Description of problem: With the update to nss-3.14 we acquired the ability to disable the ssl pkcs11 bypass at build time, a feature we requested, by setting a build time variable. A runtime variable allowed, and still does, a user to enable bypass. The upstream code quietly returns success even when the caller has requested bypass and the build doesn't have it. The user will not get the performance improvement the feature was intended to but on the on the other hand it will not break. Backward compatibility takes precedence. 


Version-Release number of selected component (if applicable):


How reproducible:


Steps to Reproduce:
1. set the runtime environment variable SSLBYPASS to 1
2. run any of the ssl client/server tests, selserv and testclnt.
  
Actual results: the test will fail as either the server or the client will trigger an error.

Expected results: The test will complete and pass.

Additional info: The Test Plan Review allowed us to catch this mistake. This was due to a misunderstanding of guidance given. Revisiting the thread in the upstream, https://bugzilla.mozilla.org/show_bug.cgi?id=745281#c22
states: 

"... you will need to return SECSuccess in RHEL because you can't change the API in RHEL releases. You can remove Bypass because is doesn't actually change the semantics, only the timing. ... "

Having done it via an isolated patch, nss-ssl-enforce-no-pkcs11-bypass.path, makes the fix easy. The fix consist on removing this undesired patch.