333408 - native uconv: character encoding name issues

Status: RESOLVED WONTFIX

(Core :: Internationalization, defect)

Description

[Jungshik Shin]

There are a couple of issues with character encoding names in the current implementation of native uconv. 

Firstly, no name mapping is done to bridge the differences in naming between native character encoding names (as understood by iconv) and our canonical names. See, for instnace, 

http://lxr.mozilla.org/seamonkey/source/intl/uconv/src/nsCharsetConverterManager.cpp#194

191 #ifdef MOZ_USE_NATIVE_UCONV
192   if (mNativeUC) {
193     nsCOMPtr<nsISupports> supports;
194     mNativeUC->GetNativeConverter("UCS-2", 
195                                   aDest,
196                                   getter_AddRefs(supports));

This doesn't work in some cases although GNU iconv does recognize many different names. It's far far worse with other implementations of iconv (as found on Solaris, AIX, Tru64, etc). 

Secondly, what PRUnichar* contains(points to) is NOT UCS-2 BUT UTF-16 (in the native endian) !! However, just using 'UTF-16' in place of 'UCS-2' when calling |GetNativeConverter| (which is just a wrapper over iconv_open) doesn't work on many platforms, either. See what we have done in xpcom/io/nsNativeCharsetUtils.cpp to work around this problem.

Comment 1

[Doug Turner (:dougt)]

mass reassigning to nobody.

Comment 2

[Simon Montagu :smontagu]

Native uconv is gone.