Fix ModelloCli broken after moving from Plexus to JSR330

[fridrich]

This is my attempt at least to make the ModelloCli.java at least start and run. It still throws an exception on simple models. But at least the guice bindings look correct and the thing is at least starting and running.

[fridrich]

#434

[fridrich]

This version is now working for me and from my side, it is ready to get integrated. I don't know which kind of test to write for a command-line tool so that it does not regress any more. Please, advise.
Btw, the guice path I tried before is largely convoluted, and complicated. After I studied the commit 3c05d7a, I understood the problem.

[fridrich]

The latest version actually restores the possibility to instantiate directly the Modello class, since I see now that some Fedora tools in javapackages-bootstrap do it. So this version partially reverts commit 8804652 concerning the two files and adds the configuration including the autowiring and index scanning. Like that, stuff out there that was using Modello() constructor directly, will be able to use it from this commit on again.

[fridrich]

It would be good to have some test for this. Minimum that it can generate simple java files from a model. But for that, one needs basically the whole classpath including the plugins.

[fridrich]

This is running in openSUSE since and generates on command-line the correct Java files. Any chance to have it reviewed?

[fridrich]

Where is this standing?

[fridrich]

ping?

[cstamas]

@fridrich do you have any needed change in here, or goal was really just to make it work (again)? As if latter, I'd like to have this PR superseded by #452 as it does more correct approach IMO (using shade plugin that is able to process Sisu index and Plexus XMLs).

[fridrich]

I would really love to have at least this in 90352d3
The reasons are 1) normally, we in distributions (SUSE for me) unshade the things anyway. 2) then javapackages-bootstrap in Fedora, as I saw call the new Modello() construtor. Also this change is in line with what was done in the tests when moving from plexus-containers to jsr330. After, I don't mind any other solution that works and that works also unshaded.
I don't care about the commit that adds the exception. But I would really love to have it working as it was before. Just adding cdi-api for something that can be done with existing dependencies...
Now, I am not the one who decides. I would still prefer this, since the tests use exactly the same.

[fridrich]

or if I missunderstood something, I am fine with whichever that makes this work. In SUSE we anyway use jpackage_script to create the classpath for the modello script.

[fridrich]

I am just wondering whether one cannot just simply apply that one commit and then maybe shade artifacts, which means that I can unshade it in the package without having to patch the logic that relied on shade plugin?

[cstamas]

Yes, will see about that. Am not at desk, so you can do it as well (pr against branchin this pr), will see tomorrow.

Or just wait until tomorrow and i will sort it out.

[fridrich]

Here are fedora tools that call the Modello constructor:
https://github.com/fedora-java/javapackages-bootstrap/blob/master/mbi/modello/src/org/fedoraproject/mbi/tool/modello/ModelloTool.java#L80

[fridrich]

Closing in favour of #452