[
https://issues.jboss.org/browse/WFLY-5473?page=com.atlassian.jira.plugin....
]
Richard Janík updated WFLY-5473:
--------------------------------
Attachment: reproducer.zip
I've attached the reproducer. Unzip it into a clean installation of EAP/Wildfly and
run the reproducer.sh script.
I've been reproducing the issue manually before. I've been investigating some more
while writing the reproducer and found that there are cases where it works - invalidation
destroys SSO context - and cases where it doesn't work. It works only if you
invalidate the first session of the given SSO context. Here's an example (roughly the
same thing the reproducer does):
* Start two servers, each with <single-sign-on/>, a user and a non-distributable
deployment. Bind server1 to 127.0.0.1:8080 and server2 to 127.0.0.1:8180.
* Open browser (Firefox), GET 127.0.0.1:8080/deployment, authenticate - we get JSESSIONID
"x1" and JSESSIONIDSSO "y1"
* GET 127.0.0.1:8180/deployment, (no need to authenticate), we get JSESSIONID
"x2" and JSESSIONIDSSO "y1"
* GET 127.0.0.1:8080/deployment, (no need to authenticate), we get JSESSIONID
"x3" and JSESSIONIDSSO "y1"
* invalidate session on 127.0.0.1:8080 (JSESSIONID "x3"), then get
127.0.0.1:8080/deployment - we're not required to authenticate, but we should be
* when I spoofed the cookies in step 4 (before accessing 127.0.0.1:8080 and creating a
second session on that server) and invalidated session with JSESSIONID "x1", it
worked
Session.invalidate() does not invalidate SSO context for
non-distributable applications
---------------------------------------------------------------------------------------
Key: WFLY-5473
URL:
https://issues.jboss.org/browse/WFLY-5473
Project: WildFly
Issue Type: Bug
Components: Clustering, Web (Undertow)
Reporter: Richard Janík
Assignee: Paul Ferraro
Priority: Blocker
Fix For: 10.0.0.Final
Attachments: reproducer.zip
See "Steps to Reproduce" for detailed description.
According to my limited knowledge, this was also the core issue in
https://bugzilla.redhat.com/show_bug.cgi?id=924456 which has been dispatched as a one-off
to a customer. Thus, I'm setting the priority to blocker as this is a regression
against 6.4.x. No exceptions have been observed in the server output however.
Adding Clustering component as I've been trying this with standalone-ha.xml and BZ
924456 relates to clustering.
--
This message was sent by Atlassian JIRA
(v6.4.11#64026)