It ended up being a bug in Jetty. For form based auth, jetty doesn't
call user.authenticate() (except on the first login post) and so it
doesn't associat a subject with activeSubject in the security manager.
So when it tries to authorize the user in the role, the subject is null
and so it fails.

Thanks for your patience. I wonder if there could be some warning spit
out when the subject is null. I don't know if there are any legitimate
situations where that should happen, but if there aren't, it might be
good to log a warning so this type of thing would be easier to detect.

Thank you,

-- 
Joel Shellman
iKestrel, Inc.  http://www.ikestrel.com/

On Thu, 2001-10-18 at 01:07, Scott M Stark wrote:
> There are many testcases for security in the jbosstest unit tests and 2.4.3
> works fine. Look at those and if you think you really have a bug post
> an example to sourceforge.
> 
> xxxxxxxxxxxxxxxxxxxxxxxx
> Scott Stark

> > At least I'm pretty sure. I'm doing source level debugging and
> > activeSubject (the thread local variable) never gets set during a
> > request.
> > 
> > Well, this results in the role never being authorized because line 195
> > always returns null (the thread local always returns null because it's
> > never set for this thread). Thus on line 263, subject is null, and so
> > doesUserHaveRole() returns false.
> > 
> > I put in break points at the only two points that I saw that every set
> > activeSubject, and neither one was hit during a request.
> > 
> > Can anyone help me get a work around or fix for this?



_______________________________________________
Jboss-development mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-development

Reply via email to