Fallout from bug 200049. We are using the wrong charset name when creating the unicode encoder/decoder. Instead of "UCS-2" we should be using "UTF-16".

I'll upload a patch based on my investigation to fix bug 206811

This is not the best patch, but it should work as well as what we have now in most cases. One exception is when the UTF-16 module of glibc is not installed. Another case is when a binary compiled with a recent glibc (with UTF-16LE/UTF-16BE) is run where an old glibc(without UTF-16LE/BE) is used. As I noted in the comment, a better way is to add a run-time detection code to xpcom.

Comment on attachment 127855 [details] [diff] [review] a tentative patch >Index: intl/uconv/src/nsCharsetConverterManager.cpp >+#if __GLIBC_PREREQ(2, 3) >+#define UTF16_NAME UTF-16LE Ooops. Needless to say, I have to quote UTF-16LE (and other values). #define UTF16_NAME "UTF-16LE"

Comment on attachment 127855 [details] [diff] [review] a tentative patch >+#ifdef MOZ_USE_NATIVE_UCONV >+#ifdef __GLIBC__ ....... >+#else >+#define UTF16_NAME UCS-2 >+#endif Another #endif is needed here.

mass reassigning to nobody.

Native uconv is gone.