[TAPESTRY-132] Reusing pooled engines may lead to wrong generated URLs

• Type: Bug
• Status: Resolved
• Resolution: Incomplete
• Affects Version/s: 3.0
• Component/s: Framework
• Environment: Operating System: Other; Platform: Other
• Bugzilla Id: 27621
• Assignee: Unassigned
• Reporter: Michael Frericks
• Created: 12/Mar/04 16:15
• Updated: 16/Feb/11 22:30
• Resolved: 27/Aug/04 06:01

Description

1. Error description:

Reusing pooled engines may lead to wrong generated URLs.

The following servlet-mapping

<servlet-mapping>
<servlet-name>MyApplicationServlet</servlet-name>
<url-pattern>*.place</url-pattern>
</servlet-mapping>

can be used to provide different information in the servlet-path
of different users.

The AbstractEngine stores this servlet-path in _servletPath
and the engine ends up in the ApplicationServlet._enginePool
for reuse under certain conditions.

ApplicationServlet#doService:
...
// No session; the engine contains no state particular to
// the client (except for locale). Don't throw it away,
// instead save it in a pool for later reuse (by this, or another
// client in the same locale).

if (LOG.isDebugEnabled())
LOG.debug("Returning " + engine + " to pool.");

_enginePool.store(engine.getLocale(), engine);

If this engine is reused by another client, the _servletPath
will not be updated to the one of the new client, since

AbstractEngine#setupForRequest:
...
if (_servletPath == null)

{ ... }

This my lead to wrong URLs since in e.g EngineServiceLink
the servletPath of the engine is used.

EngineServiceLink#constructURL:

buffer.append(_cycle.getEngine().getServletPath());

Conclusion

a) Reset the _servletPath, when sending thee engine to the pool
b) Do not pool engines anyway