Pushed an enhancement,

idea is to check if servlet mapping explicitely maps the servlet in
the jaxrs subcontext.

I'll let you review the fix (in CxfJAXRSFilter) and the few tests I
added in arquillian-tomee-jaxrs-test (AvoidConflict*Test) - on
develop.

If it is not ok for everybody we can end up having a flag to switch
back to real servlet.



Romain Manni-Bucau
@rmannibucau
http://www.tomitribe.com
http://rmannibucau.wordpress.com
https://github.com/rmannibucau


2015-01-26 11:58 GMT+01:00 Thiago Veronezi <[email protected]>:
> Hi guys,
>
> I won't have much time to check it further today, but I was able to update
> this project so you can use as example.
>
> https://github.com/tveronezi/rssreader
>
> It works on 1.7.1, but it crashes on 1.7.2.
>
> []s,
> Thiago.
>
>
>
>
> On Mon, Jan 26, 2015 at 5:52 AM, Jean-Louis Monteiro <
> [email protected]> wrote:
>
>> From your summary Romain, I'd go back to the servlet solution.
>> Seems more deterministic and easy.
>>
>> From a performance point of view, it's seems also much better isn't it?
>> From my understanding, even a simple JS or HTML static resource needs to go
>> through the filter.
>>
>> It seems hard to get the filter work without re-implementing part or the
>> servlet mapping and the JAX RS mapping.
>> I mean really work without any tricks.
>>
>> For example, I often get a WebApplicationException when I want a simple
>> resource or a simple servlet.
>> So from my point of view the filter seems to solve interesting situations
>> for JAX-RS but at the price of adding more complexity to Tomcat servlet
>> management as well as probably impacting performances.
>>
>> And when you say the impl is more complicated, I can't imagine someone else
>> maintaining this code and logic.
>>
>> conflict immediately with resources (html, css, js)
>> >
>> --> can you add more details please?
>>
>> Probably does not bring much to the discussion, but I'm use to say, the
>> simpler the better.
>> There is another point that matters (at least to me).
>>
>> Having a simple app on Tomcat works fine. Having the same app on WebProfile
>> flavor also works fine.
>> But having the same one on JAX RS or plus may or not fail and have less
>> throughput.
>>
>> For instance, not the same stack nor process between Tomcat, TomEE WP and
>> TomEE Plus is a stopper for me.
>>
>> Jean-Louis
>>
>> --
>> Jean-Louis Monteiro
>> http://twitter.com/jlouismonteiro
>> http://www.tomitribe.com
>>
>> On Mon, Jan 26, 2015 at 10:31 AM, Romain Manni-Bucau <
>> [email protected]>
>> wrote:
>>
>> > well basically we have 2 ways of implementing JAXRS "mapping"
>> > - as the result of a servlet (old implementation)
>> > - as a filter (current)
>> >
>> > Pro for servlet:
>> > - everybody does it -> NO surprises in its usage
>> > - simpler than what we have
>> >
>> > Cons for servlet:
>> > - conflict immediately with resources (html, css, js)
>> > - standard servlet rules applies so we can conflict with servlets as well
>> >
>> >
>> > Pro for filter:
>> > - we can solve it
>> > - more complicated impl
>> >
>> > Cons:
>> > - nobody does it (means we are not as close of a standard as we could)
>> > - we loose some servlet rules like /foo/* has a higher priority than
>> > /* for request /foo/bar
>> >
>> >
>> > Current implementation simply does this logic: if a servlet is mapped
>> > and is not the default one then use this servlet, else if a resource
>> > matches the request then return its content else use jaxrs. This allow
>> > JAXRS to not hide resources even when mapping is conflicting - default
>> > case whatever we do when a user doesn't specify an Application mapping
>> > for an @ApplicationPath.
>> >
>> > I know David wanted it to work OOTB, personally I don't care that much
>> > since I consider it is a bad practise to have conflicting endpoints -
>> > the case in this example, index should be mapped to a welcome-file and
>> > that's it IMO.
>> >
>> > Last point it is still deactivable through
>> > openejb.jaxrs.static-first=false in conf/system.properties and
>> > shouldn't happen if there is a JAXRS mapping.
>> >
>> >
>> > Hope I didn't forget too much points.
>> >
>> >
>> >
>> > Romain Manni-Bucau
>> > @rmannibucau
>> > http://www.tomitribe.com
>> > http://rmannibucau.wordpress.com
>> > https://github.com/rmannibucau
>> >
>> >
>> > 2015-01-26 10:18 GMT+01:00 Jonathan Gallimore <
>> > [email protected]>:
>> > > Hi,
>> > >
>> > > Interesting issue. I thought I'd create a new thread for this issue as
>> I
>> > > think it probably needs some discussion and agreement on how best to
>> fix,
>> > > and think that's best done on a discussion thread as opposed to the
>> vote
>> > > thread.
>> > >
>> > > I'll leave the vote open at least for a few more hours to see if there
>> is
>> > > any further feedback that comes in.
>> > >
>> > > If I have understood it correctly, the issue is that you have servlet
>> > > mappings like:
>> > >
>> > >    <servlet-mapping>
>> > >      <servlet-name>default</servlet-name>
>> > >      <url-pattern>/app/*</url-pattern>
>> > >    </servlet-mapping>
>> > >    <servlet>
>> > >      <servlet-name>index</servlet-name>
>> > >      <jsp-file>/index.jsp</jsp-file>
>> > >    </servlet>
>> > >    <servlet-mapping>
>> > >      <servlet-name>index</servlet-name>
>> > >      <url-pattern>/*</url-pattern>
>> > >    </servlet-mapping>
>> > >
>> > > and an EJB or POJO like:
>> > >
>> > > @Path("rest")
>> > > public class RestService {
>> > >
>> > >     @Path("users")
>> > >     public List<Users> getUsers() {
>> > >         ....
>> > >     }
>> > > }
>> > >
>> > > and for some reason the index.jsp is "trumping" the Rest endpoint for
>> > > /rest/users.
>> > >
>> > > Is that correct? Reverting out those 2 commits does give us other
>> issues
>> > so
>> > > it sounds like this may require a different fix potentially?
>> > >
>> > > Any thoughts?
>> > >
>> > > Jon
>> > >
>> > > On Mon, Jan 26, 2015 at 8:21 AM, Romain Manni-Bucau <
>> > [email protected]>
>> > > wrote:
>> > >
>> > >> Hello Thiago,
>> > >>
>> > >> if you map
>> > >>
>> > >>  <servlet>
>> > >>     <servlet-name>index</servlet-name>
>> > >>     <jsp-file>/index.jsp</jsp-file>
>> > >>   </servlet>
>> > >>   <servlet-mapping>
>> > >>     <servlet-name>index</servlet-name>
>> > >>     <url-pattern>/*</url-pattern>
>> > >>   </servlet-mapping>
>> > >>
>> > >> then you explicitely ask all your calls to be mapped to index servlet
>> > >> so it seems ok to me.
>> > >>
>> > >> Did I misunderstand you?
>> > >>
>> > >>
>> > >>
>> > >> Romain Manni-Bucau
>> > >> @rmannibucau
>> > >> http://www.tomitribe.com
>> > >> http://rmannibucau.wordpress.com
>> > >> https://github.com/rmannibucau
>> > >>
>> > >>
>> > >> 2015-01-26 3:00 GMT+01:00 Thiago Veronezi <[email protected]>:
>> > >> > -1 -> The default servlet is catching the RS requests.
>> > >> >
>> > >> > Basically, in single-page-applications, I need to map the default
>> > servlet
>> > >> > in a special way...
>> > >> >
>> > >> > .
>> > >> > .
>> > >> > .
>> > >> >   <!-- The trick is to put all your static files under the same
>> > directory
>> > >> > and map the "default" servlet to it -->
>> > >> >   <servlet-mapping>
>> > >> >     <servlet-name>default</servlet-name>
>> > >> >     <url-pattern>/app/*</url-pattern>
>> > >> >   </servlet-mapping>
>> > >> >
>> > >> >   <!-- Any other request will point to the "index.jsp" page. This
>> way
>> > YUI
>> > >> > will be able to manage page transitions
>> > >> >         at the client side in case the user starts the application
>> > from a
>> > >> > permalink. -->
>> > >> >   <servlet>
>> > >> >     <servlet-name>index</servlet-name>
>> > >> >     <jsp-file>/index.jsp</jsp-file>
>> > >> >   </servlet>
>> > >> >   <servlet-mapping>
>> > >> >     <servlet-name>index</servlet-name>
>> > >> >     <url-pattern>/*</url-pattern>
>> > >> >   </servlet-mapping>
>> > >> > .
>> > >> > .
>> > >> > .
>> > >> >
>> > >> > TomEE 1.7.2 returns "index.jsp" for all my rest calls (GET
>> > >> > http://localhost:8080/myapp/rest/users, for example).
>> > >> >
>> > >> > After reverting back these two commits, the system gets back to
>> > normal.
>> > >> >
>> > >>
>> >
>> https://git1-us-west.apache.org/repos/asf?p=tomee.git;a=commit;h=544806da419bc2f5ab8bc936a989ff99bc9d891b
>> > >> >
>> > >>
>> >
>> https://git1-us-west.apache.org/repos/asf?p=tomee.git;a=commit;h=cb135dd6344f93ed24888cafcc05d0cfbb0c62a9
>> > >> >
>> > >> > @Romain,
>> > >> > Can you take a look? I'm not sure what was the idea.
>> > >> >
>> > >> > []s,
>> > >> > Thiago.
>> > >> >
>> > >> >
>> > >> > On Sun, Jan 25, 2015 at 9:36 AM, Jonathan Gallimore <
>> > >> > [email protected]> wrote:
>> > >> >
>> > >> >> Hi,
>> > >> >>
>> > >> >> Sorry, should have added that. It's:
>> > >> >>
>> > https://repository.apache.org/content/repositories/orgapachetomee-1045
>> > >> >>
>> > >> >> Regards
>> > >> >>
>> > >> >> Jon
>> > >> >>
>> > >> >>
>> > >> >>
>> > >> >> > On 25 Jan 2015, at 14:19, jieryn <[email protected]> wrote:
>> > >> >> >
>> > >> >> > On Sat, Jan 24, 2015 at 4:20 PM, Jonathan Gallimore
>> > >> >> > <[email protected]> wrote:
>> > >> >> >> Maven Repo:
>> > >> https://dist.apache.org/repos/dist/dev/tomee/staging-1045/
>> > >> >> >
>> > >> >> > Where is the actual Maven artifact staging repository for this?
>> So
>> > my
>> > >> >> > projects which depend on TomEE can be built and tested against
>> the
>> > >> >> > candidate. Thanks!
>> > >> >>
>> > >>
>> >
>>

Reply via email to