fix: keep UTF-8 text in run and tail output (#274)

Commit `ce8b2bd`

Jay Hemnani · 2026-10-16 09:27:21 -0400 EDT

parent 36e024e

fix: keep UTF-8 text in run and tail output (#274)

stripAnsi fed each byte to the bare VT parser, which expects decoded
input. It dropped every non-ASCII byte, and a UTF-8 continuation byte in
the C1 range (Ý is C3 9D, 0x9D is the 8-bit OSC introducer) started a
sequence that swallowed the text after it. Pass bytes >= 0x80 through
when the parser is not inside a sequence.

1 files changed, +20, -0

M src/util.zig

+20, -0

......@@ -493,6 +493,11 @@ pub fn stripAnsi(alloc: std.mem.Allocator, data: []const u8) ![]const u8 {

    var parser = ghostty_vt.Parser.init();
    for (data) |c| {
        // Bytes >= 0x80 outside a sequence are UTF-8 text, not C1 controls.
        if (c >= 0x80 and parser.state == .ground) {
            result.append(alloc, c) catch unreachable;
            continue;
        }
        const actions = parser.next(c);
        for (actions) |action_opt| {
            const action = action_opt orelse continue;

......@@ -2294,6 +2299,21 @@ test "stripAnsi: only escape sequences" {
    try testing.expectEqualStrings("", result);
}

test "stripAnsi: keeps UTF-8 text" {
    const alloc = testing.allocator;
    const result = try stripAnsi(alloc, "\x1b[1mcafé\x1b[0m – 日本 🎉\n");
    defer alloc.free(result);
    try testing.expectEqualStrings("café – 日本 🎉\n", result);
}

test "stripAnsi: UTF-8 bytes do not start escape sequences" {
    const alloc = testing.allocator;
    // Ý is C3 9D, and 0x9D on its own is the 8-bit OSC introducer
    const result = try stripAnsi(alloc, "Ýmir\nline two\n");
    defer alloc.free(result);
    try testing.expectEqualStrings("Ýmir\nline two\n", result);
}

test "isUserInput: kitty keyboard status reply is a terminal response" {
    // CSI ? flags u answers the program's CSI ? u query; no key was pressed
    try testing.expect(!isUserInput("\x1b[?1u"));

...	...		@@ -493,6 +493,11 @@ pub fn stripAnsi(alloc: std.mem.Allocator, data: []const u8) ![]const u8 {
493	493
494	494		var parser = ghostty_vt.Parser.init();
495	495		for (data) \|c\| {
	496	+	// Bytes >= 0x80 outside a sequence are UTF-8 text, not C1 controls.
	497	+	if (c >= 0x80 and parser.state == .ground) {
	498	+	result.append(alloc, c) catch unreachable;
	499	+	continue;
	500	+	}
496	501		const actions = parser.next(c);
497	502		for (actions) \|action_opt\| {
498	503		const action = action_opt orelse continue;

...	...		@@ -2294,6 +2299,21 @@ test "stripAnsi: only escape sequences" {
2294	2299		try testing.expectEqualStrings("", result);
2295	2300		}
2296	2301
	2302	+	test "stripAnsi: keeps UTF-8 text" {
	2303	+	const alloc = testing.allocator;
	2304	+	const result = try stripAnsi(alloc, "\x1b[1mcafé\x1b[0m – 日本 🎉\n");
	2305	+	defer alloc.free(result);
	2306	+	try testing.expectEqualStrings("café – 日本 🎉\n", result);
	2307	+	}
	2308	+
	2309	+	test "stripAnsi: UTF-8 bytes do not start escape sequences" {
	2310	+	const alloc = testing.allocator;
	2311	+	// Ý is C3 9D, and 0x9D on its own is the 8-bit OSC introducer
	2312	+	const result = try stripAnsi(alloc, "Ýmir\nline two\n");
	2313	+	defer alloc.free(result);
	2314	+	try testing.expectEqualStrings("Ýmir\nline two\n", result);
	2315	+	}
	2316	+
2297	2317		test "isUserInput: kitty keyboard status reply is a terminal response" {
2298	2318		// CSI ? flags u answers the program's CSI ? u query; no key was pressed
2299	2319		try testing.expect(!isUserInput("\x1b[?1u"));

Commit ce8b2bd

Commit `ce8b2bd`