[BZ #14090] Fix md5/sha512 with large block sizes

[Paul Eggert]

On 07/27/2012 09:42 AM, Roland McGrath wrote:
> the fix also touched the sha* code.  So there should be a test for
> that too.

It's not practical to test for that.
It'd need a machine where size_t is wider than 64 bits,
along with a single buffer with at least 16 EiB.

> can it be done without
> the giant buffer by calling __md5_process_bytes many times?

I'm afraid the test needs a single buffer with at least 4 GiB.