Issue1599004

test_hmac and test_pep247 fail because of problems with the SHA module.

* SHA and MD5 objects lack a .digest_size attribute.

* Successive calls to digest() return different values.

* For digest(), the byte array is converted to a string using the default encoding, which produces weird results if it's UTF-8.  It should hardwire using ISO-8859-1, because we want a 1-to-1 mapping of characters.

The attached patch fixes test_pep247.  test_hmac still fails; I'm still looking into the problem.

Revised version of the patch against the trunk; this now fixes test_hmac.  The problem was the use of .getBytes(), which uses a system-specific encoding; here we need it to use ISO-8859-1.

I only see one problem in these changes: calling String.getBytes with ISO-8859-1 uses CodingErrorAction.REPLACE in its internal CharsetEncoder.  That just replaces characters above 255 with '?'.  I think we want it to throw an exception, right?

Unfortunately, CharsetEncoder appeared in Java 1.4 so we can't just use it directly for 2.2. As far as I can tell, there's no way in versions earlier than 1.4 to get that kind of control over the encoding in Java's library.  We may have to write a bit of code in PyString or the like that handles converting Strings to bytes and reports errors instead of eating them.  This applies to http://jython.org/bugs/1599900 as well.

Committed in r3038.