ValueError: mb_detect_encoding(): Argument #2 ($encodings) contains invalid encoding "UTF8"

[nicolas-grekas]

Description

The following code: https://3v4l.org/3cnGD/rfc#vgit.master

<?php
echo mb_detect_encoding('abc', 'UTF8, ASCII');

Resulted in this output:

Fatal error: Uncaught ValueError: mb_detect_encoding(): Argument #2 ($encodings) contains invalid encoding "UTF8"

But I expected this output instead:

UTF-8

This is a regression, see https://3v4l.org/3cnGD

And this works if the list is given as an array: https://3v4l.org/lVva0/rfc#vgit.master

PHP Version

PHP 8.4

Operating System

No response

[cmb69]

Not sure when or why this has been changed, but using "UTF-8" is still supported (https://3v4l.org/8HA6Q/rfc#output).

[youkidearitai]

Hmm. Seems could not catch at mbfl_name2encoding_ex.

mbfl_name2encoding_ex (name=0x7ff545cce8 "UTF8, ASCII", name_len=4) at /home/tekimen/src/php-src/ext/mbstring/libmbfl/mbfl/mbfl_encoding.c:338
338			if (key < sizeof(mbfl_encoding_ptr_list_after_hashing) / sizeof(mbfl_encoding_ptr_list_after_hashing[0])) {
(gdb) n
339				int8_t offset = mbfl_encoding_ptr_list_after_hashing[key];
(gdb) n
340				if (offset >= 0) {
(gdb) n
341					encoding = mbfl_encoding_ptr_list + offset;
(gdb) n
342					if (strncasecmp((*encoding)->name, name, name_len) == 0) {
(gdb) p offset
$28 = 45 '-'
(gdb) p encoding
$29 = (const mbfl_encoding * const *) 0x5556c95f78 <mbfl_encoding_ptr_list+360>
(gdb) p encoding->name
$30 = 0x5556a1f828 "ISO-8859-1"
(gdb) bt
#0  mbfl_name2encoding_ex (name=0x7ff545cce8 "UTF8, ASCII", name_len=4) at /home/tekimen/src/php-src/ext/mbstring/libmbfl/mbfl/mbfl_encoding.c:342
#1  0x00000055558f8464 in php_mb_parse_encoding_list (value=0x7ff545cce8 "UTF8, ASCII", value_length=11, return_list=0x7fffffd5e8, return_size=0x7fffffd5e0, persistent=false, arg_num=2)
    at /home/tekimen/src/php-src/ext/mbstring/mbstring.c:350
#2  0x00000055559077ac in zif_mb_detect_encoding (execute_data=0x7ff5413080, return_value=0x7fffffd688) at /home/tekimen/src/php-src/ext/mbstring/mbstring.c:3428

[cmb69]

According to git bisect, 3ad422e is the culprit. Maybe @nielsdos can have a look at this issue.

[youkidearitai]

UTF8 is alias of UTF-8. Originally, catch below:

php-src/ext/mbstring/libmbfl/mbfl/mbfl_encoding.c

Lines 369 to 378 in 3ad422e

	/* search aliases */
	for (encoding = mbfl_encoding_ptr_list; *encoding; encoding++) {
	if ((*encoding)->aliases) {
	for (const char **alias = (*encoding)->aliases; *alias; alias++) {
	if (strcasecmp(*alias, name) == 0) {
	return *encoding;
	}
	}
	}
	}

However, strcasecmp can't catch variable name in "UTF8, ASCII".
I understanding only this.

[nielsdos]

The problem is simply that the loops below should use strncasecmp to take into account the length. Will fix this evening.

[youkidearitai]

@nielsdos I tried create pull request. Feel free to comment. Thank you.
#15829