Great news! I'm both sad that the annotations caused you pain (the complete opposite was intended) and happy that you managed to work around the problem.
It's not by coincidence that BND uses the intermediate state of the manifest to go between the annotations and the module info. However, I wasn't sure that this would cover all cases particularly w.r.t. the service loader handling. However, it appears it does. Sincerely, Ray On Tue, Apr 6, 2021 at 10:39 AM Mark Thomas <ma...@apache.org> wrote: > On 02/04/2021 13:58, Raymond Augé wrote: > > I just wanted to make a note that removing the annotation will also mean > > that the module-info.java will need to be manually managed since bnd also > > generated that based on the annotations. > > Good news. I compared the module-info.class files for the EL API jar and > the WebSocket API jar between the 9.0.45 release (that used the > annotation) and a current build from source (that doesn't use the > annotation) and they are identical. On that basis I think we have a > working solution. > > Mark > > > > > > Just FYI, > > - Ray > > > > On Thu, Apr 1, 2021 at 4:40 AM Jean-Louis MONTEIRO <jeano...@gmail.com> > > wrote: > > > >> That's awesome news. > >> > >> I'm glad it's something that can be achieved without too much effort. > >> I understand and buy the pragmatic approach. > >> > >> But at the same time, if we can do a step forward and get even closer to > >> being certified, that'd be great. > >> > >> Le jeu. 1 avr. 2021 à 10:06, Mark Thomas <ma...@apache.org> a écrit : > >> > >>> I've been playing with this a bit more and it appears we can simply > >>> hard-code the "Require-Capability" header in el-api.jar.bnd > >>> > >>> Having taken the time to look at the actual values generated for these > >>> API JARs, this does look like something that would be simple to > maintain > >>> manually - especially for the API JARs where adding a new use of > >>> ServiceLoader is likely to happen fairly rarely. > >>> > >>> We should be able to remove the Bnd annotation in > >>> - 10.0.x for 10.0.6 onwards > >>> - 9.0.x for 9.0.46 onwards > >>> > >>> We'll certainly do this for the API JARs. I think we'll leave it in > >>> place for the implementation JARs > >>> > >>> I have a few things on my TODO list at the moment but I don't see why I > >>> shouldn't be able to get this done for the May set of releases. > >>> > >>> Mark > >>> > >>> > >>> On 01/04/2021 08:24, Romain Manni-Bucau wrote: > >>>> Hi, > >>>> > >>>> AFAIK TomEE will try to be certified and will try to not fork as much > >> as > >>>> possible Tomcat code so can be neat to solve it in particular when > >>>> relatively easy: > >>>> > >>>> 1. compile tomcat classes with bnd as of today > >>>> 2. run bnd to get the manifest.mf ( > >>>> 3. post process tomcat classes to remove bnd from the .class > >>>> 4. jar it > >>>> > >>>> should solve the process and does not look crazy compared to what > >> tomcat > >>>> already does, no? > >>>> > >>>> Alternative is indeed to drop bnd since the manifest is quite easy to > >>> write > >>>> manually or with an ant task to filter the version for tomcat case, at > >>>> least for spec jars (it is harder for the impl but here bnd is fine). > >>>> > >>>> Romain Manni-Bucau > >>>> @rmannibucau <https://twitter.com/rmannibucau> | Blog > >>>> <https://rmannibucau.metawerx.net/> | Old Blog > >>>> <http://rmannibucau.wordpress.com> | Github < > >>> https://github.com/rmannibucau> | > >>>> LinkedIn <https://www.linkedin.com/in/rmannibucau> | Book > >>>> < > >>> > >> > https://www.packtpub.com/application-development/java-ee-8-high-performance > >>>> > >>>> > >>>> > >>>> Le jeu. 1 avr. 2021 à 09:19, Mark Thomas <ma...@apache.org> a écrit : > >>>> > >>>>> On 31/03/2021 20:14, Volodymyr Siedlecki wrote: > >>>>>> Hello, > >>>>>> > >>>>>> It appears the TCK Signature tests will not be relaxed (see 1 & 2), > >>>>>> and I'm wondering how Tomcat will proceed with the bnd annotation in > >>>>>> the EL API? Will it be removed in the next release? > >>>>> > >>>>> Currently, there are no plans to change the Tomcat code. > >>>>> > >>>>> We do run the TCKs but take a pragmatic view of any failures. For > >>>>> example, the Servlet TCK test that tests setting a context path in > >>>>> web.xml always fails because Tomcat always overrides any such > setting. > >>>>> The Servlet spec allows this setting to be overridden but the TCK > test > >>>>> doesn't consider the possibility that a container will always do > this. > >>>>> Therefore we simply treat this as an expected failure. Full details > >> for > >>>>> all the TCKs we run can be found on the Wiki: > >>>>> > >>>>> https://cwiki.apache.org/confluence/display/TOMCAT/Jakarta+EE+TCKs > >>>>> > >>>>> The EL signature test failure is another example of a failure that we > >>>>> consider can be safely ignored. > >>>>> > >>>>> Tomcat does not, and has not for many, many years, formally certified > >>>>> against the Jakarta EE / Java EE TCKs. I am not aware of any user > >>>>> request at any point in that time to complete formal certification. > >>>>> Therefore, I expect that Tomcat will continue following its current > >>>>> pragmatic approach to the TCKs. > >>>>> > >>>>> Mark > >>>>> > >>>>> --------------------------------------------------------------------- > >>>>> To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org > >>>>> For additional commands, e-mail: dev-h...@tomcat.apache.org > >>>>> > >>>>> > >>>> > >>> > >>> > >>> --------------------------------------------------------------------- > >>> To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org > >>> For additional commands, e-mail: dev-h...@tomcat.apache.org > >>> > >>> > >> > >> -- > >> Jean-Louis > >> > > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org > For additional commands, e-mail: dev-h...@tomcat.apache.org > > -- *Raymond Augé* (@rotty3000) Senior Software Architect *Liferay, Inc.* (@Liferay) OSGi Fellow
