Didn't say it was difficult ;) The point is that we can't revert the fact to use these annotations as previously explained so we must leave with them and ensure user can use its own api in mojos, this is why i suggested adding a classloader (like ClassRealm) doing the dispatching for these API.
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 lun. 14 oct. 2019 à 15:04, Tibor Digana <tibordig...@apache.org> a écrit : > Romain, The discussion is mainly about conflicts between Maven and user > plugins. > This can be fixed by either: > 1. removing javax.inject from Guice and/or using another container with > reduced dependencies, or > 2. ClassLoader level > > I wrote a container in my company and wrote a test framework using beans, > Proxy, DI and IoC, transaction boundaries. > Not that difficult at all. > > On Mon, Oct 14, 2019 at 1:56 PM Romain Manni-Bucau <rmannibu...@gmail.com> > wrote: > > > Did you investigate the isolation between IoC beans and mojos? the IoC > > beans can see the maven/lib/jsr-whatever.jar whereas mojos can't see it - > > they dont use it - but can inject the beans. > > Mojo have a dedicated API so only issue is for the IoC so I guess it > would > > make everyone happy to get the best of both worlds, no? > > > > 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 lun. 14 oct. 2019 à 13:42, Olivier Lamy <ol...@apache.org> a écrit : > > > > > +1 > > > I definitely agree with Stuart, This is a very simple API with a > minimum > > of > > > interfaces and used a lot. > > > I guess we already did such mistake with Maven2 and Plexus :) > > > So it's only a matter of not replay history. > > > Who really wants to have to maintain another already existing library > > > whereas we already need people for what we have now. > > > > > > On Mon, 14 Oct 2019 at 21:37, Stuart McCulloch <mccu...@gmail.com> > > wrote: > > > > > > > Sorry, but why would you want to write your own version of > javax.inject > > > > which is a very small static API supported by many injection systems, > > > both > > > > EE and non-EE > > > > > > > > It means you can't inject components written outside of Maven unless > > you > > > > write adapters or manually construct them. > > > > > > > > It also means that someone who wants to write an injectable component > > for > > > > Maven and re-use it elsewhere (such as with Spring) now needs to > write > > > two > > > > versions. > > > > > > > > You'll spend a lot of time and end up with duplicate APIs for doing > > > exactly > > > > the same thing, when you could have just bumped the JSR-250 version > and > > > > moved on. > > > > > > > > On Mon, 14 Oct 2019 at 12:22, Tibor Digana <tibordig...@apache.org> > > > wrote: > > > > > > > > > It would not be finally the same/identical list of annotations. > > > > > We do not have to copy everything, as for instance we do not have > to > > > copy > > > > > descriptive methods, names of annotations, packages. > > > > > So custom annotations means that: > > > > > + we have all responsibility in our hands > > > > > + we do not have to rely on dead Java EE annotations > > > > > + we can add new annottaions tailored to our Maven domain (not EE > > > domain) > > > > > > > > > > We can erase some annotations, e.g.: > > > > > 1. https://docs.oracle.com/javaee/6/api/javax/inject/Named.html - > > not > > > > sure > > > > > if we use expressions with @Named beans > > > > > instead maybe we meant something like @javax.annotation.ManagedBean > > but > > > > it > > > > > is only my guess because @Named is useless without using > expressions > > in > > > > the > > > > > code (not in POM). > > > > > 2. we can erase > > > > > https://docs.oracle.com/javaee/6/api/javax/inject/Scope.html > > > > > because we do not have e.g. HTTP Session, Conversation scope and > Html > > > > > forms, we do not have Request Sope because we have our own > lifecycle > > > with > > > > > phases. So again not very compatible with Maven domain. Usually we > > have > > > > > singletones and I have never seen scope of liveness of the bean > > > instances > > > > > in Maven. Maybe we would need to have a range of phases in the > future > > > > where > > > > > the bean would be valid (from compile to process-classes) even in > the > > > > > parallel Maven -T 2C, maybe. > > > > > > > > > > It's good that we have tried to adopt annotations from different > > domain > > > > and > > > > > IMHO the MNG-6084 is the experience that we should not repeat and > > > should > > > > > not again adopt foreign annotations (as here in non-EE container) > > from > > > > > non-Maven domains. > > > > > > > > > > > > > > > On Mon, Oct 14, 2019 at 12:43 PM Stuart McCulloch < > mccu...@gmail.com > > > > > > > > wrote: > > > > > > > > > > > There are already equivalents, adding yet another "standard" > that's > > > > > > specific to Maven is just like writing yet another logging API > > IMHO: > > > > > > > > > > > > javax.annotation.Priority > > > > org.eclipse.sisu.Priority > > > > > > > > > > > > javax.annotation.PostConstruct > > > > > > > > > org.codehaus.plexus.personality.plexus.lifecycle.phase.Initializable > > > > > > javax.annotation.PreDestroy > > > > > > > org.codehaus.plexus.personality.plexus.lifecycle.phase.Disposable > > > > > > > > > > > > (these last two are not totally equivalent since the Plexus > API > > > uses > > > > > > interfaces instead of annotations, but they support the same > goal) > > > > > > > > > > > > The container will work with or without JSR-250 on the classpath > so > > > > this > > > > > is > > > > > > more about what you want to let plugin component authors to use. > If > > > > > you're > > > > > > happy with them only using Maven specific annotations and having > to > > > > > > retro-fit or add adapters when they want to use components > outside > > of > > > > > Maven > > > > > > that use PostConstruct and PreDestroy then just roll back > MNG-6084, > > > > > making > > > > > > sure to add a release note warning any plugin authors depending > on > > > this > > > > > > feature that they will need to rewrite or adapt their plugins. > > > > > > > > > > > > Also note that this API is only exposed to plugins, it should > _not_ > > > be > > > > > > leaking onto the build classpath ... if it is then that's a bug. > So > > > > this > > > > > > situation is specific to when a plugin either actively uses a > > > > dependency > > > > > > that wants a later version of JSR-250 or wants to use that later > > > > version > > > > > > itself. > > > > > > > > > > > > Since the JSR-250 API doesn't change much I still think just > > bumping > > > > the > > > > > > version in the distro is the simplest and less disruptive option. > > > > > > > > > > > > On Mon, 14 Oct 2019 at 11:03, Tibor Digana < > tibordig...@apache.org > > > > > > > > wrote: > > > > > > > > > > > > > All these annotations are for Java EE containers and > application > > > > > servers. > > > > > > > A clear solution in 4.0 or 5.0 would be to develop our own > > > > annotations > > > > > > for > > > > > > > Maven Domain (not EE domain) within or Java package. > > > > > > > > > > > > > > This way we would reach: > > > > > > > + annotations looking similar to EE annotations > > > > > > > + isolation on the package level > > > > > > > + the purpose of the annotations would match specific Maven > > domain > > > > > > > + the JavaDoc would not be EE specific anymore > > > > > > > > > > > > > > Many coleagues have negative opinion to my proposal but this > was > > > only > > > > > > mine > > > > > > > view. > > > > > > > > > > > > > > On Mon, Oct 14, 2019 at 11:40 AM Stuart McCulloch < > > > mccu...@gmail.com > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > The JSR-250 API was exposed in > > > > > > > > https://issues.apache.org/jira/browse/MNG-6084 > > > > > > > > > > > > > > > > It provides the @PostConstruct, @PreDestroy, and @Priority > > > standard > > > > > > > > extension annotations for use by plugin components. > > > > > > > > > > > > > > > > Isolating that API would affect any plugin components that > rely > > > on > > > > > > > MNG-6084 > > > > > > > > - they would then need to revert back to Initializable and > > > > Disposable > > > > > > > from > > > > > > > > the Plexus API (although note that the benefit of using a > javax > > > API > > > > > is > > > > > > > that > > > > > > > > it makes components more reusable, especially when they come > > from > > > > > > outside > > > > > > > > the Maven ecosystem and just happen to be used in a Maven > > > plugin.) > > > > > > > > > > > > > > > > The best solution would be to either upgrade that dependency > - > > or > > > > > > > > alternatively just expose the @PostConstruct, @PreDestroy, > and > > > > > > @Priority > > > > > > > > annotation types rather than the whole package. > > > > > > > > > > > > > > > > On Mon, 14 Oct 2019 at 06:01, Romain Manni-Bucau < > > > > > > rmannibu...@gmail.com> > > > > > > > > wrote: > > > > > > > > > > > > > > > > > Guess it just comes from guice or friends and since they > dont > > > > > support > > > > > > > > > recent API it was not upgraded. > > > > > > > > > We can upgrade it or just isolate it IMO - but requires > some > > > > > careness > > > > > > > > since > > > > > > > > > custom components can use it so needs an evolution to > toggle > > it > > > > > > > probably. > > > > > > > > > > > > > > > > > > Le lun. 14 oct. 2019 à 03:43, Tibor Digana < > > > > tibordig...@apache.org > > > > > > > > > > > > a > > > > > > > > > écrit : > > > > > > > > > > > > > > > > > > > It is a similar issue to the problem with "javax.inject" > > API. > > > > > > > > > > What about relocating these two APIs to a package > > > > > > "org.apache.maven"? > > > > > > > > Any > > > > > > > > > > thoughts? > > > > > > > > > > > > > > > > > > > > On Mon, Oct 14, 2019 at 12:11 AM Petar Tahchiev < > > > > > > > paranoia...@gmail.com > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > > Hello guys, > > > > > > > > > > > > > > > > > > > > > > I just decided to upgrade my build from 3.5.0 to 3.6.2 > > last > > > > > night > > > > > > > > and I > > > > > > > > > > > stumbled across multiple problems, latest one being the > > > > > following > > > > > > > > > > > exception: > > > > > > > > > > > > > > > > > > > > > > Nested exception is java.lang.NoSuchMethodError: > > > > > > > > > > > javax.annotation.Resource.lookup()Ljava/lang/String; > > > > > > > > > > > > > > > > > > > > > > Seems like maven 3.6.2 comes with jsr-250-api-1.0.jar > in > > > the > > > > > lib/ > > > > > > > > > folder > > > > > > > > > > > and that seems to cause problems for me because I have > > > moneta > > > > > in > > > > > > my > > > > > > > > > > > classpath which brings newer version of jsr250: > > > > > > > > > > > > > > > > > > > > > > [INFO] --- maven-dependency-plugin:3.1.1:tree > > > (default-cli) @ > > > > > > > > drmartens > > > > > > > > > > --- > > > > > > > > > > > [INFO] com.demo:drmartens:war:1.0-SNAPSHOT > > > > > > > > > > > [INFO] \- > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > com.nemesis.platform:nemesis-platform-core:jar:2.1.67.BUILD-SNAPSHOT:compile > > > > > > > > > > > [INFO] \- org.javamoney:moneta:pom:1.3:compile > > > > > > > > > > > [INFO] \- > > > > > > > > javax.annotation:javax.annotation-api:jar:1.3.2:compile > > > > > > > > > > > [INFO] > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > ------------------------------------------------------------------------ > > > > > > > > > > > > > > > > > > > > > > So I was able to fix this following this > recommendation: > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > https://stackoverflow.com/questions/52607814/how-to-remove-a-maven-lib-jsr250-api-1-0-jar-from-the-plugin-classpath-jetty > > > > > > > > > > > using extensions, but I was wondering: why do we need > the > > > > > JSR250 > > > > > > in > > > > > > > > the > > > > > > > > > > lib > > > > > > > > > > > folder? Also why do we need such an old version? > > > > > > > > > > > -- > > > > > > > > > > > Regards, Petar! > > > > > > > > > > > Karlovo, Bulgaria. > > > > > > > > > > > --- > > > > > > > > > > > Public PGP Key at: > > > > > > > > > > > > > > > > > > > > > http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x19658550C3110611 > > > > > > > > > > > Key Fingerprint: A369 A7EE 61BC 93A3 CDFF 55A5 1965 > 8550 > > > > C311 > > > > > > 0611 > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > Olivier Lamy > > > http://twitter.com/olamy | http://linkedin.com/in/olamy > > > > > >
