Code Review Request, JDK-8214339 : SSLSocketImpl erroneously wraps SocketException

Jamil Nimeh jamil.j.nimeh at
Fri Dec 14 19:49:01 UTC 2018

Looks pretty good.  I did have one question about a few of the methods 
in KeyShareExtension and PreSharedKeyExtension, specifically where you 
return nulls on error branches with the make-compiler-happy comments.  
In those cases would it be a bit cleaner to set a byte[] variable to 
null at the beginning of the method and then in the successful code path 
set the value to whatever byte array comes back from hmacs or hkdf 
operations?  At the end of the method you only need one return 
statement, rather than a return null on every error branch, many of 
which won't be executed due to method calls which always throw 
exceptions.  Not a big deal if you wish to leave things as they are 
since I know the current approach is used in many places in the code 
that this review doesn't touch.


On 12/14/2018 8:14 AM, Xue-Lei Fan wrote:
> Hi,
> Please review the update:
> In some cases, the SSLProtocolException or SSLHandshakeException may 
> be thrown if the underlying socket run into problems.  An application 
> may depends on the exception class for further action, for example 
> retry the connection with different parameters.
> This update is trying to separate the socket problem from the TLS 
> protocol or handshake problem, by using different exception classes.
> Thanks,
> Xuelei

More information about the security-dev mailing list