wishlist: add support for larger message digests from SHA-2 family

[dkg]

Currently, python-crypto appears to support SHA256 as its largest message digest. it would be good to be able to support other message digests from the SHA-2 family as well (sha-512, sha-384, and sha-224, in particular).

[lorenz quack (lorenz.quack)]

I went ahead and implemented SHA-512.
I based it on the existing SHA256.c implementation which made it pretty easy.
The patch also includes some unit tests.

Caveat: It depends on the header stdint.h which is only required by C99 and furthermore it depends on stdint.h defining a uint64-t type. even in C99 this is only optional.

P.S.: I took the liberty to add myself to the ACKS file in this patch.

[lorenz quack (lorenz.quack)]

Here are some more patches which should be applied in this order:

02_SHA2_unification.patch : Introduces generic template for SHA-2 family hashes. change SHA-256 and SHA-512 to use it.
03_SHA224_SHA384.patch : Implements SHA-224 and SHA-384 also using the generic template
04_SHA2_clean.patch : Cleans up the code and adds the Public Domain disclaimers where I previously forgot them

Hope this is useful and comments are welcome!

[lorenz quack (lorenz.quack)]

Not sure if there is any interest in these patches.
But for completeness I should point out a bug I noticed in my patches.

I changed the macro DIGEST_SIZE to mean the size of the digest in Bits and introduced a macro DIGEST_SIZE_BYTES to replace the old one but forgot to change the hash_template accordingly.

The fix seems pretty trivial: Either the macros need to be renamed or the hash_template needs to be adapted.

Should I send in new patches or is this comment sufficient?

[dkg (dkg0)]

I suspect you should leave the existing DIGEST_SIZE as bytes, and introduce a new DIGEST_SIZE_BITS.

That would avoid breaking the semantics of the API for people who use the macro, if it is exposed somewhere.

[lorenz quack (lorenz.quack)]

fixed the naming issue of the macros in the SHA2-family.
Also added unittests checking for the digest_size attribute at a module and instance level.

[Darsey Litzenberger (dlitz)]

Fixed in v2.5