Re: [m2e-dev] Eclipse Code Recommenders and m2e 1.5

[Igor Fedorenko <igor@xxxxxxxxxxxxxx>]

I don't have a solution, unfortunately, but let me know if you come up
with something workable and I'll see if I can accommodate it from m2e end.

--
Regards,
Igor

On 11/12/2013, 6:06, Andreas Sewe wrote:
> Hi,
>
> > - Require-Bundle org.eclipse.m2e.maven.runtime 1.4 as an immediate
> >    workaround
>
> FYI, this solution seems straight-forward but doesn't work as smoothly
> as one would hope. :-(
>
> The big problem is that Require-Bundle imports everything -- including
> stuff like javax.inject and com.google.inject which
> o.e.m2e.maven.runtime for some reason exports. And this causes a
> uses-conflict for us, as the bundles needing m2e classes also need to
> depend on our own o.e.r.injection bundle which re-exports a different
> javax.inject and com.google.inject.
>
> And wiring (through Import-Package) o.e.r.injection to
> org.eclipse.m2e.maven.runtime's javax.inject and com.google.inject
> doesn't work, since AFAIK we need our own (i.e., Orbit's)
> com.google.inject bundle since we attach two fragments to it:
> com.google.inject.assistedinject and com.google.inject.multibindings.
> And fragments require a *specific* host bundle; you cannot just attach
> them to o.e.m2e.maven.runtime as well.
>
> Previously, with us just importing the packages we were interested in
> (org.sonatype.aether, org.apache.maven.*) this wasn't a problem. We
> could still obtain our com.google.inject package from the bundle with
> the same name. But with Require-Bundle, we have to obtain it from
> o.e.m2e.maven.runtime.
>
> Now, maybe changing the order of bundles in Require-Bundle would work,
> although it seems a bit fragile. Either way, I'll give that a try.
>
> Best wishes,
>
> Andreas