multiple utf-8 string vulnerabilities

[marcusoverhagen]

Multiple UTF-8 functions are unsafe and vulnerable to denial of service attacks or buffer overflow attacks.

This includes:

count_utf8_bytes

UTF8CountBytes

UTF8CountChars

UTF8ToCharCode

UTF8ToLength

It is dangerous to look only at the first byte to determine how long an character sequence is. A malformed sequence can be used to skip a string's terminating zero byte.

Example (2 bytes and terminating zero):

0xE0 0x81 0x00

The first byte tells that it's 3 bytes long, and the above mentioned functions will skip the terminating zero.

[axeld]

Yes, I noticed that, too, when I fixed that byte count bug a few days ago. I'll give Michael another try :-)

[mmlr]

Fixed in r19624.