Instantiate both the RSA private and public keys when setting parameters

[philr]

The RSA e (exponent) and n (modulus) parameters are initially backed by the rsa_e and rsa_n fields. These fields are then used to instantiate the internal privateKey and publicKey objects in generatePrivateKeyIfParams() and generatePublicKeyIfParams(). However, both rsa_e and rsa_n are set to null after use. This means that once either of the private or public keys has been created, the other cannot be. Consequently, the RSA object can only be used for private or public functions, but not for both.

For example:

orig = OpenSSL::PKey::RSA.generate(2048)
copy = OpenSSL::PKey::RSA.new
[:e, :n, :d, :p, :q, :iqmp, :dmp1, :dmq1].each do |param|
  copy.send("#{param}=", orig.send(param))
end

# e and n were set first, copy will now act only as a public key

# public_encrypt will succeed
copy.public_encrypt('Test')

# private_encrypt will fail with an 'OpenSSL::PKey::RSAError: private key needed' exception
copy.private_encrypt('Test')

When the above code is run using the MRI standard library, both the public_encrypt and private_encrypt calls succeed.

This pull request modifies generatePrivateKeyIfParams() and generatePublicKeyIfParams() so that they will obtain the e and n parameters from either privateKey or publicKey if rsa_e and rsa_n have already been consumed.

[philr]

The build failure seems to be unrelated to this change (the same job was failing with the same error before).

[kares]

Thanks Phil, this is looking good. we probably do not need those private methods to be synchronized.

[philr]

@kares I have amended my commit to remove synchronized from the getPublicExponent() and getModulus() private methods.

Travis CI has done a new build. However, this time the jobs have all failed with an unrelated issue initializing the build environment.

[kares]

yep, restarted the build and all seems well, thank you for your effort.