Ok, now I think I got what you mean. So it could regress in two ways,
one is what you wrote below, and the other I am seeing. So instead of
modifying the existing test case, I just add a new one which basically
is the same one in the previous webrev as follows:
http://cr.openjdk.java.net/~naoto/8038092/webrev.3/
This way they can detect those two possible regressions.
Naoto
On 7/3/14, 9:37 AM, Alan Bateman wrote:
On 03/07/2014 17:16, Naoto Sato wrote:
With the original test case, webrev.1 and webrev.2 both succeed. With
the modified test case (in webrev.2), webrev.1 fails but webrev.2
succeeds. The reason I changed the test case is to catch possible
regression where someone makes changes to the SharedSecret being
initialized at BidiBase class loading time (as in webrev.1).
I'm not sure that I get this, it actually looks to me that this change
to the test could mask a problem with a future change that would break
the shared secrets setup.
-Alan
