Severity: CRITICAL (server panic on untrusted bytes)
Status: ✅ FIXED
Files: src/format.zig:735-768
expandTruncateInto truncates a status-format value to n characters:
var idx: usize = 0;
while (idx < value.len) {
const len = std.unicode.utf8ByteSequenceLength(value[idx]) catch 1;
if (count >= n) break;
count += 1;
idx += len;
}
...
try out.appendSlice(allocator, value[0..idx]);
utf8ByteSequenceLength validates only the lead byte. If value ends with a truncated sequence (e.g. a trailing 0xC3 with its continuation byte missing), len comes back 2, idx overshoots value.len, and value[0..idx] panics with a slice-out-of-bounds. The negative branch has the same shape: k overshoots before value[k..].
The values come from ctx.get() — status variables like pane_title, which the child process controls via OSC sequences. A child emitting a truncated UTF-8 sequence in its title crashes the server on the next status render. Any pane in any session can trigger it.
Clamp before slicing:
idx = @min(idx, value.len);
try out.appendSlice(allocator, value[0..idx]);
(same for k in the negative branch), or iterate with std.unicode.Utf8View so malformed sequences decode as U+FFFD instead of overshooting. A regression test should feed a value ending in 0xC3 / 0xE2 0x82 through the truncation path.
expandTruncateInto in src/format.zig clamps idx, j, and k to value.len using @min(idx + len, value.len) during iteration, and explicitly clamps indices before slicing value[0..idx] and value[k..].
Tests: expandTruncateInto with malformed UTF-8 does not panic — bug #484 tests positive and negative truncation on truncated 2-byte (\xC3) and 3-byte (\xE2\x82) sequences.