ucnv_toUChars from UTF32 doesn't work with srcLength=-1

This works as designed. This isn't the only function that works like this.

If you really have data written as UChar, then you're wasting a lot of time during the conversion process.

Generally speaking, UChar32 (UTF-32) isn't used by most programs unless it's used for wchar_t. ICU has almost no APIs that use UChar32 strings. If you're interpreting wchar_t strings, then this conversion process your performing is incorrect. You should be using u_strToWCS and u_strFromWCS instead. The wchar_t type isn't always Unicode based, and these functions handle the non-Unicode whcar_t types. You can use UnicodeString::getBuffer(minCapacity) and UnicodeString::releaseBuffer with these functions to read or write to a UnicodeString.