[Botan-devel] Status of ECDSA (working)
lloyd at randombit.net
Sat Oct 11 15:49:09 EDT 2008
On Thu, Oct 09, 2008 at 09:43:19AM -0400, Jack Lloyd wrote:
> Just FYI, ECDSA/ECDH in 1.7.15 is not functional, so don't even bother
> (actually missing several vital classes from the source completely).
> n.r.b head is looking happier. However I'm seeing several test
> failures in the GF(p) math (originally found because the ECDSA key
> generation would fail its self-test and thrown an exception).
> InSiTo passes, so it appears I broke something.
Some debugging help from Falko Strenzke has helped me solve most of
these bugs. Currently on mainline, ECDSA is passing the suite of tests
provided with InSiTo. One notable failure: using the GCC-provided TR1
shared_ptr on 32-bit x86 with GCC 4.3 seems to cause memory corruption
(at least I am seeing as such on my machine). I was not able to
replicate the corruption with earlier versions of GCC, with Boost's
TR1 and any version of GCC, or with GCC 4.3 on x86-64. Currently Botan
defaults to using Boost's TR1 to avoid this.
Before I would really consider ECDSA working I want to add a much
larger set of testcases, generated by a third party (either standard
test vectors from IEEE 1363 or ANSI X9, or a set generated using
Crypto++). But everything I have tried so far works.
More information about the botan-devel