feat(update): replace cmd.exe handoff with native Win32 spawn

Both the in-app restart and the auto-update install previously
shelled out to cmd.exe so the new instance could wait for the old
one to release the singleton mutex and the locked exe file. On
some Windows configurations the `start ""` inside `cmd /c ...` can
flash a console window despite CREATE_NO_WINDOW + DETACHED_PROCESS
flags. The replacement spawns the child binary directly via
CreateProcessW; since the main exe is built with
windows_subsystem = "windows", no console is ever allocated.

- New `src/update/handoff.rs` exposes `spawn_detached`,
  `wait_for_parent_exit`, and `cleanup_stale_old_exes`.
- New CLI flags `--wait-pid <pid>` and `--updated-to <version>`
  parsed early in `main`; the child waits up to 5s on the parent
  PID via OpenProcess+WaitForSingleObject before falling through
  to a 3s mutex-acquisition retry.
- `restart_app` and `install::begin` both spawn detached children
  using the new helper.
- Update install now uses MoveFileExW twice (rename running exe
  sideways, then move staged exe into place with
  MOVEFILE_REPLACE_EXISTING | MOVEFILE_COPY_ALLOWED so portable
  installs on non-system drives still work). Rollback restores
  the backup if either the swap OR the post-swap detached spawn
  fails, and a MessageBoxW modal surfaces the backup path if the
  rollback itself fails.
- First launch after an auto-update shows a blue-info tray
  balloon "Updated to vX.Y.Z" via a new `tray::notify_info` (the
  existing `tray::notify` is split into `notify_warning` +
  `notify_info` sharing a `notify_inner`).
- Startup sweeps stale `<exe>.old.<pid>` siblings left by past
  in-place updates.
- Three new `LocaleStrings` fields translated across all 8
  supported locales (en/nl/es/fr/de/ja/ko/zh-TW).
This commit is contained in:
2026-05-21 16:28:25 +07:00
parent 1ba2883989
commit 1cd5b778f4
23 changed files with 1306 additions and 112 deletions
+1
View File
@@ -29,6 +29,7 @@ features = [
"Win32_UI_WindowsAndMessaging",
"Win32_System_Registry",
"Win32_System_Threading",
"Win32_Storage_FileSystem",
"Win32_Security",
"Win32_UI_HiDpi",
"Win32_UI_Input_KeyboardAndMouse",
@@ -0,0 +1,104 @@
---
phase: 1
title: "Foundation: CLI flags + native spawn helper + mutex retry"
status: complete
priority: P1
effort: "3h"
dependencies: []
---
# Phase 1: Foundation: CLI flags + native spawn helper + mutex retry
## Overview
Build the primitives that phases 2-4 reuse: a CLI argument parser for `--wait-pid <pid>` and `--updated-to <version>`, a `spawn_detached_self` helper that calls `CreateProcessW` directly (no cmd.exe), and mutex-acquisition retry logic that activates only when `--wait-pid` was passed.
## Requirements
**Functional**
- Parse `--wait-pid <u32>` and `--updated-to <version-string>` from `std::env::args` without breaking existing flags (`--diagnose`, `--apply-update`).
- Expose `spawn_detached(exe: &Path, args: &[OsString]) -> io::Result<()>` that uses `CreateProcessW` with `CREATE_NO_WINDOW | DETACHED_PROCESS | CREATE_NEW_PROCESS_GROUP`.
- Expose `wait_for_parent_exit(pid: u32, timeout_ms: u32)` that uses `OpenProcess(SYNCHRONIZE)` + `WaitForSingleObject`. Returns silently on timeout (best-effort).
- Mutex acquisition in `app::run` retries `CreateMutexW` for ~3 seconds (200ms backoff) ONLY when `--wait-pid` was present; preserves today's immediate-fail behavior for normal startup.
**Non-functional**
- No new external dependencies. All Win32 calls via the existing `windows = "0.58"` crate features.
- Helper module ≤ ~120 lines total.
## Architecture
```
src/
├── update/
│ └── handoff.rs ← NEW: spawn_detached, wait_for_parent_exit, cleanup_stale_old_exes
├── main.rs ← parse --wait-pid early, call wait_for_parent_exit BEFORE app::run
└── app.rs ← run() reads a static "wait_pid_was_passed" flag, retries mutex if set
```
Rationale for `src/update/handoff.rs`: keeps low-level Win32 process/file ops alongside the update module that uses them most. `app.rs` and `main.rs` import it for restart + post-update bootstrap.
## Related Code Files
- **Create**: `src/update/handoff.rs`
- **Modify**: `src/main.rs` (early arg parse + wait + flag handoff to app)
- **Modify**: `src/update/mod.rs` (declare `pub mod handoff`)
- **Modify**: `src/app.rs` (mutex retry loop, gated on flag from main)
- **Modify**: `Cargo.toml` if a new `windows` feature is needed (likely `Win32_System_Threading` already covers `OpenProcess`/`WaitForSingleObject`/`CreateProcessW`)
## Implementation Steps
1. **Audit `windows` crate features.** Confirm `Win32_System_Threading` is in `Cargo.toml` (it is — line 31). Verify `CreateProcessW`, `STARTUPINFOW`, `PROCESS_INFORMATION` are accessible. Add `Win32_Storage_FileSystem` if not already present (needed for phase 3's `MoveFileExW`).
2. **Create `src/update/handoff.rs`** with three pub fns:
- `pub fn spawn_detached(exe: &Path, args: &[OsString]) -> io::Result<()>`
- Build a wide-char command line: quoted exe path + space-joined args, NUL-terminated.
- `STARTUPINFOW` zero-initialized, `cb` set.
- `CreateProcessW(NULL, cmdline_wide.as_mut_ptr(), NULL, NULL, FALSE, CREATE_NO_WINDOW | DETACHED_PROCESS | CREATE_NEW_PROCESS_GROUP, NULL, NULL, &si, &pi)`.
- Close `pi.hProcess` and `pi.hThread` immediately (fire-and-forget).
- `pub fn wait_for_parent_exit(pid: u32, timeout_ms: u32)`
- `OpenProcess(SYNCHRONIZE, FALSE, pid)`. If it fails (parent already gone), return immediately.
- `WaitForSingleObject(h, timeout_ms)`. Ignore return value.
- `CloseHandle(h)`.
- `pub fn cleanup_stale_old_exes(current_exe: &Path)` (used by phase 4; stub here, fill in phase 4)
- Stub: returns `Ok(())`.
3. **Modify `src/update/mod.rs`** to add `pub mod handoff;`.
4. **Modify `src/main.rs`** — insert BEFORE `app::run()`:
```rust
let wait_pid = args.iter()
.position(|a| a == "--wait-pid")
.and_then(|i| args.get(i + 1))
.and_then(|s| s.parse::<u32>().ok());
if let Some(pid) = wait_pid {
update::handoff::wait_for_parent_exit(pid, 5_000);
}
```
Note: keep this AFTER `update::run_cli` so the legacy `--apply-update` path still short-circuits cleanly.
5. **Modify `src/app.rs::run`** — gate mutex retry on whether `--wait-pid` appeared. Cheapest implementation: re-parse `std::env::args` once at the top of `run()`. Then around line 156:
```rust
let retry_mutex = std::env::args().any(|a| a == "--wait-pid");
let _mutex = acquire_singleton_mutex(retry_mutex)?; // new helper
```
New helper `acquire_singleton_mutex(retry: bool)`:
- If `!retry`: today's behavior (fail immediately on `ERROR_ALREADY_EXISTS`).
- If `retry`: loop CreateMutexW → on `ALREADY_EXISTS`, `Sleep(200)` and retry. Budget 15 iterations = ~3 seconds. Log every retry. After budget exhausted, return error and exit cleanly.
6. **Compile check.** `cargo build --release`. Fix any feature gaps. No behavior should change yet — `--wait-pid` arg is parsed but no caller passes it yet.
## Success Criteria
- [ ] `cargo build --release` succeeds with zero warnings beyond the existing `dead_code` allow.
- [ ] Running the binary normally (no flags) behaves identically to today (mutex check is immediate, no retry).
- [ ] Running the binary with `--wait-pid <pid-of-running-instance>` against a live instance: the new process waits ≤5s for old one to exit, then acquires the mutex within ~200ms of its release. Verifiable by killing the original after 2s and watching the new one continue.
- [ ] `update::handoff::spawn_detached` smoke test: from a small one-off snippet in `main` (gated behind a never-used flag) verify CreateProcessW returns success and pid increments. Remove before commit.
## Risk Assessment
| Risk | Mitigation |
|---|---|
| Wide-char cmdline construction has off-by-one / missing NUL | Hand-test with a path containing spaces; assert `OsStringExt::encode_wide` produces expected bytes |
| `CREATE_NO_WINDOW | DETACHED_PROCESS` combination misbehaves on GUI subsystem binaries | Documented Windows behavior: for a GUI subsystem child, both flags are effectively no-ops (no console requested), but combining them is harmless. Tested by phase 5 |
| Mutex retry loop hangs forever if budget logic wrong | Hard cap = 15 iterations × 200ms = 3.0s. After that, exit. Add log line per retry so a stuck loop is visible in `--diagnose` |
| `OpenProcess(SYNCHRONIZE, ...)` returns access-denied for cross-session | Fallback: `WaitForSingleObject` simply isn't called; mutex retry compensates |
@@ -0,0 +1,113 @@
---
phase: 2
title: "Restart path: replace restart_app with native CreateProcessW"
status: complete
priority: P1
effort: "1h"
dependencies: [1]
---
# Phase 2: Restart path: replace restart_app with native CreateProcessW
## Overview
Replace `src/app.rs::restart_app`'s `cmd.exe /c "timeout & start ..."` handoff with a direct `spawn_detached(current_exe, ["--wait-pid", our_pid])` call. The new instance handles the parent-exit wait itself using phase 1's helper; no timer needed.
## Requirements
**Functional**
- Restart triggered from the tray menu produces zero console flash.
- New instance acquires `Global\ClaudeCodeUsageBubble` mutex successfully every time.
- Settings still flushed to disk before exit (preserve current `snap + save` defensive write).
- Path-with-`%` defense becomes unnecessary (no cmd.exe); the check is removed.
**Non-functional**
- `restart_app` function shrinks from ~40 lines to ~25 lines.
## Architecture
```
restart_app():
1. settings::save(snap) (unchanged)
2. exe = current_exe() (unchanged)
3. our_pid = GetCurrentProcessId()
4. handoff::spawn_detached(exe, [--wait-pid, our_pid])
5. on success: PostQuitMessage(0)
6. on failure: log error, do NOT quit
```
Mutex release is implicit on process exit — no explicit `ReleaseMutex` needed because the `_mutex` handle in `app::run` is dropped when `run()` returns after `PostQuitMessage`. `Drop` closes the handle, which releases the mutex.
## Related Code Files
- **Modify**: `src/app.rs::restart_app` (lines ~1378-1432)
- **Remove**: the `RESTART_CREATE_NO_WINDOW` / `RESTART_DETACHED_PROCESS` constants (now in handoff.rs)
- **Remove**: the `%`-rejection defense (no cmd.exe to exploit)
- **Remove**: the `replace('"', "")` quote-stripping (handoff.rs handles quoting)
## Implementation Steps
1. **Read current `restart_app`** at `src/app.rs:1378-1432` to confirm exact bounds.
2. **Rewrite `restart_app`** to:
```rust
fn restart_app() {
// Defensive settings flush (unchanged)
let snap = lock_state().as_ref().map(|s| s.settings.clone());
if let Some(s) = snap {
settings::save(&s);
}
let exe = match std::env::current_exe() {
Ok(p) => p,
Err(e) => {
log::error!("restart: current_exe failed: {e}");
return;
}
};
let pid = unsafe { GetCurrentProcessId() };
let args = vec![
OsString::from("--wait-pid"),
OsString::from(pid.to_string()),
];
match update::handoff::spawn_detached(&exe, &args) {
Ok(()) => {
log::info!("restart: spawned detached child, posting quit");
unsafe { PostQuitMessage(0) };
}
Err(e) => log::error!("restart: spawn failed: {e}"),
}
}
```
3. **Drop the two `RESTART_*` const declarations** above the function — they were specific to the old cmd.exe path. Phase 1's helper has its own.
4. **Drop the `%`-rejection block** in the new `restart_app`. The brainstorm doc keeps the equivalent check in `install.rs` for defense-in-depth; here it's pure dead weight without cmd.exe.
5. **Imports**: add `use std::ffi::OsString;` and `use windows::Win32::System::Threading::GetCurrentProcessId;` if not already in scope.
6. **Compile check**: `cargo build --release`.
7. **Manual smoke test**: run the binary, click Restart in the tray menu, verify:
- No console window flashes
- New instance appears within ~1s
- Old instance log shows "spawned detached child, posting quit"
- New instance log shows mutex acquired (via Phase 1's retry path)
## Success Criteria
- [ ] `cargo build --release` clean.
- [ ] Manual restart from menu produces ZERO visible console window across 20 consecutive triggers.
- [ ] New instance window appears within 1500ms of menu click.
- [ ] Settings file (`settings.json`) shows updated mtime after restart, confirming defensive save still runs.
- [ ] `restart_app` function is ≤25 lines.
## Risk Assessment
| Risk | Mitigation |
|---|---|
| Mutex handle not yet dropped when child tries to acquire | Phase 1's `--wait-pid` + 5s `WaitForSingleObject` + 3s mutex retry covers race comfortably (8s total budget vs <1s actual parent exit) |
| `PostQuitMessage` doesn't immediately exit; window-message pump may process more events | Phase 1 mutex retry tolerates up to 3s of overlap |
| `current_exe()` returns a path that the child can't load (rare: deleted exe, fileshare disconnect) | Existing behavior preserved: log error, do not quit. User can manually retry |
@@ -0,0 +1,156 @@
---
phase: 3
title: "Update install: rename + move + native spawn"
status: complete
priority: P1
effort: "2h"
dependencies: [1]
---
# Phase 3: Update install: rename + move + native spawn
## Overview
Replace `src/update/install.rs::begin`'s `cmd.exe /c "timeout & move & start ..."` handoff with native steps: `MoveFileExW` to rename the running exe sideways, `MoveFileExW` to move the staged exe into place, then `spawn_detached` of the new binary. Removes the only remaining cmd.exe invocation in the update flow.
## Requirements
**Functional**
- Update install produces zero console flash.
- New binary version starts after auto-update without user interaction.
- SHA-256 verification continues to gate the swap (no swap on checksum mismatch).
- On any failure step, the original exe must remain runnable (no half-state).
**Non-functional**
- `install.rs` net change ≈ -30 lines (cmd-quoting code is gone, replaced by short Win32 calls).
- New code paths use the windows crate; no new dependencies.
## Architecture
```
install::begin(http, release):
1. current = current_exe()
2. ensure_writable(current.parent()) (unchanged)
3. staging = stage_path()
4. reject_unsafe_path(current) (kept for defense-in-depth; harmless now)
5. reject_unsafe_path(staging)
6. create_dir_all(staging.parent())
7. download(http, asset_url, staging, sha256) (unchanged)
8. backup = current.with_file_name(format!("{}.old.{}", filename, pid))
9. MoveFileExW(current, backup, 0) ← NEW: rename running exe sideways
10. MoveFileExW(staging, current, MOVEFILE_REPLACE_EXISTING) ← NEW
11. our_pid = GetCurrentProcessId()
12. handoff::spawn_detached(current,
["--wait-pid", our_pid, "--updated-to", version_str]) ← NEW
13. return Ok(())
```
Caller (`app.rs` Apply action) is responsible for `PostQuitMessage` after `begin` returns Ok — same as today.
### Rollback semantics
| Step that failed | State | Recovery |
|---|---|---|
| 7 (download) | Original exe untouched | Existing behavior: error surfaced, user retries |
| 9 (rename current → backup) | Original exe untouched | Surface `Error::NotWritable`; do not proceed |
| 10 (move staging → current) | Original exe is at backup path, current path empty | Best-effort revert: rename backup back to current; surface error |
| 10 + revert (both fail) | Original at backup path; current path empty; user has no runnable binary at the install location | **Per Validation Session 1 decision:** show a Windows `MessageBoxW` (MB_OK \| MB_ICONERROR) telling the user where the backup is, then exit. Message: "Update failed. Your original binary is saved as `{backup_path}`. Please rename it back to `{exe_name}` manually." |
| 12 (spawn child) | New exe at correct path, but app didn't restart | Log + tray balloon "Update applied; restart manually". Rare — `CreateProcessW` on a fresh fully-written exe almost never fails |
<!-- Updated: Validation Session 1 - Rollback escalation MessageBox added -->
### Rollback escalation helper
Add a private `surface_rollback_failure(backup_path: &Path, target_name: &str)` helper that calls `MessageBoxW` with `MB_OK | MB_ICONERROR` and the localized message. Adds a new `LocaleStrings` field `update_rollback_failed_body` parameterized with `{backup_path}` and `{exe_name}` (Rust `format!` substitution at call site). The plain MessageBox uses the Win32 dialog, so no console can flash.
## Related Code Files
- **Modify**: `src/update/install.rs::begin`
- **Modify**: `src/update/install.rs::spawn_handoff` → REMOVED entirely
- **Modify**: `src/update/install.rs` imports (drop `os::windows::process::CommandExt`, `process::{Command, Stdio}`; add `MoveFileExW`, `MOVEFILE_REPLACE_EXISTING`, `GetCurrentProcessId`, `MessageBoxW`, `MB_OK`, `MB_ICONERROR`)
- **Modify**: `src/update/mod.rs::Error` — add `Error::SwapFailed(String)` variant if MoveFileExW failures don't fit existing variants cleanly
- **Modify**: `src/i18n/mod.rs::LocaleStrings` — add `update_rollback_failed_body: String` (also belongs to Phase 4 i18n group, but Phase 3 is the consumer)
## Implementation Steps
1. **Read current `install.rs::begin` + `spawn_handoff`** to confirm exact bounds (lines 19-36 + 99-121).
2. **Decide path-safety policy**: keep `reject_unsafe_path` (the `%`-check) as defense-in-depth even though no cmd.exe runs. Update the function-level comment to reflect new reality (kept for paranoia, not strict need).
3. **Add `swap_and_spawn` private helper** (replaces `spawn_handoff`):
```rust
fn swap_and_spawn(
source: &Path,
target: &Path,
version: &super::release::Version,
) -> Result<(), super::Error> {
let backup = backup_path(target);
move_file(target, &backup, 0)?;
if let Err(e) = move_file(source, target, MOVEFILE_REPLACE_EXISTING) {
// Best-effort revert
let _ = move_file(&backup, target, MOVEFILE_REPLACE_EXISTING);
return Err(e);
}
let pid = unsafe { GetCurrentProcessId() };
let args = vec![
OsString::from("--wait-pid"),
OsString::from(pid.to_string()),
OsString::from("--updated-to"),
OsString::from(format!("{}.{}.{}", version.major, version.minor, version.patch)),
];
super::handoff::spawn_detached(target, &args)
.map_err(super::Error::Io)
}
fn move_file(src: &Path, dst: &Path, flags: MOVE_FILE_FLAGS) -> Result<(), super::Error> {
let src_w = to_utf16_nul(src);
let dst_w = to_utf16_nul(dst);
let r = unsafe {
MoveFileExW(
PCWSTR::from_raw(src_w.as_ptr()),
PCWSTR::from_raw(dst_w.as_ptr()),
flags,
)
};
r.ok().map_err(|e| super::Error::SwapFailed(e.to_string()))
}
fn backup_path(target: &Path) -> PathBuf {
let pid = unsafe { GetCurrentProcessId() };
let mut p = target.to_owned();
let fname = target.file_name().map(|s| s.to_string_lossy().into_owned())
.unwrap_or_else(|| "exe".to_string());
p.set_file_name(format!("{fname}.old.{pid}"));
p
}
```
Use the existing `os::to_utf16_nul` helper for wide-char conversion (already used in `app.rs::run`).
4. **Rewrite `begin`** to call `swap_and_spawn(&staging, &current, &release.version)` instead of `spawn_handoff(&staging, &current)`. Pass the version from the `Release` struct already in hand.
5. **Add `Error::SwapFailed(String)` variant** to `src/update/mod.rs` if no existing variant fits the move-failure semantics. The `#[error(...)]` message should be `"file swap failed: {0}"`.
6. **Remove `spawn_handoff` function** entirely. Remove now-unused imports (`std::os::windows::process::CommandExt`, `Command`, `Stdio`, `CREATE_NO_WINDOW`, `DETACHED_PROCESS` constants).
7. **Compile check**: `cargo build --release`. Address any feature-flag gaps (likely need `Win32_Storage_FileSystem` added to Cargo `windows` features for `MoveFileExW`).
8. **Test rollback path manually**: write a temp .exe to staging that is read-only or has wrong permissions to force step 10 to fail; verify backup is restored and original is still runnable.
## Success Criteria
- [ ] `cargo build --release` clean.
- [ ] Manual auto-update from a test-tagged v0.1.99 produces ZERO visible console window across 5 consecutive runs.
- [ ] SHA-256 mismatch still rejects the swap (verify by tampering with a downloaded asset before swap).
- [ ] On forced step-10 failure (simulated): backup is restored, original binary still launches.
- [ ] `spawn_handoff` function no longer exists in the codebase (`grep -r spawn_handoff src/` returns empty).
- [ ] No `cmd.exe` string remains in `src/update/install.rs`.
## Risk Assessment
| Risk | Mitigation |
|---|---|
| `MoveFileExW` fails because AV holds a handle on the running exe | Surface `Error::SwapFailed`; user retries. Rare on Defender (which scans on read, not perpetually) |
| Two updates triggered in quick succession leave two `.old.<pid>` files | Phase 4 cleanup at startup handles this — glob removes ALL `.old.*` siblings |
| User's install dir is on a network share where renaming-while-open is forbidden | Existing `ensure_writable` probe catches read-only / no-write cases. Network FS oddities → surface as `NotWritable` |
| `MoveFileExW` with `REPLACE_EXISTING` on a non-NTFS volume | Works on FAT32 (per MS docs); the renaming-while-running concern is NTFS-specific but step 9 always renames an empty (just-emptied) path in step 10 |
| Release-build inlines + LTO breaks symbol-level rollback assumption | Functional rollback path is exercised by phase 5's manual test under release profile |
@@ -0,0 +1,165 @@
---
phase: 4
title: "Cleanup + tray notification"
status: complete
priority: P2
effort: "1.5h"
dependencies: [1, 3]
---
# Phase 4: Cleanup + tray notification
## Overview
Two additions that make the silent update visible to the user without being intrusive: (a) startup cleanup of stale `bubble.exe.old.<pid>` files from previous updates, and (b) a tray balloon "Updated to vX.Y.Z" on first launch after auto-update (driven by the `--updated-to` flag passed by phase 3).
<!-- Updated: Validation Session 1 - i18n approach corrected to match TOML struct-field architecture -->
## Requirements
**Functional**
- On every startup, scan `current_exe().parent()` for files matching `<exe-stem>.exe.old.*` and remove them silently. Errors logged at debug level, never surfaced to user.
- When `--updated-to vX.Y.Z` is passed AND the tray subsystem is initialized, show a balloon notification with localized title + body.
- Add **3 new fields** to `LocaleStrings` (`src/i18n/mod.rs:23-71`): `update_applied_title`, `update_applied_body`, `update_rollback_failed_body` (the last one is consumed by Phase 3 but the i18n change belongs to this phase's pattern). Body strings use Rust `format!` at call site — TOML strings hold raw text (e.g. body = `"Updated to v"`, then call site does `format!("{}{}", strings.update_applied_body, version)`). Choice of suffix vs prefix vs `{}` placeholder substitution is dialect-sensitive; use literal positional substitution via `format!` because TOML doesn't support template placeholders the i18n loader recognizes.
- Translate the 3 new strings in all **8 existing locale files**: `src/i18n/locales/{en,nl,es,fr,de,ja,ko,zh-TW}.toml`.
**Non-functional**
- Cleanup runs in the foreground startup path (it's a few file operations — no need for a thread).
- Balloon uses `NIIF_INFO` (blue info icon) not `NIIF_WARNING` (yellow triangle). The existing `tray::notify` (`src/tray/mod.rs:85`) hardcodes `NIIF_WARNING` and has **2 callers** (`src/app.rs:831` usage threshold, `src/app.rs:859` token expired) — both correctly semantically "warning". Rename existing `notify``notify_warning`; add new sibling `notify_info`.
## Architecture
### Cleanup
Triggered from `app::run` after tray icons register but before main message loop. Implementation lives in `update::handoff::cleanup_stale_old_exes` (stub was added in phase 1).
```rust
pub fn cleanup_stale_old_exes(current_exe: &Path) {
let Some(dir) = current_exe.parent() else { return };
let Some(stem) = current_exe.file_name() else { return };
let prefix = format!("{}.old.", stem.to_string_lossy());
let entries = match std::fs::read_dir(dir) {
Ok(e) => e,
Err(_) => return,
};
for entry in entries.flatten() {
let name = entry.file_name();
if name.to_string_lossy().starts_with(&prefix) {
if let Err(e) = std::fs::remove_file(entry.path()) {
log::debug!("cleanup_stale_old_exes: remove {:?} failed: {e}", entry.path());
}
}
}
}
```
### Balloon notification
Parse `--updated-to` argument early (alongside `--wait-pid` in main.rs), stash it on `AppState`. After the tray icons are registered for the first time, if the version string is present, call `tray::notify_info(hwnd, kind, title, body)`.
```rust
// main.rs (after --wait-pid parse):
let updated_to = args.iter()
.position(|a| a == "--updated-to")
.and_then(|i| args.get(i + 1))
.cloned();
// pass into app::run via existing arg-threading or a static OnceLock
```
```rust
// tray/mod.rs: split notify into two variants
pub fn notify_info(owner: HWND, kind: IconKind, title: &str, body: &str) {
notify_inner(owner, kind, title, body, NIIF_INFO);
}
pub fn notify_warning(owner: HWND, kind: IconKind, title: &str, body: &str) {
notify_inner(owner, kind, title, body, NIIF_WARNING);
}
fn notify_inner(owner: HWND, kind: IconKind, title: &str, body: &str, flags: NOTIFY_ICON_INFOTIP_FLAGS) {
// existing body, but use `flags` instead of hardcoded NIIF_WARNING
}
```
If `tray::notify` has only one caller today (which the brainstorm scout suggested), just rename it to `notify_warning` and add `notify_info`.
## Related Code Files
- **Modify**: `src/update/handoff.rs::cleanup_stale_old_exes` (fill in phase-1 stub)
- **Modify**: `src/app.rs::run` (call cleanup; call tray::notify_info after tray registration if updated_to is set)
- **Modify**: `src/app.rs:831,859` (rename `tray::notify``tray::notify_warning` at both existing call sites)
- **Modify**: `src/main.rs` (parse `--updated-to`, stash for app)
- **Modify**: `src/tray/mod.rs` — rename `notify``notify_warning`; add `notify_info`; extract shared `notify_inner(... flags: NOTIFY_ICON_INFOTIP_FLAGS)`
- **Modify**: `src/i18n/mod.rs::LocaleStrings` — add 3 new `String` fields: `update_applied_title`, `update_applied_body`, `update_rollback_failed_body`
- **Modify**: `src/i18n/locales/{en,nl,es,fr,de,ja,ko,zh-TW}.toml` — 8 files, add the 3 new keys to each. Use machine translation for non-English where idiomatic translation unavailable; flag with comment for native-speaker review later
## Implementation Steps
1. **i18n: add 3 new fields to `LocaleStrings`** in `src/i18n/mod.rs:23-71`. After the existing `threshold_95_body` field, add:
```rust
pub update_applied_title: String,
pub update_applied_body: String,
pub update_rollback_failed_body: String,
```
2. **Translate in 8 locale files** (`src/i18n/locales/{en,nl,es,fr,de,ja,ko,zh-TW}.toml`). Suggested English values (mirror style of existing `token_expired_title`/`token_expired_body`):
```toml
update_applied_title = "Update applied"
update_applied_body = "Updated to v" # call site appends version string
update_rollback_failed_body = "Update failed. Your original binary is saved as " # call site appends backup path + suffix
```
For non-English, machine-translate body+title, leave the trailing space/prefix structure intact. Comment in PR description: "Translations machine-generated; flagged for native-speaker review".
3. **Split `tray::notify`** (`src/tray/mod.rs:85-94`) into:
- rename existing fn → `notify_warning` (preserves current `NIIF_WARNING` semantics)
- extract shared `fn notify_inner(owner, kind, title, body, flags: NOTIFY_ICON_INFOTIP_FLAGS)`
- add new `pub fn notify_info(owner, kind, title, body)` that calls `notify_inner(..., NIIF_INFO)`
Update both existing call sites (`src/app.rs:831,859`) to call `notify_warning` instead of `notify`.
4. **Fill in `cleanup_stale_old_exes`** per architecture section.
5. **Parse `--updated-to` in main.rs**, thread it into `app::run`. Two options:
- Pass as a new arg to `pub fn run(updated_to: Option<String>)`.
- Store in a `OnceLock<Option<String>>` inside `update::handoff`, set in main, read in app.
Pick the simpler one — direct function arg is preferred unless `run`'s signature is already heavily used elsewhere.
6. **In `app::run`** after tray icons register and the main window message loop is about to enter:
```rust
if let Some(v) = updated_to.as_ref() {
let strings = i18n.strings();
let title = strings.update_applied_title.clone();
let body = format!("{}{}", strings.update_applied_body, v);
tray::notify_info(msg_hwnd, IconKind::ClaudeCode, &title, &body);
}
update::handoff::cleanup_stale_old_exes(&exe);
```
Use `ClaudeCode` IconKind because it's always present when Claude is enabled (default). If user disabled Claude and enabled only Codex, fall back to Codex kind. Cheapest: try ClaudeCode first; if `tray::notify_info` fails silently, no harm.
7. **Compile check**: `cargo build --release`.
8. **Manual test cleanup**:
- Create a file `claude-code-usage-bubble.exe.old.1234` next to the running binary.
- Launch the app.
- Verify the file is gone after launch.
9. **Manual test notification**:
- Launch with `--updated-to 9.9.9` flag.
- Verify Windows notification appears with the title + version body.
- Verify it uses the blue info icon, not yellow warning.
## Success Criteria
- [ ] `cargo build --release` clean.
- [ ] Stale `.old.<pid>` files in install dir are removed on every startup (idempotent).
- [ ] Launching with `--updated-to vX.Y.Z` shows a tray balloon with localized title and version body.
- [ ] Balloon icon is blue info (NIIF_INFO), not yellow warning.
- [ ] Launching WITHOUT `--updated-to` shows no balloon (existing behavior preserved).
- [ ] All 8 locale TOML files (`en, nl, es, fr, de, ja, ko, zh-TW`) contain the 3 new keys (`update_applied_title`, `update_applied_body`, `update_rollback_failed_body`); `cargo build --release` would fail to deserialize otherwise.
## Risk Assessment
| Risk | Mitigation |
|---|---|
| Cleanup removes a `.old.<pid>` file that another running instance still depends on | Impossible by construction: only the spawning instance writes `.old.<pid>` and it has already exited by the time the new instance runs cleanup |
| Glob false-positive (e.g. a user-created file matching the pattern) | Pattern requires `.old.` literal AND a numeric pid-like suffix is implied; we don't pattern-match the suffix strictly. Risk is theoretical; user-created files matching `claude-code-usage-bubble.exe.old.*` is extremely unlikely |
| Tray balloon doesn't show because NIM_MODIFY runs before NIM_ADD completes | Defer the `notify_info` call by one tick (PostMessage to message loop) if testing shows races. Likely unnecessary because tray icons register synchronously |
| i18n loader is struct-field based, no template substitution at the loader level | Confirmed by Validation Session 1: append/prepend the version via Rust `format!` at the call site. TOML strings are static text fragments only |
| Translations diverge from idiomatic expression in non-English locales | Machine-translate for first cut, mark "FIXME: review by native speaker" in commit message. Subsequent crowd-sourced fixes are out of this plan's scope |
@@ -0,0 +1,104 @@
---
phase: 5
title: "Manual end-to-end verification"
status: pending
priority: P2
effort: "1h"
dependencies: [1, 2, 3, 4]
---
# Phase 5: Manual end-to-end verification
## Overview
Functional sign-off across all changed code paths. No unit tests added (the changed surface is Win32-heavy and effectively integration territory); instead, a documented manual checklist that the maintainer runs once before committing.
## Requirements
**Functional**
- All success criteria from phases 1-4 verified on a real Windows machine (Win11 preferred, Win10 as secondary if available).
- Build artifact runs without errors when launched plainly (no flags).
- No regression in existing update / restart / refresh paths.
**Non-functional**
- Verification log saved as a section in `phase-05` after run, with date + outcome per item.
## Test Matrix
### Group A: Restart path (phase 2)
| # | Step | Expected |
|---|---|---|
| A1 | Launch binary, open right-click menu, click "Restart" | No console flash; new instance appears within 1.5s |
| A2 | Repeat A1 twenty times back-to-back | Zero flashes observed; settings.json mtime updates each time |
| A3 | Launch with `--diagnose`, click Restart, inspect `%TEMP%\claude-code-usage-bubble.log` | Shows "restart: spawned detached child, posting quit" and new instance shows mutex acquired (within 200ms of the wait completing) |
### Group B: Update install path (phase 3)
Setup: tag a test release `v0.1.99` via the existing GitHub Actions workflow (per `plans/260516-1730-github-release-auto-update`). Bump local `Cargo.toml` back to `v0.1.0` before running. Build the local v0.1.0 with this plan's changes.
| # | Step | Expected |
|---|---|---|
| B1 | Launch v0.1.0 build, set update channel to "Hourly", manually trigger "Check for updates" | "Update available" shown; click "Apply" |
| B2 | During B1 apply | No console flash; new v0.1.99 instance appears |
| B3 | After B1/B2 | Tray balloon "Update applied — Updated to v0.1.99" appears (blue info icon) |
| B4 | Inspect install dir after B1/B2 | NO `.old.*` files remain (cleanup removed them) |
| B5 | Repeat B1 four more times (after re-tagging v0.1.100 etc.) | Zero flashes across 5 update cycles |
| B6 | Tamper test: edit downloaded asset on disk before swap (would require pausing between download and swap — gate via `--diagnose` log timestamps) | SHA-256 mismatch raises `Error::ChecksumMismatch`; swap is NOT performed; original binary still runs |
| B7 | Rollback test: make staging path read-only or simulate step-10 failure | Backup restored; original binary still runs after a manual restart |
### Group C: Cleanup + notification (phase 4)
| # | Step | Expected |
|---|---|---|
| C1 | Manually drop `claude-code-usage-bubble.exe.old.9999` next to the running binary; launch app | Stale file is removed within 1s of launch |
| C2 | Launch app with `--updated-to 9.9.9` arg from a terminal | Tray balloon shows title + "Updated to v9.9.9" in current UI language |
| C3 | Repeat C2 with each supported locale | Each locale shows correct translation |
| C4 | Launch normally (no `--updated-to`) | No balloon appears |
### Group D: Regression smoke
| # | Step | Expected |
|---|---|---|
| D1 | Launch binary fresh (cold start, no flags) | Bubble appears in <2s; usage refresh fires once; tray icon registers |
| D2 | Open settings (right-click → Language → switch), confirm restart happens | Restart works without flash (this is the same `restart_app` path) |
| D3 | Disable auto-update ("Disabled"), wait 30s, re-enable Hourly | No state corruption; check timer resets |
| D4 | Run with `--diagnose --apply-update <some-path> <pid>` (legacy compat) | Returns exit code 0 cleanly (per `update::install::run_cli`) — unchanged |
## Implementation Steps
1. Build `cargo build --release`.
2. Copy `target/release/claude-code-usage-bubble.exe` to `%LOCALAPPDATA%\ClaudeCodeUsageBubble\` (fresh test directory).
3. Run through Test Matrix A → B → C → D in order.
4. For each test row, record outcome (PASS/FAIL + notes) in the Verification Log section below.
5. If any FAIL: open an issue describing the failure, do NOT mark phase complete.
6. If all PASS: mark phase complete, commit changes following project commit conventions (no `chore:` or `docs:` per CLAUDE.md).
## Success Criteria
- [ ] All Group A tests PASS.
- [ ] All Group B tests PASS (B6 + B7 may be skipped if test scaffolding too costly; document as such).
- [ ] All Group C tests PASS.
- [ ] All Group D tests PASS.
- [ ] Verification Log filled in with date + outcomes.
- [ ] No console flash observed across the entire test session.
## Risk Assessment
| Risk | Mitigation |
|---|---|
| Manual testing skips Group B because tagging real releases is annoying | Alternative: build two local copies (v0.1.0 + v0.1.99), set up a local HTTP server with a fake GitHub-Releases-shaped JSON, point the app at it via a debug flag. Out of scope for this plan but noted |
| "Flash" is subjective at 60Hz | Record screen with OBS at 60fps for one test run, scrub frame-by-frame to confirm zero console window appearance |
| Win10-only flash regression (only test on Win11) | Document Win10 testing as "best effort"; primary target is Win11. Note Win10 result in Verification Log |
## Verification Log
<!-- Fill in after running the test matrix -->
### Session 1 — TBD
- Group A: TBD
- Group B: TBD
- Group C: TBD
- Group D: TBD
- Flash observed: TBD
- Notes: TBD
@@ -0,0 +1,96 @@
---
title: "Silent in-app update + restart (no cmd.exe)"
description: "Replace cmd.exe handoff in update install + app restart paths with native Win32 (MoveFileExW + CreateProcessW) so no terminal window can ever flash. Add tray-balloon notification after auto-updates."
status: in_progress
priority: P2
branch: "main"
tags: ["update", "restart", "win32", "ux"]
blockedBy: []
blocks: []
created: "2026-05-21T07:43:16.332Z"
createdBy: "ck:plan"
source: skill
---
# Silent in-app update + restart (no cmd.exe)
## Overview
Two paths today spawn `cmd.exe /c "timeout ... & start ..."` for update install (`src/update/install.rs::begin`) and app restart (`src/app.rs::restart_app`). Combination of `CREATE_NO_WINDOW | DETACHED_PROCESS` + inner `start ""` can still flash a console window on some Windows configs. This plan replaces both with native `MoveFileExW` + `CreateProcessW` (the main exe is `windows_subsystem = "windows"`, so direct spawn never allocates a console). Also wires a tray balloon "Updated to vX.Y.Z" on first launch after an auto-update.
Brainstorm context: [`plans/reports/brainstormer-260521-1530-silent-update-no-cmd.md`](../reports/brainstormer-260521-1530-silent-update-no-cmd.md).
Supersedes the cmd.exe mechanism decision in [`260518-0945-menu-restart-button`](../260518-0945-menu-restart-button/plan.md) (that plan picked cmd.exe deliberately, modeled on `update::install`; this plan replaces both call sites with the native path).
## Phases
| Phase | Name | Status |
|-------|------|--------|
| 1 | [Foundation: CLI flags + native spawn helper + mutex retry](./phase-01-foundation-cli-flags-native-spawn-helper-mutex-retry.md) | Complete |
| 2 | [Restart path: replace restart_app with native CreateProcessW](./phase-02-restart-path-replace-restart-app-with-native-createprocessw.md) | Complete |
| 3 | [Update install: rename + move + native spawn](./phase-03-update-install-rename-move-native-spawn.md) | Complete |
| 4 | [Cleanup + tray notification](./phase-04-cleanup-tray-notification.md) | Complete |
| 5 | [Manual end-to-end verification](./phase-05-manual-end-to-end-verification.md) | Pending (user-driven) |
## Key contracts (must hold across plan)
| Contract | Source today | Invariant |
|---|---|---|
| Singleton mutex name | `app.rs` `APP_MUTEX_NAME` = `Global\ClaudeCodeUsageBubble` | New instance must wait for parent to release before acquiring |
| Main binary subsystem | `main.rs:1` `#![windows_subsystem = "windows"]` | Direct spawn allocates no console |
| Asset filename | `release.rs:7` `claude-code-usage-bubble.exe` | Unchanged |
| SHA-256 verification | `install.rs:64-73` | Unchanged — still verified before swap |
| Settings save on shutdown | `app.rs::restart_app` snap+save | Preserved in new restart helper |
## Dependencies
No cross-plan dependencies. Related (superseded mechanism): `260518-0945-menu-restart-button`.
## Out of Scope
- `src/usage/refresh.rs` CLI spawns (`claude.cmd`, `codex.cmd`, `powershell.exe`, `wsl.exe`) — already use `CREATE_NO_WINDOW`; tracked for follow-up.
- `src/creds/wsl_bridge.rs` `wsl.exe` calls — same.
- Code signing / SmartScreen suppression — separate roadmap item.
- Cross-platform restart/update — Windows-only by design.
## Unresolved Questions
None.
## Validation Log
### Session 1 — 2026-05-21
#### Verification Results
- **Tier**: Full (5 phases)
- **Claims checked**: 8
- **Verified**: 6 | **Failed**: 2 | **Unverified**: 0
- Verified: `app.rs:155` mutex creation; `app.rs:1378-1432` restart_app bounds; `install.rs:42-48` `--apply-update` exit-clean handler; `install.rs:99-121` `spawn_handoff`; `os::to_utf16_nul` exists and is in use; `windows = 0.58` features in Cargo.toml include `Win32_System_Threading` but NOT `Win32_Storage_FileSystem` (phase 3 must add it).
- Failed: (V1) phase 4 said `tray::notify` has "one current caller" — actually 2 (`app.rs:831`, `app.rs:859`); (V2) phase 4 said i18n uses key-based template lookup with `{v}` placeholder — actually uses struct-field-based `LocaleStrings` with TOML, no template engine.
#### Decisions
1. **i18n approach: add 3 fields to `LocaleStrings` + translate 8 locale TOML files; use Rust `format!` at call site for version substitution.**
Reason: existing pattern is struct-field-based via `include_str!` of `src/i18n/locales/{en,nl,es,fr,de,ja,ko,zh-TW}.toml`. No template substitution at the loader level.
→ Propagated to `phase-04-cleanup-tray-notification.md`: Requirements, Related Code Files, Implementation Steps, Success Criteria, Risk Assessment all updated.
2. **Rollback failure escalation: Windows `MessageBoxW` (MB_OK | MB_ICONERROR) when MoveFileExW step 10 AND best-effort revert both fail.**
Reason: silent failure here would orphan the user — they'd have no exe at install path. A clear modal tells them where the backup is (`bubble.exe.old.<pid>`).
→ Propagated to `phase-03-update-install-rename-move-native-spawn.md`: rollback table extended; imports list updated; new helper `surface_rollback_failure` added; new `LocaleStrings` field `update_rollback_failed_body` added to phase 4's i18n work.
3. **Stuck-parent fallback: accept 8s total budget (5s `WaitForSingleObject` + 3s mutex retry), then exit cleanly. No `TerminateProcess`.**
Reason: forcing process termination would defeat the `settings::save` defensive flush guarantee. The 8s ceiling is generous for normal Windows scheduling; if a real hang persists, user can manually kill old process via Task Manager — acceptable failure mode.
→ No phase file change needed. Phase 1 budget numbers (5s + 3s) already match.
4. **Auto-update timing: apply when check fires, no idle-window deferral.**
Reason: matches user's brainstorm-phase choice ("Fully silent auto-update"). Idle detection adds complexity (state tracking, defer-budget, defer-never-applies edge case) for marginal UX gain — bubble flicker during ~1s restart is acceptable.
→ No phase file change needed.
#### Whole-Plan Consistency Sweep
- Files reread: `plan.md`, `phase-01-…md`, `phase-02-…md`, `phase-03-…md`, `phase-04-…md`, `phase-05-…md`
- Decision deltas checked: 4
- Reconciled stale references: 2
- Phase 4 i18n section rewritten from key-based template to struct-field + Rust `format!`
- Phase 4 `tray::notify` caller count corrected from "1" to "2" with explicit file:line citations
- Cross-phase touchpoints verified: Phase 3 adds `update_rollback_failed_body` to `LocaleStrings`, Phase 4 owns the full i18n change (3 fields + 8 TOMLs) — both phases reference the same struct, consistent
- Unresolved contradictions: 0
@@ -0,0 +1,124 @@
# Silent in-app update + restart (no cmd.exe)
**Date:** 2026-05-21
**Author:** brainstormer
**Status:** approved (ready for `/ck:plan`)
## Problem
User sees an occasional flash terminal window. Two paths today spawn `cmd.exe /c "timeout ... & start ..."` for update install and app restart, with `CREATE_NO_WINDOW | DETACHED_PROCESS`. Combination of those flags + the inner `start ""` invocation can still emit a brief console flash on some Windows configs (Defender hooks, conhost init, AV inspection). Goal: zero-flash, fully silent auto-update + restart, with in-app notification.
## Scope
**In scope**
- `src/update/install.rs::begin` — kill cmd.exe handoff
- `src/app.rs::restart_app` — kill cmd.exe handoff
- New CLI flag `--wait-pid <pid>` on the main binary (cooperates with itself across update)
- Cleanup of stale `bubble.exe.old.*` siblings at startup
- Tray balloon "Updated to vX.Y.Z" on first run after auto-update
**Out of scope (split as follow-up)**
- `src/usage/refresh.rs` CLI spawns (`claude.cmd`, `codex.cmd`, `powershell.exe`, `wsl.exe`) — already use `CREATE_NO_WINDOW`; address separately if flash persists after this change
- `src/creds/wsl_bridge.rs` `wsl.exe` calls — same reasoning
## Approaches evaluated
| Approach | Decision | Rationale |
|---|---|---|
| A: Native rename + direct `CreateProcessW` + `--wait-pid` | **CHOSEN** | Zero cmd.exe ⇒ zero flash possible; single-binary preserved; matches existing Win32 style; recoverable on interrupt |
| B: Helper-exe pattern (`bubble-updater.exe`) | rejected | Reverses deliberate "no helper exe" decision (`src/update/install.rs:3-5`); release-pipeline change; helper bootstrap problem |
| C: NTFS POSIX atomic replace (`FileRenameInfoEx`) | rejected | Obscure API; harder failure modes with AV / image-protection; not worth the elegance trade-off |
## Final design — Approach A
### Update install flow
```
1. fetch_latest() (unchanged — pure HTTP via WinHTTP)
2. download(release_url, staging_path) (unchanged — sha256 verify)
3. rename current.exe -> current.exe.old.<pid> (MoveFileExW, allowed while running)
4. move staging.exe -> current.exe (MoveFileExW REPLACE_EXISTING)
5. settings::save(snap) (defensive flush, unchanged)
6. release singleton mutex (explicit ReleaseMutex + CloseHandle)
7. CreateProcessW(current.exe, "--wait-pid <our_pid> --updated-to vX.Y.Z",
CREATE_NO_WINDOW | DETACHED_PROCESS)
8. PostQuitMessage(0)
```
### Restart flow (settings-change path)
```
1. settings::save(snap)
2. CreateProcessW(current.exe, "--wait-pid <our_pid>",
CREATE_NO_WINDOW | DETACHED_PROCESS)
3. release singleton mutex
4. PostQuitMessage(0)
```
### New instance startup additions
```rust
// Before acquiring Global\ClaudeCodeUsageBubble mutex:
if let Some(parent_pid) = parse_wait_pid_arg() {
let h = OpenProcess(SYNCHRONIZE, FALSE, parent_pid)?;
WaitForSingleObject(h, 5000); // 5s cap; proceed regardless
CloseHandle(h);
}
// After main window is up:
cleanup_old_exes(current_dir, "bubble.exe.old.*");
if let Some(version) = parse_updated_to_arg() {
tray::show_balloon(t!("update.toast.updated_to", v = version));
}
```
### Why `--wait-pid` instead of cmd's `timeout /t 2`
- `timeout` is a cmd.exe builtin; using it requires cmd.exe.
- `WaitForSingleObject` on the parent process handle is the canonical Win32 idiom: zero delay if parent already exited, exact timing when it actually exits, no console involvement.
- Bonus: removes the magic "2 seconds is enough" guess.
### Path safety
The current `reject_unsafe_path` (`%` rejection) becomes unnecessary — no cmd.exe to expand `%var%`. Keep the function for defense-in-depth; revisit in code review.
## Files touched
| File | Change |
|---|---|
| `src/update/install.rs` | Replace `spawn_handoff` with `swap_and_spawn` using `MoveFileExW` + `CreateProcessW`; drop `cmd` arg-quoting code |
| `src/update/mod.rs` | Add `Error::SwapFailed` variant if needed |
| `src/app.rs::restart_app` | Replace cmd.exe spawn with `CreateProcessW` + mutex release ordering |
| `src/app.rs` | Add CLI flag parsing for `--wait-pid` and `--updated-to`; cleanup pass for `.old.*` siblings; balloon tray call on successful update boot |
| `src/main.rs` | Wire `--wait-pid` into the early-startup mutex-acquisition path (BEFORE `update::run_cli` check) |
| `src/tray/mod.rs` (or `tray/badge.rs`) | Confirm `Shell_NotifyIconW` with `NIF_INFO` balloon is supported by current tray code; add helper if missing |
| `src/i18n/*` | New string keys: `update.toast.updated_to` |
## Risks + mitigations
| Risk | Mitigation |
|---|---|
| `MoveFileExW` rename fails (NTFS permission denied, AV scanner holding handle) | Surface `Error::NotWritable` to user; do NOT proceed to step 4 (still recoverable — original exe untouched at this point) |
| New instance crashes before clearing old exe ⇒ `.old.*` accumulates | Cleanup glob `bubble.exe.old.*` on every startup is idempotent and cheap |
| Mutex race: new instance acquires before old releases | `--wait-pid` + `WaitForSingleObject(5000ms)` covers it; if it times out the new instance retries `CreateMutexW` in a 200ms loop for ~3s before giving up |
| User on FAT32 / non-NTFS volume | `MoveFileExW` with `MOVEFILE_REPLACE_EXISTING` still works on FAT32; renaming-while-running is the NTFS-specific concern but the .exe is rarely on FAT32. Document the edge case |
| `--wait-pid` arg parsed in legacy build that doesn't recognize it | Old builds will ignore unknown args (cargo CLI parser behavior — verify). If they crash, the user can manually launch. Acceptable: this is a one-way migration; once the new flag is in a release, future updates are smooth |
## Success criteria
1. Manual update from v0.1.9 → test-tagged v0.1.99 produces ZERO visible console window across 20 consecutive runs on Win11 + Win10
2. Restart triggered from menu (e.g. language change) produces ZERO visible console
3. Auto-update at scheduled interval (Hourly) produces a tray balloon "Updated to v0.1.99" on next launch
4. `.old.*` files do not accumulate after 5 update cycles
5. App still launches cleanly when no parent PID was passed (i.e. fresh user start)
6. SHA-256 verification path unchanged and still rejects tampered binaries
## Out-of-scope follow-ups
1. **Audit `usage::refresh::spawn_local` / `spawn_wsl`**: the `wsl.exe` invocation in particular has known console-flash quirks even with `CREATE_NO_WINDOW`. If the user still sees occasional flashes after this change ships, that is the next investigation target.
2. **`creds::wsl_bridge::wsl_run`**: same family of `wsl.exe` invocations.
## Unresolved questions
- None blocking implementation. (Open follow-up: whether to also pipe `--wait-pid` into the legacy `--apply-update` compatibility branch in `update::run_cli`, in case a very old build is doing the spawning.)
+97 -59
View File
@@ -6,8 +6,7 @@
// message-only window owned by this module.
use std::collections::HashMap;
use std::os::windows::process::CommandExt;
use std::process::{Command, Stdio};
use std::ffi::OsString;
use std::sync::atomic::{AtomicBool, Ordering};
use std::sync::{Arc, Mutex, MutexGuard, OnceLock};
use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};
@@ -15,7 +14,7 @@ use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};
use windows::core::PCWSTR;
use windows::Win32::Foundation::*;
use windows::Win32::System::LibraryLoader::GetModuleHandleW;
use windows::Win32::System::Threading::CreateMutexW;
use windows::Win32::System::Threading::{CreateMutexW, GetCurrentProcessId, Sleep};
use windows::Win32::UI::HiDpi::{
SetProcessDpiAwarenessContext, DPI_AWARENESS_CONTEXT_PER_MONITOR_AWARE_V2,
};
@@ -36,10 +35,11 @@ use crate::usage::{self, ProviderId, Registry, UsageWindows};
// Win32 message IDs owned by this module.
pub const WM_APP_USAGE_UPDATED: u32 = 0x8001;
// Posted from the update worker thread when the swap-and-restart cmd
// handoff has been launched successfully. The UI thread responds by
// calling PostQuitMessage(0) to release the file lock on the running
// .exe so cmd.exe can overwrite it.
// Posted from the update worker thread after `install::begin` has
// already swapped the binary on disk and spawned the new detached
// child. The UI thread responds with PostQuitMessage(0) so the old
// instance exits cleanly and releases the singleton mutex for the
// child waiting on `--wait-pid`.
pub const WM_APP_UPDATE_APPLIED: u32 = 0x8002;
// Timer IDs used with `SetTimer(msg_hwnd, …)`.
@@ -147,27 +147,52 @@ fn lock_state() -> MutexGuard<'static, Option<AppState>> {
// ---------- Entry ----------
pub fn run() {
/// Acquire the singleton mutex, optionally retrying for ~3s if the
/// caller passed `--wait-pid` (i.e. we just spawned from an exiting
/// parent that has not yet released its handle).
fn acquire_singleton_mutex(retry: bool) -> Option<HANDLE> {
let mutex_name_w = os::to_utf16_nul(APP_MUTEX_NAME);
let max_attempts = if retry { 15 } else { 1 };
for attempt in 0..max_attempts {
let handle = unsafe { CreateMutexW(None, false, PCWSTR::from_raw(mutex_name_w.as_ptr())) };
match handle {
Ok(h) => {
let already = unsafe { GetLastError() } == ERROR_ALREADY_EXISTS;
if !already {
return Some(h);
}
// Mutex still held by parent. Close this handle and retry.
unsafe {
let _ = CloseHandle(h);
}
if attempt + 1 == max_attempts {
log::info!("another instance already running; exiting");
return None;
}
log::debug!(
"mutex still held by parent (attempt {}/{}), waiting 200ms",
attempt + 1,
max_attempts
);
unsafe { Sleep(200) };
}
Err(e) => {
log::error!("CreateMutex failed: {e}");
return None;
}
}
}
None
}
pub fn run(args: crate::AppArgs) {
unsafe {
let _ = SetProcessDpiAwarenessContext(DPI_AWARENESS_CONTEXT_PER_MONITOR_AWARE_V2);
}
let mutex_name_w = os::to_utf16_nul(APP_MUTEX_NAME);
let _mutex = unsafe {
let handle = CreateMutexW(None, false, PCWSTR::from_raw(mutex_name_w.as_ptr()));
match handle {
Ok(h) => {
if GetLastError() == ERROR_ALREADY_EXISTS {
log::info!("another instance already running; exiting");
return;
}
h
}
Err(e) => {
log::error!("CreateMutex failed: {e}");
return;
}
}
let _mutex = match acquire_singleton_mutex(args.wait_pid_present) {
Some(h) => h,
None => return,
};
let settings = settings::load();
@@ -208,6 +233,16 @@ pub fn run() {
create_initial_bubbles();
refresh_tray_icons();
// Post-update tasks: show "Updated to vX.Y.Z" balloon (driven by
// --updated-to passed by the previous instance) and sweep any
// stale `<exe>.old.<pid>` siblings left by past in-place swaps.
if let Some(v) = args.updated_to.as_ref() {
announce_update_applied(msg_hwnd, v);
}
if let Ok(exe_path) = std::env::current_exe() {
update::handoff::cleanup_stale_old_exes(&exe_path);
}
let poll_interval = lock_state()
.as_ref()
.map(|s| s.settings.poll_interval_ms)
@@ -828,7 +863,7 @@ fn show_threshold_balloon(provider: ProviderId, threshold: u8) {
};
(s.msg_hwnd, provider, title, body)
};
tray::notify(payload.0.to_hwnd(), payload.1, &payload.2, &payload.3);
tray::notify_warning(payload.0.to_hwnd(), payload.1, &payload.2, &payload.3);
}
fn show_token_expired_balloon(failed: ProviderId) {
@@ -856,7 +891,30 @@ fn show_token_expired_balloon(failed: ProviderId) {
};
(s.msg_hwnd, failed, title, body)
};
tray::notify(payload.0.to_hwnd(), payload.1, &payload.2, &payload.3);
tray::notify_warning(payload.0.to_hwnd(), payload.1, &payload.2, &payload.3);
}
/// Show the "Updated to vX.Y.Z" balloon on first launch after an
/// auto-update. Picks Claude as the host icon if it's registered;
/// otherwise falls back to Codex. If neither is registered the
/// notification silently drops — better than crashing.
fn announce_update_applied(_msg_hwnd: HWND, version: &str) {
let payload = {
let s = lock_state();
let Some(s) = s.as_ref() else {
return;
};
let strings = s.i18n.strings();
let title = strings.update_applied_title.clone();
let body = format!("{}{}", strings.update_applied_body, version);
let host = if s.settings.show_claude_code {
ProviderId::Claude
} else {
ProviderId::ChatGpt
};
(s.msg_hwnd, host, title, body)
};
tray::notify_info(payload.0.to_hwnd(), payload.1, &payload.2, &payload.3);
}
// ---------- Context menu ----------
@@ -1377,17 +1435,9 @@ fn set_update_check_interval(value: Option<u64>) {
// ---------- Restart ----------
// Windows CreateProcess flags. Match the values used by `update::install`
// so the cmd-handoff child detaches cleanly without flashing a console.
const RESTART_CREATE_NO_WINDOW: u32 = 0x0800_0000;
const RESTART_DETACHED_PROCESS: u32 = 0x0000_0008;
/// Relaunch the running binary via a detached cmd.exe handoff.
///
/// The 1-second `timeout` gives the current process time to release the
/// `Global\ClaudeCodeUsageBubble` mutex before the relaunched instance's
/// `CreateMutexW` runs, otherwise the new instance would see
/// `ERROR_ALREADY_EXISTS` and exit immediately.
/// Relaunch the running binary by spawning a detached child via
/// `CreateProcessW`. The child waits on our PID before acquiring the
/// singleton mutex, so no shell handoff or timer is required.
fn restart_app() {
// Defensive flush — bubble positions and most settings already persist
// on change, but a final save is cheap insurance. Snapshot then drop the
@@ -1404,30 +1454,18 @@ fn restart_app() {
return;
}
};
let exe_str = exe.to_string_lossy();
// cmd.exe expands `%var%` inside double quotes, so a path containing `%`
// would let the environment leak into the relaunch. Refuse — matches the
// defense already used in `update::install`.
if exe_str.contains('%') {
log::error!("restart: refusing path containing '%': {exe_str}");
return;
}
let exe_str = exe_str.replace('"', "");
let cmd = format!(r#"timeout /t 1 /nobreak >nul & start "" "{exe_str}""#);
let spawned = Command::new("cmd.exe")
.raw_arg("/c")
.raw_arg(format!("\"{cmd}\""))
.creation_flags(RESTART_CREATE_NO_WINDOW | RESTART_DETACHED_PROCESS)
.stdin(Stdio::null())
.stdout(Stdio::null())
.stderr(Stdio::null())
.spawn();
match spawned {
Ok(_) => {
log::info!("restart: cmd handoff spawned, posting quit");
let pid = unsafe { GetCurrentProcessId() };
let args = vec![
OsString::from("--wait-pid"),
OsString::from(pid.to_string()),
];
match update::handoff::spawn_detached(&exe, &args) {
Ok(()) => {
log::info!("restart: spawned detached child (parent pid={pid}), posting quit");
unsafe { PostQuitMessage(0) };
}
Err(e) => log::error!("restart: cmd spawn failed: {e}"),
Err(e) => log::error!("restart: spawn_detached failed: {e}"),
}
}
+3
View File
@@ -44,3 +44,6 @@ chatgpt_token_expired_title = "Codex-Sitzung abgelaufen"
chatgpt_token_expired_body = "Melde dich erneut an, um die Nutzung weiter zu verfolgen."
threshold_80_body = "5-Stunden-Limit naht."
threshold_95_body = "Limit fast erreicht — gönn dir eine Pause."
update_applied_title = "Update angewendet"
update_applied_body = "Aktualisiert auf v"
update_rollback_failed_body = "Update fehlgeschlagen. Deine ursprüngliche Binärdatei liegt unter: "
+3
View File
@@ -44,3 +44,6 @@ chatgpt_token_expired_title = "Codex session expired"
chatgpt_token_expired_body = "Sign in again to keep tracking your usage."
threshold_80_body = "Approaching the 5-hour limit."
threshold_95_body = "Limit is close — consider easing up."
update_applied_title = "Update applied"
update_applied_body = "Updated to v"
update_rollback_failed_body = "Update failed. Your original binary is saved at: "
+3
View File
@@ -44,3 +44,6 @@ chatgpt_token_expired_title = "Sesión de Codex caducada"
chatgpt_token_expired_body = "Vuelve a iniciar sesión para seguir registrando el uso."
threshold_80_body = "Cerca del límite de 5 horas."
threshold_95_body = "Límite casi alcanzado — reduce el ritmo."
update_applied_title = "Actualización aplicada"
update_applied_body = "Actualizado a v"
update_rollback_failed_body = "Actualización fallida. Tu binario original está guardado en: "
+3
View File
@@ -44,3 +44,6 @@ chatgpt_token_expired_title = "Session Codex expirée"
chatgpt_token_expired_body = "Reconnectez-vous pour continuer à suivre votre utilisation."
threshold_80_body = "Approche de la limite de 5 heures."
threshold_95_body = "Limite proche — pensez à lever le pied."
update_applied_title = "Mise à jour appliquée"
update_applied_body = "Mis à jour vers v"
update_rollback_failed_body = "Échec de la mise à jour. Votre binaire d'origine est enregistré à : "
+3
View File
@@ -44,3 +44,6 @@ chatgpt_token_expired_title = "Codexのセッションが切れました"
chatgpt_token_expired_body = "使用状況の追跡を続けるには再度サインインしてください。"
threshold_80_body = "5時間の上限に近づいています。"
threshold_95_body = "上限に近づきました — ペースを落としましょう。"
update_applied_title = "更新を適用しました"
update_applied_body = "バージョン v"
update_rollback_failed_body = "更新に失敗しました。元のバイナリは次の場所に保存されています: "
+3
View File
@@ -44,3 +44,6 @@ chatgpt_token_expired_title = "Codex 세션 만료"
chatgpt_token_expired_body = "사용량을 계속 추적하려면 다시 로그인하세요."
threshold_80_body = "5시간 한도에 가까워지고 있어요."
threshold_95_body = "한도 임박 — 잠시 쉬어가세요."
update_applied_title = "업데이트가 적용되었습니다"
update_applied_body = "버전 v"
update_rollback_failed_body = "업데이트 실패. 원본 바이너리는 다음 위치에 저장되었습니다: "
+3
View File
@@ -44,3 +44,6 @@ chatgpt_token_expired_title = "Codex-sessie verlopen"
chatgpt_token_expired_body = "Meld je opnieuw aan om gebruik te blijven volgen."
threshold_80_body = "Je nadert de 5-uurslimiet."
threshold_95_body = "Limiet bijna bereikt — overweeg even gas terug te nemen."
update_applied_title = "Update toegepast"
update_applied_body = "Bijgewerkt naar v"
update_rollback_failed_body = "Update mislukt. Je oorspronkelijke bestand staat op: "
+3
View File
@@ -44,3 +44,6 @@ chatgpt_token_expired_title = "Codex 工作階段已過期"
chatgpt_token_expired_body = "請重新登入以繼續追蹤使用量。"
threshold_80_body = "接近 5 小時上限。"
threshold_95_body = "上限將至 — 建議稍作休息。"
update_applied_title = "已套用更新"
update_applied_body = "已更新至 v"
update_rollback_failed_body = "更新失敗。您的原始執行檔已保存於: "
+7
View File
@@ -68,6 +68,13 @@ pub struct LocaleStrings {
pub threshold_80_body: String,
/// Body text for the 95% threshold balloon.
pub threshold_95_body: String,
/// Title for the tray balloon shown on first launch after an auto-update.
pub update_applied_title: String,
/// Prefix for the tray balloon body. Call site appends the version (e.g. "0.1.10").
pub update_applied_body: String,
/// Prefix for the rollback-failed MessageBox body. Call site appends
/// the backup path and a separator with the expected target filename.
pub update_rollback_failed_body: String,
}
#[derive(Deserialize)]
+29 -3
View File
@@ -36,8 +36,34 @@ fn main() {
std::process::exit(exit_code);
}
if diagnose_enabled {
log::info!("entering app::run");
let wait_pid = args
.iter()
.position(|a| a == "--wait-pid")
.and_then(|i| args.get(i + 1))
.and_then(|s| s.parse::<u32>().ok());
if let Some(pid) = wait_pid {
if diagnose_enabled {
log::info!("waiting up to 5s for parent pid {pid} to exit");
}
update::handoff::wait_for_parent_exit(pid, 5_000);
}
app::run();
let updated_to = args
.iter()
.position(|a| a == "--updated-to")
.and_then(|i| args.get(i + 1))
.cloned();
if diagnose_enabled {
log::info!("entering app::run (wait_pid={wait_pid:?} updated_to={updated_to:?})");
}
app::run(AppArgs {
wait_pid_present: wait_pid.is_some(),
updated_to,
});
}
pub struct AppArgs {
pub wait_pid_present: bool,
pub updated_to: Option<String>,
}
+20 -5
View File
@@ -11,8 +11,8 @@ use std::sync::{Mutex, OnceLock};
use windows::Win32::Foundation::HWND;
use windows::Win32::UI::Shell::{
Shell_NotifyIconW, NIF_ICON, NIF_INFO, NIF_MESSAGE, NIF_TIP, NIIF_WARNING, NIM_ADD,
NIM_DELETE, NIM_MODIFY, NOTIFYICONDATAW,
Shell_NotifyIconW, NIF_ICON, NIF_INFO, NIF_MESSAGE, NIF_TIP, NIIF_INFO, NIIF_WARNING,
NIM_ADD, NIM_DELETE, NIM_MODIFY, NOTIFYICONDATAW, NOTIFY_ICON_INFOTIP_FLAGS,
};
use windows::Win32::UI::WindowsAndMessaging::DestroyIcon;
@@ -81,13 +81,28 @@ pub fn sync(owner: HWND, desired: &[TrayIcon]) {
}
}
/// Show a balloon notification on an already-registered icon.
pub fn notify(owner: HWND, kind: IconKind, title: &str, body: &str) {
/// Show a yellow-warning balloon on an already-registered icon.
pub fn notify_warning(owner: HWND, kind: IconKind, title: &str, body: &str) {
notify_inner(owner, kind, title, body, NIIF_WARNING);
}
/// Show a blue-info balloon on an already-registered icon.
pub fn notify_info(owner: HWND, kind: IconKind, title: &str, body: &str) {
notify_inner(owner, kind, title, body, NIIF_INFO);
}
fn notify_inner(
owner: HWND,
kind: IconKind,
title: &str,
body: &str,
flags: NOTIFY_ICON_INFOTIP_FLAGS,
) {
let mut data = build_data(owner, kind);
data.uFlags = NIF_INFO;
write_utf16(&mut data.szInfoTitle, title);
write_utf16(&mut data.szInfo, body);
data.dwInfoFlags = NIIF_WARNING;
data.dwInfoFlags = flags;
unsafe {
let _ = Shell_NotifyIconW(NIM_MODIFY, &data);
}
+133
View File
@@ -0,0 +1,133 @@
// Native Win32 process + file handoff primitives used by the in-app
// restart path and the auto-update install path. The main binary uses
// `windows_subsystem = "windows"`, so spawning the child directly via
// `CreateProcessW` allocates no console — nothing can flash.
use std::ffi::OsString;
use std::io;
use std::os::windows::ffi::OsStrExt;
use std::path::Path;
use windows::core::PCWSTR;
use windows::Win32::Foundation::{CloseHandle, FALSE, HANDLE, WAIT_OBJECT_0};
use windows::Win32::System::Threading::{
CreateProcessW, OpenProcess, WaitForSingleObject, CREATE_NEW_PROCESS_GROUP,
CREATE_NO_WINDOW, DETACHED_PROCESS, PROCESS_INFORMATION, PROCESS_SYNCHRONIZE,
STARTUPINFOW,
};
/// Spawn `exe` with the supplied args as a detached, console-less child.
///
/// Caller is fire-and-forget: the child's handles are closed immediately
/// so no zombie wait is required.
pub fn spawn_detached(exe: &Path, args: &[OsString]) -> io::Result<()> {
let mut cmdline = build_command_line(exe, args);
let si = STARTUPINFOW {
cb: std::mem::size_of::<STARTUPINFOW>() as u32,
..Default::default()
};
let mut pi = PROCESS_INFORMATION::default();
let flags = CREATE_NO_WINDOW | DETACHED_PROCESS | CREATE_NEW_PROCESS_GROUP;
let ok = unsafe {
CreateProcessW(
PCWSTR::null(),
windows::core::PWSTR(cmdline.as_mut_ptr()),
None,
None,
FALSE,
flags,
None,
PCWSTR::null(),
&si,
&mut pi,
)
};
if ok.is_err() {
return Err(io::Error::last_os_error());
}
unsafe {
if !pi.hThread.is_invalid() {
let _ = CloseHandle(pi.hThread);
}
if !pi.hProcess.is_invalid() {
let _ = CloseHandle(pi.hProcess);
}
}
// Suppress the unused-variable warning until si.lpReserved fields ever matter.
let _ = &si;
Ok(())
}
/// Wait up to `timeout_ms` for `pid` to exit. Silent on any failure —
/// caller treats this as a best-effort barrier before acquiring the
/// singleton mutex.
pub fn wait_for_parent_exit(pid: u32, timeout_ms: u32) {
let handle: HANDLE = match unsafe { OpenProcess(PROCESS_SYNCHRONIZE, FALSE, pid) } {
Ok(h) if !h.is_invalid() => h,
_ => return,
};
unsafe {
let res = WaitForSingleObject(handle, timeout_ms);
if res != WAIT_OBJECT_0 {
log::debug!("wait_for_parent_exit pid={pid} timeout/err res={:?}", res.0);
}
let _ = CloseHandle(handle);
}
}
/// Remove leftover `<exe>.old.<pid>` siblings from previous in-place updates.
/// Filled in by phase 4; stubbed here so phase 1 can wire the call sites.
pub fn cleanup_stale_old_exes(current_exe: &Path) {
let Some(dir) = current_exe.parent() else {
return;
};
let Some(stem) = current_exe.file_name() else {
return;
};
let prefix = format!("{}.old.", stem.to_string_lossy());
let entries = match std::fs::read_dir(dir) {
Ok(e) => e,
Err(_) => return,
};
for entry in entries.flatten() {
let name = entry.file_name();
if name.to_string_lossy().starts_with(&prefix) {
if let Err(e) = std::fs::remove_file(entry.path()) {
log::debug!(
"cleanup_stale_old_exes: remove {:?} failed: {e}",
entry.path()
);
}
}
}
}
fn build_command_line(exe: &Path, args: &[OsString]) -> Vec<u16> {
// CreateProcessW parses argv[0] from a quoted exe path. We wrap the
// exe in `"…"` and join args separated by spaces. Args are quoted
// only when they contain whitespace; our callers pass simple tokens
// (--wait-pid <number>, --updated-to <version>) so naive quoting is
// sufficient.
let mut line = String::new();
line.push('"');
line.push_str(&exe.to_string_lossy());
line.push('"');
for a in args {
line.push(' ');
let s = a.to_string_lossy();
if s.chars().any(|c| c.is_whitespace()) {
line.push('"');
line.push_str(&s);
line.push('"');
} else {
line.push_str(&s);
}
}
let mut wide: Vec<u16> = std::ffi::OsString::from(line).encode_wide().collect();
wide.push(0);
wide
}
+127 -43
View File
@@ -1,42 +1,47 @@
// Download a release asset and hand off via inline `cmd /c`.
// Download a release asset and swap it in via native Win32 calls.
//
// We avoid the helper-exe pattern entirely: after writing the new .exe
// to a staging path, we spawn cmd.exe with a one-liner that waits 2 s,
// moves the new binary over the running one (Windows releases the file
// lock when our process exits), and relaunches it.
// After writing the new .exe to a staging path and verifying its
// SHA-256, we `MoveFileExW` the running exe sideways (so Windows
// releases the file lock on our own image), then `MoveFileExW` the
// staged exe into place, then spawn the new binary detached via
// `handoff::spawn_detached`. No shell, no console allocation.
use std::os::windows::process::CommandExt;
use std::path::PathBuf;
use std::process::{Command, Stdio};
use std::ffi::OsString;
use std::path::{Path, PathBuf};
use sha2::{Digest, Sha256};
use crate::net::Client;
use windows::core::PCWSTR;
use windows::Win32::Storage::FileSystem::{
MoveFileExW, MOVEFILE_COPY_ALLOWED, MOVEFILE_REPLACE_EXISTING, MOVE_FILE_FLAGS,
};
use windows::Win32::System::Threading::GetCurrentProcessId;
use windows::Win32::UI::WindowsAndMessaging::{
MessageBoxW, MB_ICONERROR, MB_OK,
};
const CREATE_NO_WINDOW: u32 = 0x0800_0000;
const DETACHED_PROCESS: u32 = 0x0000_0008;
use crate::net::Client;
use crate::os::to_utf16_nul;
pub fn begin(http: &Client, release: &super::Release) -> Result<(), super::Error> {
let current = std::env::current_exe()?;
ensure_writable(&current)?;
let staging = stage_path()?;
// Refuse to proceed if either path contains `%`. Inside double quotes
// cmd.exe still expands `%var%` references, so a path containing `%`
// would let cmd substitute environment variables into the swap step.
// Such paths are vanishingly rare on real Windows installs; failing
// fast is safer than rolling a bespoke cmd-escape layer.
// Defense in depth: `MoveFileExW` itself is immune to `%`-expansion
// (no shell parses our paths), but the existing rejection guards
// future code paths that might invoke external tools, so keep it.
reject_unsafe_path(&current)?;
reject_unsafe_path(&staging)?;
if let Some(parent) = staging.parent() {
std::fs::create_dir_all(parent)?;
}
download(http, &release.asset_url, &staging, release.asset_sha256.as_ref())?;
spawn_handoff(&staging, &current)?;
swap_and_spawn(&staging, &current, &release.version)?;
Ok(())
}
/// CLI entry-point compatibility for `--apply-update <target> <source> <pid>`.
/// The inline-cmd handoff already does the swap-and-restart; if this binary
/// The native handoff already does the swap-and-restart; if this binary
/// is invoked with the legacy flag (e.g. from an older release's helper)
/// just exit cleanly so the upgrade still completes.
pub fn run_cli(args: &[String]) -> Option<i32> {
@@ -50,7 +55,7 @@ pub fn run_cli(args: &[String]) -> Option<i32> {
fn download(
http: &Client,
url: &str,
to: &std::path::Path,
to: &Path,
expected_sha256: Option<&[u8; 32]>,
) -> Result<(), super::Error> {
let resp = http
@@ -86,40 +91,119 @@ fn hex_encode(bytes: &[u8]) -> String {
out
}
fn reject_unsafe_path(p: &std::path::Path) -> Result<(), super::Error> {
fn reject_unsafe_path(p: &Path) -> Result<(), super::Error> {
let s = p.to_string_lossy();
if s.contains('%') {
return Err(super::Error::UnsafePath(format!(
"path contains '%' which cmd.exe expands as a variable: {s}"
"path contains '%': {s}"
)));
}
Ok(())
}
fn spawn_handoff(source: &std::path::Path, target: &std::path::Path) -> Result<(), super::Error> {
let src_str = source.to_string_lossy().replace('"', "");
let tgt_str = target.to_string_lossy().replace('"', "");
// 2-second wait gives the current process time to exit and release the
// file lock before `move` overwrites it.
let cmd = format!(
r#"timeout /t 2 /nobreak >nul & move /y "{src_str}" "{tgt_str}" & start "" "{tgt_str}""#
);
// raw_arg bypasses Rust's std auto-escaping which would turn the inner
// `"` characters into `\"`. cmd.exe does not recognise `\"`, so the
// escaped form makes `start` see the path as `\\` and emit a
// "Windows cannot find '\\'" dialog. Feeding the command line raw
// preserves the quotes cmd.exe actually expects.
Command::new("cmd.exe")
.raw_arg("/c")
.raw_arg(format!("\"{cmd}\""))
.creation_flags(CREATE_NO_WINDOW | DETACHED_PROCESS)
.stdin(Stdio::null())
.stdout(Stdio::null())
.stderr(Stdio::null())
.spawn()?;
fn swap_and_spawn(
source: &Path,
target: &Path,
version: &super::release::Version,
) -> Result<(), super::Error> {
let backup = backup_path(target);
// Step 1: rename running exe sideways. Windows allows renaming a
// file even while its image is mapped into memory; this releases
// the lock on the original `target` path. Same directory by
// construction, so plain MoveFileExW with no flags is sufficient.
move_file(target, &backup, MOVE_FILE_FLAGS(0))?;
// Step 2: move staged exe into place. Staging lives under
// %LOCALAPPDATA%, target lives wherever the user installed —
// COPY_ALLOWED lets MoveFileExW fall back to copy+delete when
// the two paths cross volumes (portable installs on D:/E:/etc.).
let step2_flags = MOVEFILE_REPLACE_EXISTING | MOVEFILE_COPY_ALLOWED;
if let Err(swap_err) = move_file(source, target, step2_flags) {
// Best-effort revert. Same volume, no COPY_ALLOWED needed.
if let Err(revert_err) = move_file(&backup, target, MOVEFILE_REPLACE_EXISTING) {
log::error!("rollback also failed: {revert_err}; surfacing modal");
let target_name = target
.file_name()
.map(|s| s.to_string_lossy().into_owned())
.unwrap_or_else(|| "claude-code-usage-bubble.exe".to_string());
surface_rollback_failure(&backup, &target_name);
}
return Err(swap_err);
}
// Step 3: spawn the new exe detached with --wait-pid + --updated-to.
let pid = unsafe { GetCurrentProcessId() };
let version_str = format!("{}.{}.{}", version.major, version.minor, version.patch);
let args = vec![
OsString::from("--wait-pid"),
OsString::from(pid.to_string()),
OsString::from("--updated-to"),
OsString::from(version_str),
];
if let Err(spawn_err) = super::handoff::spawn_detached(target, &args) {
// New binary is on disk but won't auto-launch. Roll back so
// the user's next "Restart" stays on the known-good version.
log::error!("spawn_detached failed after swap: {spawn_err}; attempting revert");
if let Err(revert_err) = move_file(&backup, target, MOVEFILE_REPLACE_EXISTING) {
log::error!("post-spawn revert failed: {revert_err}");
}
return Err(super::Error::Io(spawn_err));
}
Ok(())
}
fn move_file(src: &Path, dst: &Path, flags: MOVE_FILE_FLAGS) -> Result<(), super::Error> {
let src_w = to_utf16_nul(&src.to_string_lossy());
let dst_w = to_utf16_nul(&dst.to_string_lossy());
let result = unsafe {
MoveFileExW(
PCWSTR::from_raw(src_w.as_ptr()),
PCWSTR::from_raw(dst_w.as_ptr()),
flags,
)
};
result.map_err(|e| {
super::Error::SwapFailed(format!(
"MoveFileExW({} -> {}): {e}",
src.display(),
dst.display()
))
})
}
fn backup_path(target: &Path) -> PathBuf {
let pid = unsafe { GetCurrentProcessId() };
let fname = target
.file_name()
.map(|s| s.to_string_lossy().into_owned())
.unwrap_or_else(|| "exe".to_string());
let mut p = target.to_owned();
p.set_file_name(format!("{fname}.old.{pid}"));
p
}
fn surface_rollback_failure(backup: &Path, target_name: &str) {
// Pull the localized body from i18n; the caller passes the
// user-meaningful filename so we can format it in-place.
let strings = crate::i18n::I18n::load(None).strings().clone();
let body = format!(
"{}{}\n\n{}",
strings.update_rollback_failed_body,
backup.display(),
target_name
);
let title_w = to_utf16_nul(&strings.update_failed);
let body_w = to_utf16_nul(&body);
unsafe {
MessageBoxW(
None,
PCWSTR::from_raw(body_w.as_ptr()),
PCWSTR::from_raw(title_w.as_ptr()),
MB_OK | MB_ICONERROR,
);
}
}
fn stage_path() -> Result<PathBuf, super::Error> {
let base = dirs::data_local_dir().ok_or_else(|| {
super::Error::NotWritable("no local data directory available".to_string())
@@ -130,7 +214,7 @@ fn stage_path() -> Result<PathBuf, super::Error> {
.join("update.exe"))
}
fn ensure_writable(target: &std::path::Path) -> Result<(), super::Error> {
fn ensure_writable(target: &Path) -> Result<(), super::Error> {
let parent = target.parent().ok_or_else(|| {
super::Error::NotWritable("could not resolve install directory".to_string())
})?;
+6 -2
View File
@@ -1,10 +1,12 @@
// Self-update subsystem.
//
// Two stages: `release::fetch_latest` checks GitHub releases for a newer
// build; `install::begin` downloads the .exe and hands off to a detached
// `cmd /c` script that swaps the binary and restarts.
// build; `install::begin` downloads the .exe, swaps it in via native
// `MoveFileExW`, then spawns the new binary detached via
// `CreateProcessW`. No shell handoff — nothing can flash a console.
pub mod channel;
pub mod handoff;
pub mod install;
pub mod release;
@@ -24,6 +26,8 @@ pub enum Error {
ChecksumMismatch { expected: String, actual: String },
#[error("path rejected for safety: {0}")]
UnsafePath(String),
#[error("file swap failed: {0}")]
SwapFailed(String),
}
pub use channel::{current as current_channel, Channel};