Subj : Re: Mixing up parameters in setProperty callback
To   : Sterling Bates <sblistserv@bigbangco.com>
From : Brendan Eich <brendan@meer.net>
Date : Mon Dec 15 2003 11:21 pm

Sterling Bates wrote:

>>The reason that symptom bites is because without the JS_ShutDown call,
>>the deflated string cache holds onto char[] strings decimated from JS
>>strings due to calls to JS_GetStringBytes, an ancient API that "cannot
>>fail" and that takes no cx parameter, associating each such
>>chopped-down-to-8-bit-char string copy with the 16-bit-char JSString
>>GC-thing whence it was chopped.
> 
> 
> Gotcha.  I'm guessing that JS_GetStringChars is the replacement.  That'll
> help tremendously.

Right -- JS_GetStringBytes goes back to 1996, when the engine lacked 
Unicode support(!).  JS started out using ISO-Latin-1 chars stored in 
bytes, in the Nav2/3 era.  Some pages stored locale-specific, multibyte 
character encodings in strings.  The dark ages.

JS_GetStringChars is the only way to get the full, unchopped chars, 
although sometimes, when you're debugging or hacking in known-ASCII 
settings, JS_GetStringBytes is still a win.  But it should be avoided in 
general for good i18n hygiene.

/be

.