| Re: [jgit-dev] Is it actually possible to set up a new JGit developer environment these days? |
Well that's kind of my point, that setting up an environment from scratch is not a hermetic, reproducible process, and depends on whether you are lucky enough to have accumulated the right state in your IDE over the years to make things work.
>> * in case a dependency is missing go to Orbit repo (CVS) and check it out
> http://wiki.eclipse.org/EGit/Contributor_Guide#Option_2:_Install_from_Orbit_P2_Repository
> lists 8 projects you would need to do this for, which is I guess maybe ok, as long as we don't add new dependencies?
Maybe I'm just lucky, but currently I have no extra dependencies from
Orbit and see no compile errors in the workspace.
On Tue, Mar 25, 2014 at 8:30 PM, Dave Borowitz <dborowitz@xxxxxxxxxx> wrote:
>
>
>
> On Tue, Mar 25, 2014 at 12:25 PM, Tomasz Zarna <tzarna@xxxxxxxxx> wrote:
>>
>> Hmm, from my recent experience setting up a JGit dev workspace isn't that
>> hard:
>> * clone the repo
>> * import all the projects into eclipse
>
>
> * ignore all errors about specific Maven features not being supported (e.g.
> maven-antrun-plugin).
>
>> * activate jgit-4.4.target
>> * in case a dependency is missing go to Orbit repo (CVS) and check it out
>
>
> http://wiki.eclipse.org/EGit/Contributor_Guide#Option_2:_Install_from_Orbit_P2_Repository
> lists 8 projects you would need to do this for, which is I guess maybe ok,
> as long as we don't add new dependencies? If you wanted to actually get the
> versions JGit expects instead of whatever the latest R-build has, you would,
> what, inspect the .p2f file for the version it's expecting?
>
> I'm liking Roberto's suggestion more and more...
>
>>
>> * done!
>>
>> Running a single test (case) is as easy as creating a JUnit launch
>> config in Eclipse.
>
>
> Yeah, pretty the only reason I want to use an IDE is because I already knew
> this part was easy.
>
>>
>> fwiw, I'm on Luna M6
>>
>> Cheers,
>>
>> On Tue, Mar 25, 2014 at 7:17 PM, Dave Borowitz <dborowitz@xxxxxxxxxx>
>> wrote:
>> > The contributor guide is...disheartening.
>> >
>> > "Note: Eclipse Kepler currently has a bug (see bug 409073) which causes
>> > the
>> > instructions below to fail. You can manually install the tools (e.g.
>> > from
>> > the marketplace)"
>> >
>> > Really? No thank you. Ok, so maybe I just won't use Eclipse.
>> >
>> > First things first, alias mvn='JAVA_HOME=/path/to/jdk6 mvn', because
>> > some
>> > java7 bug in plexus-io or something makes maven-source-plugin take
>> > literally
>> > hours on my laptop.
>> >
>> > So mvn package at least works now, except that we have a lot of tests
>> > (which
>> > is good!) so it takes a long time.
>> >
>> > Now, can I run a specific test case? Like, the one corresponding to the
>> > bugfix I'm working on?
>> >
>> > Google tells me this would be 'mvn test -Dtest=MyTest', but no,
>> > maven-surefire-plugin says no tests are found. Something must be busted
>> > because -Dtest is supposed to support wildcards and 'mvn test -Dtest=*'
>> > gives me the same error. Perhaps 'mvn -pl org.eclipse.jgit.test test
>> > -Dttest', but then it expects all the other modules to be installed in
>> > my
>> > local repo. Again, no thank you.
>> >
>> > I have a theory that everybody who actively develops JGit is using an
>> > Eclipse project they set up >1 year ago after hours of manual labor, and
>> > we'd each be screwed if that project ever got lost. Which is pretty much
>> > what happened to me here :)
>> >
>> > https://git.eclipse.org/r/17955 can't get merged fast enough IMHO. (Wish
>> > I
>> > had more time to help.)
>> >
>> > _______________________________________________
>> > jgit-dev mailing list
>> > jgit-dev@xxxxxxxxxxx
>> > https://dev.eclipse.org/mailman/listinfo/jgit-dev
>> >
>
>