[wildfly-dev] Taglibs - correct location of TLD files packaged in JARs

[Tomas Hofman]

Adding back the list...

So unless anyone posts any objections I will prepare a PR to that effect.

Thanks!

On 12/06/18 16:12, Remy Maucherat wrote:
> On Tue, Jun 12, 2018 at 4:06 PM, Tomas Hofman <thofman at redhat.com 
> <mailto:thofman at redhat.com>> wrote:
> 
>     Hello Remy,
> 
>     just to explain, Jean Clere mentioned you might have an opinion on this,
>     which is why I included you directly. Feel free to ignore this if you think
>     this is more for other people.
> 
> 
> Right now I think this is a grey area, and both ways sound reasonably 
> legitimate until there's a real clarification.
> 
> Rémy
> 
> 
>     Tomas
> 
> 
>     On 12/06/18 15:58, Tomas Hofman wrote:
> 
>         Hi folks,
> 
>         we have an issue [0] where a web app with JAR packaged taglib is
>         migrated from
>         Tomcat to Wildfly/EAP and the two servers are resolving tablib location
>         differently.
> 
>         The taglib file in the JAR is located in
>         META-INF/resources/WEB-INF/tlds/hi.tld.
> 
>         It's added to the taglib map explicitly via <taglib> element in
>         web-fragment.xml (in the same JAR):
> 
>                    <taglib>
>                        <taglib-uri>/HiTag</taglib-uri>
>                        <taglib-location>/WEB-INF/tlds/hi.tld</taglib-location>
>                    </taglib>
> 
>         This works in Tomcat, which takes *META-INF/resources/* as the root for
>         resolving <taglib-location>, and doesn't work in Wildfly, which takes
>         *META-INF/* as the root.
> 
>         The functionality is described by JSP SPEC section 7.3 (I don't want to
>         quote
>         any parts here, because I would have to quote all of it) [1]. After going
>         through it I think our interpretation is the correct one, though Apache
>         think
>         the contrary [2].
> 
>         However, would it hurt if we added META-INF/resources/ as another root
>         to allow
>         both options?
> 
> 
>         [0] The issue: https://issues.jboss.org/browse/JBEAP-14757
>         <https://issues.jboss.org/browse/JBEAP-14757>
>         [1] JSP spec:
>         http://download.oracle.com/otndocs/jcp/jsp-2_3-mrel2-eval-spec/
>         <http://download.oracle.com/otndocs/jcp/jsp-2_3-mrel2-eval-spec/>
>         [2] I wrote to Apache mailing list and Jeremy Boynes replied:
> 
>         """
>         Here’s how I read the spec.
> 
>         JSP 7.3.3 talks about a “taglib map” using elements in the deployment
>         descriptor per your example. Per JSP 7.3.2 the URI is mapped to a context
>         relative path interpreted relative to the root of the web application.
>         Resources contained in a web fragment jar are located in its
>         /META-INF/resources directory. Tomcat/Jasper are synthesizing the
>         effective web
>         application from the resources in the fragments, then resolving the URI
>         in the
>         taglib map against that merged application. I think that’s a correct
>         interpretation of the spec.
> 
>         JSP 7.3.1 refers to implicit map entries found by scanning the web
>         application’s jars from WEB-INF/lib.
>         """
> 
>         http://mail-archives.apache.org/mod_mbox/tomcat-taglibs-user/201805.mbox/%3C4E3C404C-1D9F-4292-BE77-C443C3C61671%40apache.org%3E
>         <http://mail-archives.apache.org/mod_mbox/tomcat-taglibs-user/201805.mbox/%3C4E3C404C-1D9F-4292-BE77-C443C3C61671%40apache.org%3E>
> 
> 
> 
>     -- 
>     Tomas Hofman
>     Software Engineer, JBoss SET
>     Red Hat
> 
> 

-- 
Tomas Hofman
Software Engineer, JBoss SET
Red Hat