The couch tiles stop minting a console window #411

Merged

enricobuehler merged 1 commits from worktree-win-msix-console-flash into main 2026-08-27 11:34:20 +00:00

Conversation 0 Commits 1 Files Changed 4

+83 -57

enricobuehler commented 2026-08-27 11:05:50 +00:00

Owner

Copy Link

Copy Source

Field report: launching the "Punktfunk Console" entry on Windows shows a terminal window before the couch UI appears.

Root cause

punktfunk-session.exe is deliberately console-subsystem — its stdout IS the shell↔session contract. Both couch entry points are GUI-subsystem processes with no console to lend it, so a spawn without CREATE_NO_WINDOW makes Windows mint a fresh one: a black terminal that then sits behind the couch UI for the whole session.

The shell's own stream spawns (clients/windows/src/spawn.rs) have carried the flag since the Windows port. The two --browse hand-offs never did:

• punktfunk-console.exe — the "Punktfunk Console" MSIX tile (AppxManifest.xml) and the Inno installer's {userprograms}\Punktfunk Console shortcut
• punktfunk-client --console — the Steam-shortcut / HTPC-startup-item path

Both are shipping today.

The second bug, in the same spawn

🛑 CREATE_NO_WINDOW on its own turns the cosmetic bug into a crash. stdout was inherited. With a console that was merely invisible output; with the flag, the inherited handle is invalid — and match_window (clients/session/src/main.rs:501) println!s the settled window size regardless of --json-status, so a launched title would panic the child mid-stream on the first report. stdout is now an explicit Stdio::null(), not an inheritance.

The shape of the fix

The two hand-offs were copy-paste twins, which is precisely how they drifted apart on one flag — so they become one file, clients/windows/src/couch.rs, that both roots include: the standalone punktfunk-console.exe bin by #[path], the shell as a module. session_binary() folds in there too, so there is one resolver rather than two. Net −41 lines.

Verification

• couch.rs type-checks and passes clippy -D warnings -W pedantic against x86_64-pc-windows-msvc, via a scratch crate that pulls the file in over a stubbed parent module — the crate itself cannot be cross-checked from macOS (SDL3, aws-lc-rs do not cross-build).
• cargo check -p punktfunk-client-windows (the non-Windows stub build) is clean.
• rustfmt --check clean on all four touched files.

⏳ On-glass verification on Windows is owed — please confirm the Console tile comes up with no terminal behind it, and that launching a title from the couch library still survives the match_window report.

Two notes for the reviewer

1. The plain "Punktfunk" tile was reported as flashing a console too, and this PR does not explain that one. Audited: main() → AttachConsole (attaches to an existing parent console, never allocates; a no-op under MSIX) → identity → deeplink → windows_reactor::bootstrap() (a no-op under package identity) → SDL in-process → WinUI. No child process is spawned on that path at all. The console window's title bar names the process that owns it; that string is what the next round needs.
2. Same trap, still latent: crates/pf-paths/src/lib.rs spawns icacls.exe at four sites with no creation flags. punktfunk-tray is GUI-subsystem and depends on pf-paths — it currently only calls config_dir()/published_mgmt_port() so it dodges them, but any tray call to create_private_dir/write_secret_file would pop 2–4 console windows. Left out of this PR deliberately: not reachable today, and not the reported bug.

Field report: launching the **"Punktfunk Console"** entry on Windows shows a terminal window before the couch UI appears. ## Root cause `punktfunk-session.exe` is deliberately **console-subsystem** — its stdout IS the shell↔session contract. Both couch entry points are GUI-subsystem processes with no console to lend it, so a spawn without `CREATE_NO_WINDOW` makes Windows mint a fresh one: a black terminal that then sits behind the couch UI for the whole session. The shell's own stream spawns (`clients/windows/src/spawn.rs`) have carried the flag since the Windows port. The two `--browse` hand-offs never did: - `punktfunk-console.exe` — the **"Punktfunk Console" MSIX tile** (`AppxManifest.xml`) *and* the Inno installer's `{userprograms}\Punktfunk Console` shortcut - `punktfunk-client --console` — the Steam-shortcut / HTPC-startup-item path Both are shipping today. ## The second bug, in the same spawn 🛑 **`CREATE_NO_WINDOW` on its own turns the cosmetic bug into a crash.** stdout was inherited. With a console that was merely invisible output; with the flag, the inherited handle is **invalid** — and `match_window` (`clients/session/src/main.rs:501`) `println!`s the settled window size **regardless of `--json-status`**, so a launched title would panic the child mid-stream on the first report. stdout is now an explicit `Stdio::null()`, not an inheritance. ## The shape of the fix The two hand-offs were copy-paste twins, which is precisely how they drifted apart on one flag — so they become **one file**, `clients/windows/src/couch.rs`, that both roots include: the standalone `punktfunk-console.exe` bin by `#[path]`, the shell as a module. `session_binary()` folds in there too, so there is one resolver rather than two. Net −41 lines. ## Verification - `couch.rs` type-checks **and** passes `clippy -D warnings -W pedantic` against `x86_64-pc-windows-msvc`, via a scratch crate that pulls the file in over a stubbed parent module — the crate itself cannot be cross-checked from macOS (SDL3, aws-lc-rs do not cross-build). - `cargo check -p punktfunk-client-windows` (the non-Windows stub build) is clean. - `rustfmt --check` clean on all four touched files. ⏳ **On-glass verification on Windows is owed** — please confirm the Console tile comes up with no terminal behind it, and that launching a title from the couch library still survives the `match_window` report. ## Two notes for the reviewer 1. The **plain "Punktfunk" tile was reported as flashing a console too, and this PR does not explain that one.** Audited: `main()` → `AttachConsole` (attaches to an existing parent console, never allocates; a no-op under MSIX) → identity → deeplink → `windows_reactor::bootstrap()` (a no-op under package identity) → SDL in-process → WinUI. No child process is spawned on that path at all. The console window's title bar names the process that owns it; that string is what the next round needs. 2. **Same trap, still latent:** `crates/pf-paths/src/lib.rs` spawns `icacls.exe` at four sites with no creation flags. `punktfunk-tray` is GUI-subsystem and depends on pf-paths — it currently only calls `config_dir()`/`published_mgmt_port()` so it dodges them, but any tray call to `create_private_dir`/`write_secret_file` would pop 2–4 console windows. Left out of this PR deliberately: not reachable today, and not the reported bug.

enricobuehler added 1 commit 2026-08-27 11:05:53 +00:00

fix(clients/windows): the couch tiles stop minting a console window ... de19a98f1b

`punktfunk-session` keeps the CONSOLE subsystem on purpose — its stdout is the
shell/session contract. Both couch entry points are GUI processes with no console
to lend it, so a spawn without CREATE_NO_WINDOW makes Windows mint one: a black
terminal that then sits behind the couch UI for the whole session. The shell's
own stream spawns (spawn.rs) have carried the flag since the port; the two
`--browse` hand-offs never did — the "Punktfunk Console" MSIX tile and
`punktfunk-client --console` (Steam shortcut, HTPC startup item) hit it every launch.

The two hand-offs were copy-paste twins, which is how they drifted apart on one
flag, so they become one file (couch.rs) that both roots include — the standalone
`punktfunk-console.exe` bin by `#[path]`, the shell as a module. `session_binary()`
folds in there too, so there is one resolver rather than two.

Second, quieter bug in the same spawn: stdout was inherited. With a console that
was merely invisible output; with CREATE_NO_WINDOW the inherited handle is invalid,
and `match_window` reports the settled window size on stdout REGARDLESS of
`--json-status` — a `println!` onto a dead handle panics the child mid-stream. So
stdout is an explicit `Stdio::null()`, not an inheritance.

Verified: couch.rs type-checks and clippies clean against x86_64-pc-windows-msvc in
an isolated harness (the crate itself cannot be cross-checked from macOS — SDL3 and
aws-lc-rs do not cross-build). On-glass verification on Windows is owed.

enricobuehler merged commit 0b473cdb38 into main 2026-08-27 11:34:20 +00:00

enricobuehler deleted branch worktree-win-msix-console-flash 2026-08-27 11:34:26 +00:00

Sign in to join this conversation.

Reviewers

No Reviewers

Labels

No items

No labels

Milestone

No items

No Milestone

Projects

Clear projects

No projects

Assignees

Clear assignees

FernandoMX FlipperDan HitFrostbite bluelightspecial enricobuehler (Enrico Bühler) kszlim luxus nannku neoney

No Assignees

1 Participants

Notifications

Subscribe

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: unom/punktfunk#411