Re: [JGIT RFC] JGit mavenization done right

From: Shawn O. Pearce <spearce@spearce.org>
Date: Aug 19, 2008, 14:44 UTC
Message-ID: <20080819144424.GC20947@spearce.org>
In-Reply-To: <7bfdc29a0808172333g1cbe2102n6e76ae98a1943411@mail.gmail.com>

Imran M Yousuf <imyousuf@gmail.com> wrote:

Show 9 quoted lines

> On Mon, Aug 18, 2008 at 11:55 AM, Shawn O. Pearce <spearce@spearce.org> wrote:
> > Imran M Yousuf <imyousuf@gmail.com> wrote:
> >> I would like to request you all to test out JGit from
> >> http://repo.or.cz/w/egit/imyousuf.git. Please checkout the branch
> >> 'unified_tst_rsrc' and try to build it with both maven and Eclipse
> >> (i.e. as was built earlier)
> 
> Thanks, it would nice to know whether it works in the original build
> process or not :).

Well, it did break it in Eclipse:

$ git diff-tree --abbrev -r -M --diff-filter=D orcz-pub/master HEAD :100644 000000 9d7d138... 0000000... D org.spearce.jgit.test/.gitignore :100644 000000 987d6be... 0000000... D org.spearce.jgit.test/.project :100644 000000 8bfa5f1... 0000000... D org.spearce.jgit.test/.settings/org.eclipse.jdt.core.prefs :100644 000000 fce94cf... 0000000... D org.spearce.jgit.test/.settings/org.eclipse.jdt.ui.prefs :100644 000000 304e861... 0000000... D org.spearce.jgit/.classpath :100644 000000 ba077a4... 0000000... D org.spearce.jgit/.gitignore :100644 000000 7d38455... 0000000... D org.spearce.jgit/.project :100644 000000 709a440... 0000000... D org.spearce.jgit/.settings/org.eclipse.jdt.ui.prefs

Removing this stuff was not so good. The Eclipse projects are busted and can't be used anymore. We need them back.

The make_jgit.sh however seems to produce a valid JAR. Given the file-level differences I didn't expect it to fail.

Also, I wonder if JGitTestUtil is better handled by placing the method in RepositoryTestCase and making sure everyone subclasses that if they need a test resource file? I'm fairly certain they already do, and its a lot easier to invoke a method you inherited than one in another class. (Well, easier for the guy writing the test case anyway, Java obviously doesn't care either way.)

If we are going to take this in upstream I'd like a flattened/cleaned up history. Being able to bisect the misstep of using symlinks (the old Maven approach) isn't very valuable in the long-term view of the history.

Robin, any thoughts?

-- 
Shawn.