input, and unicode output, which have been broken due to the change of the size of Rune from two bytes to four bytes. (Unicode-le and unicode-be output have not been affected, since they work different, and produce correct output as long as sizeof(Rune) >= 2.) In all of the affected functions an extra array of N elements of type u16int is used to temporarily store the 16-bit unicode data. As this increases stack usage, perhaps a different, slightly more complicated solution might be preferred, where only a single `Rune buf[N]' is used for both temporary storage and the result. R=rsc CC=codebot http://codereview.appspot.com/1574041