allow opening browser on windows through wsl

[The64thRealm]

No description provided.

[cursor]

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

[claude]

Claude finished @The64thRealm's task in 1m 41s —— View job

---

Code Review

Nice, focused change. The WSL problem is real and well-diagnosed, and the dev command is now resilient to a browser that won't open (matching the PR intent). Both call sites (auth.rs, dev/mod.rs) are updated consistently, server::run still executes regardless of whether the browser opens, and the PowerShell single-quote escaping is handled correctly. The module doc comment explaining why the open crate falls short under WSL is excellent.

A few observations, none of them blockers:

1. Child process stdio is inherited — may print noise to the terminal (src/browser.rs:50-52)
run_ok uses cmd.status(), which lets the child inherit the CLI's stdout/stderr. If powershell.exe (or wslview) emits a warning/error, it lands in the user's terminal uninvited. Since you only care about the exit code, capturing output keeps things clean:

fn run_ok(cmd: &mut Command) -> bool {
    cmd.output().map(|o| o.status.success()).unwrap_or(false)
}

Fix this →

2. .status()/.output() blocks until the child exits (src/browser.rs:50)
wslview and Start-Process normally fork-and-return immediately, so this is fine in practice. Just flagging that if a launcher were to block (e.g., waits on the spawned browser), the CLI would hang. Low risk given the specific commands used.

3. .gitattributes is unrelated scope creep
The line-ending normalization is reasonable on its own, but it's orthogonal to "open browser on WSL." Two thoughts: (a) it may be cleaner as its own PR/commit; (b) * text=auto eol=lf can produce a large one-time renormalization diff on existing checkouts — consider a follow-up git add --renormalize . commit so the normalization is explicit rather than trickling in file-by-file.

4. Minor: is_wsl() "wsl" check is redundant (src/browser.rs:22)
WSL2 kernel strings already contain microsoft (e.g. ...-microsoft-standard-WSL2), so the extra s.contains("wsl") is harmless but never the deciding factor. Fine to leave.

Correctness check I couldn't fully verify locally: the dev command opens http://localhost:{port}. Under WSL2 localhost is forwarded between Windows and the WSL VM, so opening it from the Windows-side browser should reach the WSL-hosted server — worth a quick manual confirmation on a real WSL host if you haven't already.

Nit: std::io::Error::other requires Rust ≥ 1.74 — fine for any recent toolchain, just make sure your MSRV (if you track one) allows it.

Overall this is a solid, well-documented improvement. Only #1 is worth addressing before merge, and even that is cosmetic.

Note: I reviewed by reading the diff and surrounding context; I did not build/run since this environment doesn't have Doppler configured (doppler run -- cargo ... per CLAUDE.md).

[franzwarning]

looks good! go ahead and merge and I'll cut a release