crypto: clear error stack in ECDH::Initialize

[indutny]

Clean up OpenSSL error stack in ECDH::Initialize, some curves have
faulty implementations that are leaving dangling errors after
initializing the curve.

Fix: #4686

[indutny]

R: @nodejs/crypto
CI: https://ci.nodejs.org/job/node-test-pull-request/1232/

[thefourtheye]

LGTM if CI is green.

[shigeki]

Does it mean the Oakley curve is broken in openssl?

$ openssl ecparam -list_curves
...
Oakley-EC2N-3:
IPSec/IKE/Oakley curve #3 over a 155 bit binary field.
Not suitable for ECDSA.
Questionable extension field!
Oakley-EC2N-4:
IPSec/IKE/Oakley curve #4 over a 185 bit binary field.
Not suitable for ECDSA.
Questionable extension field!

Instead of ignoring it, how about the idea to have ERR_peek_error() and throw an error in ECDH::New?

[indutny]

@shigeki I don't really want to introduce more ERR_peek_error() checks until we decide what to do with the ERR stack in general. Every check added means that it may throw error on dirty stack.

[deian]

thanks!

[shigeki]

@indutny I found the error is intentional and ignoring it is a right choice, which comes from

https://github.com/openssl/openssl/blob/OpenSSL_1_0_2-stable/crypto/ec/ec_lib.c#L321-L325

We ignore the return value because some groups have an order with factors of two, which makes the Montgomery setup fail. |group->mont_data| will be NULL in this case.

CI was green except existing failures on WIn. LGTM and landed in ebd9add. Thanks.

[indutny]

Thank you!