= 0 {
p = p[i+1:]
}
const exeSuffix = ".exe"
if len(p) >= len(exeSuffix) && strings.EqualFold(p[len(p)-len(exeSuffix):], exeSuffix) {
p = p[:len(p)-len(exeSuffix)]
}
return strings.ToLower(p)
}
+22
View File
@@ -0,0 +1,22 @@
package winapi
import "testing"
func TestClassFromImagePath(t *testing.T) {
cases := []struct{ in, want string }{
{`C:\Program Files\Microsoft VS Code\Code.exe`, "code"},
{`C:\Windows\explorer.exe`, "explorer"},
{`chrome.exe`, "chrome"},
{`C:\x\FOO.EXE`, "foo"},
{`C:\x\My.App.exe`, "my.app"},
{`C:/forward/slash/Code.exe`, "code"},
{`firefox`, "firefox"},
{``, ""},
{`.exe`, ""}, // degenerate: only the extension
}
for _, c := range cases {
if got := ClassFromImagePath(c.in); got != c.want {
t.Errorf("ClassFromImagePath(%q) = %q, want %q", c.in, got, c.want)
}
}
}
+75
View File
@@ -0,0 +1,75 @@
//go:build windows
package winapi
import (
"unsafe"
"golang.org/x/sys/windows"
)
var (
user32 = windows.NewLazySystemDLL("user32.dll")
procGetWindowTextW = user32.NewProc("GetWindowTextW")
procShowWindow = user32.NewProc("ShowWindow")
)
// ForegroundWindow returns the current foreground window's handle (as a uintptr
// so platform-neutral callers need not import windows), its title, and its
// on-task class (process exe base name; see ClassFromImagePath). ok is false
// when GetForegroundWindow returns null, which happens transiently during focus
// changes and some secure-desktop transitions (UAC / lock screen).
func ForegroundWindow() (hwnd uintptr, title, class string, ok bool) {
h := windows.GetForegroundWindow()
if h == 0 {
return 0, "", "", false
}
return uintptr(h), windowTitle(h), windowClass(h), true
}
func windowTitle(h windows.HWND) string {
// 512 uint16s; titles over 511 chars are truncated — acceptable for a sensor tag.
const max = 512
buf := make([]uint16, max)
n, _, _ := procGetWindowTextW.Call(
uintptr(h),
uintptr(unsafe.Pointer(&buf[0])),
uintptr(max),
)
// n==0 covers both call failure and a genuinely empty title; both are fine here.
return windows.UTF16ToString(buf[:n])
}
func windowClass(h windows.HWND) string {
var pid uint32
if _, err := windows.GetWindowThreadProcessId(h, &pid); err != nil || pid == 0 {
return ""
}
proc, err := windows.OpenProcess(windows.PROCESS_QUERY_LIMITED_INFORMATION, false, pid)
if err != nil {
return ""
}
defer windows.CloseHandle(proc)
buf := make([]uint16, windows.MAX_PATH)
size := uint32(len(buf))
// MAX_PATH suffices for almost all paths; an over-long path fails here and
// degrades windowClass to "" (window left unclassified), never a crash.
if err := windows.QueryFullProcessImageName(proc, 0, &buf[0], &size); err != nil {
return ""
}
return ClassFromImagePath(windows.UTF16ToString(buf[:size]))
}
// MinimizeForeground minimizes the current foreground window. It returns nil
// when nothing is focused. ShowWindow's BOOL return reports prior visibility,
// not success, so there is nothing to error-check; minimize is best-effort by
// the Guard contract.
func MinimizeForeground() error {
h := windows.GetForegroundWindow()
if h == 0 {
return nil
}
procShowWindow.Call(uintptr(h), uintptr(windows.SW_MINIMIZE))
return nil
}