In the bug report, the following fix was suggested:

"The fix to the issue should be simple, just move the initialization of the verificationResults Map BEFORE the SecureRandom initialization in JceSecurity.java"

Does that not work for some reason?

--Sean

On 8/19/20 1:13 AM, Xuelei Fan wrote:
On 8/18/2020 2:43 PM, Valerie Peng wrote:

Using a shared instance is surely faster. However, the API specified that the most preferred SecureRandom impl will be used. To ensure this for all scenarios, creating default SecureRandom obj will provide correct result but shared instance may not.
I understand your point.  It might not break the spec if a shared instance is used.  It depends on the understanding of "most preferred SecureRandom impl" in the context.


Apps can call other init functions which takes SecureRandom objects to avoid this default SecureRandom obj creation if needed.

Yes, it's an alternative solution.  If an application used the default SecureRandom, it would be nice if there is no performance regression.

The SecureRandom initialization may be not cheap in some circumstances. As this bug did not complain about the use of shared instance, it may be fine if we want to avoid the performance impact if the impact exists.

Just for your consideration.

Xuelei

Valerie

On 8/18/2020 2:10 PM, Xuelei Fan wrote:
Is there any performance impact?

Xuelei

On 8/18/2020 12:51 PM, Valerie Peng wrote:

Anyone has cycles to review this somewhat trivial changes? JceSecurity has this shared SecureRandom instance which may lead to NPE when certain 3rd party JCE provider is set as most preferred. Removing this shared instance and change to create default SecureRandom obj when needed.

Bug: https://bugs.openjdk.java.net/browse/JDK-8246383

Webrev: http://cr.openjdk.java.net/~valeriep/8246383/webrev.00/

Thanks,
Valerie

Reply via email to