[Patch][BZ 6530] *printf and incomplete multibyte sequences returns -1 bogusly

[Jeff Law]

I've reviewed Andreas S.'s patch (as has Jonathan Nieder) and it's also 
the same patch that Rich Felker submitted (as far as I can tell 
independently).  I've also reviewed Jonathan's testsuite addition.

Based on the above, I'm approving and installing these changes to fix 
this ANSI compliance issue.

If anyone thinks I'm overstepping here, please let me know.