[RFC] mb_str_split

Help me to find memory leak. Travis CI says:
010+ [Sun Jan 13 18:49:49 2019] Script:
'/home/travis/build/php/php-src/ext/mbstring/tests/mb_str_split_jp.php'
011+ /home/travis/build/php/php-src/ext/mbstring/mbstring.c(646) : Freeing
0x00007f59d2c02540 (66 bytes),

https://travis-ci.org/php/php-src/jobs/479101356

What to do next?

Hello, internals!
NikiC wrote very detailed review about my mb_str_split. So i rewrote the
function completely. While i was working on new implementation i've noticed
something in the mbfl library functions: mbfl_substr and mbfl_strlen.

if (encoding->flag & MBFL_ENCTYPE_SBCS) {
len = string->len;
} else if (encoding->flag & (MBFL_ENCTYPE_WCS2BE | MBFL_ENCTYPE_WCS2LE)) {
len = string->len/2;
} else if (encoding->flag & (MBFL_ENCTYPE_WCS4BE | MBFL_ENCTYPE_WCS4LE)) {
len = string->len/4;
}

There is 2 more 2-bytes width encodings: MBFL_ENCTYPE_MWC2BE (UTF16-BE),
MBFL_ENCTYPE_MWC2LE (UTF16-LE).

Is this a mistake or not?

Please check:
https://github.com/php/php-src/blob/30668755b64aa732246d952451f89d1fcfe581f0/ext/mbstring/libmbfl/mbfl/mbfilter.c#L659

There is 2 more 2-bytes width encodings: MBFL_ENCTYPE_MWC2BE
(UTF16-BE),
MBFL_ENCTYPE_MWC2LE (UTF16-LE).

UTF-16 is not a fixed-width 2-byte encoding.
Just like UTF-8 is not a fixed-width 1-byte encoding.

--
Lauri Kenttä