On 20/03/2009, sebb <seb...@gmail.com> wrote: > On 20/03/2009, sebb <seb...@gmail.com> wrote: > > On 20/03/2009, Niall Pemberton <niall.pember...@gmail.com> wrote: > > > > > On Thu, Mar 19, 2009 at 3:06 PM, Stephen Colebourne > > > > > > <scolebou...@btopenworld.com> wrote: > > > > sebb wrote: > > > >> > > > >> On 19/03/2009, Stephen Colebourne <scolebou...@btopenworld.com> > wrote: > > > >>> > > > >>> So, overall, I'm dubious as to whether the value is sufficient to > > > >>> compilcate the compliation and to field the inevitable > > > >>> confusion/questions > > > >>> as to 'why we added a dependency' (when we didn't add one > really...) > > > >> > > > >> Again, I'm not sure I follow. > > > >> > > > >> I don't see how the addition of a single new dependency complicates > > > >> the compilation. > > > > > > > > Because [lang] has no dependencies at present. That is a feature. > > > > > > > >> Nor do I see why users will be confused, so long as the site shows > > > >> that LANG depends on Java 1.5 only. Many of them will just use > Maven > > > >> to pick up the new version. If necessary one can always add some > > > >> information on the site as to how annotations behave. > > > > > > > > But due to the way maven generates documentation, and the data in > the pom, > > > > it will /appear/ like [lang] does have a dependency. > > > > > > > > Since most users are unaware that annotation dependencies are not > needed at > > > > runtime, they will take the belt and braces approach and include the > > > > 'dependency'. Or stop using [lang]. > > > > > > > >> Indeed hopefully users will start adding annotations to their own > code... > > > > > > > > This change doesn't actually help with that, other than providing > > > > advertising for JCIP. > > > > > > > > I'm basically -0 to this change, as I think the confusion outweighs > the > > > > gains. > > > > > > > > > I agree with Stephen. > > > > > > As well as the point he makes its also causing the > > > net.jcip.annotations package to be included in the OSGi Import-Package > > > statement in the manifest which I assume will make this a required > > > dependency when using lang in an OSGi environment. I guess that the > > > maven-bundle-plugin can probably be configured to stop that happening > > > but even if it can then I don't really see the point of using this > > > over just plain comments in the javadocs. > > > > > > > > > How did you generate the OSGI stuff? > > > > I've been experimenting with HC, and I don't see the same behaviour. > > > > > Take 2: > > Just found a problem when using compile+optional in HC: > > The generated bundle: > org.apache.httpcomponents.httpcore_4.1-SNAPSHOT.jar > contains a valid DEPENDENCIES file. > The MANIFEST looks OK too, no mention of jcip > > However, the included archive: > httpcore-nio-4.1-SNAPSHOT.jar > has a DEPENDENCIES file which refers to JCIP. > > Not sure if this is a problem for OSGI or not, but it is not consistent. >
It gets messier - when I then build HC Client, the OSGI jar org.apache.httpcomponents.httpclient_4.0-beta3-SNAPSHOT.jar has a DEPENDENCIES file that is OK, but the MANIFEST now includes JCIP. The embedded jars httpclient-4.0-beta3-SNAPSHOT.jar httpmime-4.0-beta3-SNAPSHOT.jar have DEPENDENCIES files that include JCIP Looks like there is a problem with handling optional dependencies across projects. > > > Niall > > > > > > > > > > > > > Stephen > > > > > > --------------------------------------------------------------------- > > > To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org > > > For additional commands, e-mail: dev-h...@commons.apache.org > > > > > > > > > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
