fix: stop libghostty stream logs from corrupting the boo ui viewport

[raphaelk-medsimples]

What

boo ui renders its live viewport through an in-process libghostty stream parser, and the UI runs in the user's terminal with stderr inherited. libghostty logs unimplemented sequences at info level under the stream scope (e.g. info(stream): OSC 1 (change icon) received and ignored), so those lines paint directly over the rendered pane and corrupt it until a manual C-a l redraw.

The session daemon doesn't have this problem because startDaemon already redirects its stderr to BOO_LOG//dev/null. The in-process UI path has no equivalent guard.

This is very visible with Powerlevel10k, which emits an OSC 1 on every prompt and on preexec — so cat ~/.zshrc (or any command) floods the viewport with received and ignored lines.

Fixes #64.

How

boo installs no std_options, so logging uses Zig defaults. This adds a std_options.logFn that drops info-and-below from the stream scope, leaving every other scope (.ui, .daemon, .window, …) untouched:

pub const std_options: std.Options = .{ .logFn = filteredLog };

fn filteredLog(comptime level, comptime scope, comptime format, args) void {
    if (scope == .stream and @intFromEnum(level) >= @intFromEnum(std.log.Level.info)) return;
    std.log.defaultLog(level, scope, format, args);
}

A more targeted alternative would be to redirect the UI process's stderr the way the daemon does, but that would also hide genuinely useful warn/err output; filtering only info(stream) keeps everything else visible.

Verification

Built v0.5.18 from source (Zig 0.15.2) and drove a session emitting one OSC 1, capturing BOO_LOG:

build	info(stream) lines
ReleaseSafe, before	1
ReleaseSafe, after (this PR)	0

• zig fmt --check build.zig build.zig.zon src test — clean
• zig build test — passes
• zig build -Doptimize=ReleaseSafe — compiles

Confirmed the cause is independent of optimize mode (both Debug and ReleaseSafe emit info(stream) without the filter), so this is the right lever rather than a build-flag change.

Notes

Opening as a draft for maintainer direction on the preferred approach (logFn filter vs. UI stderr redirect vs. log_scope_levels). Also — thanks for boo; the libghostty-backed state model and the automation surface are a pleasure to work with.

[kylecarbs]

Thank you! Will publish a patch shortly.