On 02/12/12 22:00, Konstantin Kolinko wrote:
2012/11/29 <ma...@apache.org>:
Author: markt
Date: Thu Nov 29 14:35:02 2012
New Revision: 1415184
URL: http://svn.apache.org/viewvc?rev=1415184&view=rev
Log:
Fix https://issues.apache.org/bugzilla/show_bug.cgi?id=54190
Improve unit tests.
Patch by Brian Burch.
Modified:
tomcat/tc7.0.x/trunk/ (props changed)
tomcat/tc7.0.x/trunk/test/org/apache/catalina/authenticator/TestNonLoginAndBasicAuthenticator.java
tomcat/tc7.0.x/trunk/webapps/docs/changelog.xml
Propchange: tomcat/tc7.0.x/trunk/
------------------------------------------------------------------------------
Merged /tomcat/trunk:r1415177-1415179
+ private static final int SHORT_TIMEOUT_MINS = 1;
+ private static final int LONG_TIMEOUT_MINS = 2;
+ private static final int MANAGER_SCAN_DELAY_SECS = 60;
+ private static final int EXTRA_DELAY_SECS = 5;
+ private static final long TIMEOUT_DELAY_MSECS =
+ (((SHORT_TIMEOUT_MINS * 60)
+ + MANAGER_SCAN_DELAY_SECS + EXTRA_DELAY_SECS) * 1000);
...
+ // allow the session to time out and lose authentication
+ Thread.sleep(TIMEOUT_DELAY_MSECS);
According to Buildbot logs, testBasicLoginSessionTimeout() runs for 2
minutes (125 seconds), mainly due to a sleep() call.
I wish there were a way to speed up this test.
My thoughts
a) Maybe use reflection to reduce StandardSession.lastAccessedTime and
thisAccessedTime by some fixed amount (60000) instead of waiting for
that time to pass.
That would speed up the test... but it sounds like adding a time machine
to the test class! My feeling is this would add inappropriate logical
complexity to a test that has always created and destroyed a tomcat
instance for each test case (there are now 15).
However, I intend to replicate most of the improvements from this test
class into the other authenticator tests, so I am already apprehensive
about adding too many 60+ second delays to the entire suite.
Mark and I briefly discussed adding a new protected method for use by
these timeout tests:
org.apache.catalina.session.StandardSession.setSessionTimeoutSecs(int secs)
What do you think?
b) Maybe call ManagerBase.setProcessExpiresFrequency(1) instead of the
default value (6) and reduce MANAGER_SCAN_DELAY_SECS from 60 to 10.
Mark and I discussed this idea earlier, but I was reluctant to increase
complexity in an already radical change to the original.
Your suggestion showed me how to achieve faster detection in a fairly
simple manner, so I have reopened the bug and provided a new patch that
shaves 50 seconds off the run time of the timeout test. Thanks!
Best regards,
Konstantin Kolinko
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
