Re: libnettle/libhogweed WIP

[Eli Zaretskii]

> From: Ted Zlatanov <address@hidden>
> Date: Fri, 14 Jul 2017 15:43:53 -0400
> 
> EZ> Also, the 7th test fails for me as below; any idea?
> EZ>   Test test-gnutls-005-aead-ciphers condition:
> EZ>       (ert-test-failed
> EZ>        ((should
> EZ>    (gnutls-tests-hexstring-equal input reverse))
> EZ>   :form
> EZ>   (gnutls-tests-hexstring-equal "                " "                ►   ► 
>  ∟\231\202    \200")
> EZ>   :value nil))
> EZ>      FAILED  7/7  test-gnutls-005-aead-ciphers
> 
> It works for me, on Ubuntu 17.04, GnuTLS 3.5.6 and a fresh checkout.
> 
> Could you run with `make gnutls-tests GNUTLS_TEST_VERBOSE=1' and look at
> the output? This full output is normally silenced because it's so
> verbose, but in this case we'll need it because so many different test
> cases are attempted.

I've run the test under a debugger.  Tell me what you want to know.

What I see (without really understanding what should happen) is that
gnutls-symmetric-decrypt returns a string that is 32-byte long,
whereas the original input before encryption was a 16-byte string.  In
the decrypted string, the first 16 bytes are identical to the input
before encryption, the rest seem to be binary garbage.  When the
original 16-byte input is encrypted by gnutls-symmetric-encrypt, the
result is a 32-byte string.  All of this happens in the first call to
encryption and decryption functions produced by the loop in
test-gnutls-005-aead-ciphers.  Does this makes sense and/or ring any
bells?  Do you need me to provide any further information?

The version of GnuTLS library I have here is 3.4.15, btw.