[yxshee]

strip_ansi_codes (and so measure_text_width, truncate_str, pad_str) only accepts ; between CSI parameters. SGR sequences that use the colon form from ITU T.416, like curly underline \x1b[4:3m or \x1b[38:2::255:128:0m, are cut off after the first number and the rest leaks into the text:

let s = "\x1b[4:3mwavy\x1b[0m \x1b[38:2::255:128:0morange\x1b[0m";
strip_ansi_codes(s)    // ":3mwavy :2::255:128:0morange", expected "wavy orange"
measure_text_width(s)  // 28, expected 11

This lets the parser take : wherever it takes ;, and makes the same change in the reference regex the DFA tests compare against (plus : in the exhaustive small-strings test).

The new strip_sgr_with_colon_separated_params test fails on main with left: ":3mwavy :2::255:128:0morange". make test, make lint and make format-check pass.

[djc]

Thanks.