Topic: mod_proxy_html tag mangling

[pshenkyr]

Hello,

Basics first, this installation is on windows 2000,with one server running apache 2.2.6 from apachelounge with the mod_proxy_html-3.0.0 from apachelounge as a reverse proxy to jboss 4.2.2. Everything installs and activates properly, the application works properly when accessing the jboss server directly. I'm using tcpmon from https://tcpmon.dev.java.net/ to capture datastreams.

The issue crops up when accessing the application through the reverse proxy with SetOutputFilter proxy-html active.

Configuration snippet:
<VirtualHost _default_:8080>
DocumentRoot "C:\valid\path\to\placeholder\docroot"
ServerName host1.domain.tld:8080
ErrorLog logs/xxxxxxxxx.log
LogLevel Debug

ProxyRequests off
<Proxy *>
Order deny,allow
Allow from all
</Proxy>
ProxyPass /fake http://host2.domain.tld:8080/real
ProxyPassReverse /fake http://host2.domain.tld:8080/real
SetOutputFilter proxy-html
ProxyHTMLLogVerbose On
ProxyHTMLDoctype HTML Legacy
ProxyHTMLBufSize 65536
ProxyHTMLURLMap /real/ /fake/
<Location /fake>
Order deny,allow
Allow from all
</Location>
CacheRoot C:\valid\path\to\cache
</VirtualHost>

The raw source:
<custom:link page="/viewSomethingInteresting.action?new" title="Search Something" prefix="<li>" postfix="">Something</custom:link>

The stream to the browser without proxy-html:
<a href="/real/viewSomethingInteresting.action?new" title="Search Something">Something</a>

The stream to the browser with proxy-html active:
<a href="/fake/viewSomethingInteresting.action?new" title="Search Something"></a>Something

So, if it is not obvious, the effect is to create an inactive link from a functioning one by swapping the sequence of Something and </a>.

I have tried several combinations of /real and /fake to no avail.
This happens to several different lines, the one included is only one instance.
I have tried seperating the lines and stringing them together to no avail.

I'll be glad to be called a n00b for some easy Doh! configuration issue but I can't see it. I believe if this is cleared up the reverse proxy will function properly with both static and dynamic content for this application.

My thanks in advance for any assistance.

[pshenkyr]

This was a ~5yo application migrated to newer deployment components. One of these was upgrading Struts from v1.0.1 to v1.2.9, with the application including some custom tag libraries. As a result the text sent into mod_proxy_html included an incorrect self closing tag so the example listed as (v1.0.1):
<a href="/real/viewSomethingInteresting.action?new" title="Search Something">Something</a>

became (v1.2.9):
<a href="/real/viewSomethingInteresting.action?new" title="Search Something"/>Something</a>

The cause was a bug in a custom tag which was ignored/handled by Struts v1.0.1 but not by v1.2.9. The invalid result was being corrected by quirks mode in browsers, but was confusing mod_proxy_html into generating the invalid links.

Once the custom tag was corrected, mod_proxy_html worked properly.

Thanks to anyone who spent brain cells on this.