dmitry, plz look at this, thanks :)

if I change the script to:
<?php
declare(encoding="gb2312");
$s = "洪仁玕";
var_dump($s);
?>

also set fenc=cp936, then a memory leak will be reported:

$ php54 -d mbstring.internal_encoding=utf8 -dzend.script_encoding=cp936 
/tmp/1.php 
string(7) "洪仁?"
[Tue Nov 15 16:26:36 2011]  Script:  '/tmp/1.php'
***php-src/trunk/ext/mbstring/mbstring.c(612) :  Freeing 0x2A95DDDF68 (131 
bytes), script=/tmp/1.php
=== Total 1 memory leaks detected ===

the same script as above, will trigger a abort in this way:

$php54 -d mbstring.internal_encoding=cp936 /tmp/1.php 

php: Zend/zend_language_scanner.l:126: encoding_filter_script_to_internal: 
Assertion `internal_encoding && 
zend_multibyte_check_lexer_compatibility(internal_encoding)' failed.
Aborted (core dumped)

seems the characters is lost in the mbfl_buffer_converter_feed2 (called in 
zend_multibyte_encoding_converter)

actully, there is a more simple reproduce script:
<?php
$s = "洪仁";
var_dump(mb_convert_encoding($s, "utf8", "gbk"));
?>

save script in fenc=cp936

than, in php 5.4 this result a :
string(3) "洪"

characters lost while converting from cp936 to utf8. this is really a big 
problem.....

I have make a fix, re-assign to me, thanks dmitry ;)

This bug has been fixed in SVN.

Snapshots of the sources are packaged every three hours; this change
will be in the next snapshot. You can grab the snapshot at
http://snaps.php.net/.

 For Windows:

http://windows.php.net/snapshots/
 
Thank you for the report, and for helping us make PHP better.