[
https://issues.apache.org/jira/browse/WICKET-3258?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13453038#comment-13453038
]
Adam Lewandowski commented on WICKET-3258:
------------------------------------------
I also saw this problem with Websphere 8 (and 6 and 7), although in a slightly
different context: redirects to mounted bookmarkable pages. There are a couple
of odd/non-standard behaviors in WAS that cause this:
1) When the request is not handled by a registered servlet (as is usually the
case when using the Wicket filter), then WAS considers the request to be
handled by a default servlet with a servlet mapping of "" (empty string). It
considers the entire request path after the context root to be 'path info'. So
request.getServletPath() returns "" and getPathInfo() returns the full
context-relative path.This is exactly the opposite of what every other
container (including the reference implementation) does.
2) When calculating a relative URL for redirect, WAS is stripping the 'path
info' off of the request path and tacking on the relative portion. Because WAS
considers the entire request path to be 'path info', this results in the
redirect URL being relative to the context root, instead of the request URI. I
did discover an undocumented custom property
('com.ibm.ws.webcontainer.redirectwithpathinfo') that, when set to true, makes
it behave according to spec: “If the location is relative without a leading '/'
the container interprets it as relative to the current request URI”.
I didn't fully test with the custom property, but it did appear to be working
correctly in my limited testing. Instead we opted to use a different rendering
strategy (ONE_PASS_RENDER) to eliminate the problematic redirects in the first
place.
> Allow for non-relative redirects for buggy servlet containers
> -------------------------------------------------------------
>
> Key: WICKET-3258
> URL: https://issues.apache.org/jira/browse/WICKET-3258
> Project: Wicket
> Issue Type: Improvement
> Components: wicket
> Affects Versions: 1.4.14, 1.5-M3
> Environment: Websphere
> Reporter: Jeremy Thomerson
> Assignee: Jeremy Thomerson
> Priority: Minor
> Fix For: 1.4.18, 1.5.0
>
> Attachments: relativeurlproblem.tar.gz
>
>
> There appears to be a bug in the way WebSphere handles relative URLs. The
> servlet spec
> (http://jcp.org/aboutJava/communityprocess/final/jsr154/index.html), page 260
> specifies the behavior for relative URLs, which Wicket uses, but WebSphere
> appears to ignore. If we send a relative URL like "../../Something.html" to
> HttpServletResponse#sendRedirect, the relative path should be relative *to
> the requested URI* (the page I was requesting that resulted in the redirect).
> WebSphere isn't implementing this behavior correctly. Instead, it appears
> to interpret it as relative *to the app context and servlet mapping*.
>
> More details:
>
> We request a page like:
> http://[server]:[serverport]/[appcontext]/[servletmapping]/somefolder/someotherfolder/foo.html
>
> That page results in an error, and Wicket tries to redirect to an unmounted
> bookmarkable page by doing
> sendRedirect("../../?wicket:bookmarkablePage=:com.FooPage"). This is correct
> according to the servlet spec, because the url passed to sendRedirect should
> be relative *to the requested URI* (<--- this is the key). So, we expect the
> container to redirect to
> http://[server]:[serverport]/[appcontext]/[servletmapping]/?wicket:bookmarkable=com.FooPage
> . Jetty does (and Tomcat and others). But, WebSphere does not.
>
> Although I cannot look at the Websphere code, I suspect that it is
> interpreting this relative URL as being relative to the root of the servlet
> mapping. It is then removing two folders, which takes us back to the server
> root (instead of the servlet mapping, like it should), removing the app
> context and servlet mapping (i.e.
> http://[server]:[serverport]/?wicket:bookmarkable=com.FooPage) This is
> obviously incorrect.
>
> We have tried this compatibility setting (true and false), but it didn't
> work:
> https://publib.boulder.ibm.com/infocenter/wasinfo/v6r1/index.jsp?topic=/com.ibm.websphere.nd.doc/info/ae/ae/xrun_jvm_sendredirect.html
>
> There are a lot of reports about the problem:
> http://www.mail-archive.com/[email protected]/msg68864.html
> https://issues.apache.org/jira/browse/STR-1843?page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel
> http://www.mail-archive.com/[email protected]/msg06308.html
> Although we obviously can't fix the issue, we should provide a way for users
> of these servlet containers to redirect to an absolute URL rather than a
> relative URL.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
