On 8/23/10 2:20 PM, Carsten Ziegeler wrote:
> Justin Edelson wrote
>> I fixed the two failing test in JCR Installer. As mentioned last week,
>> one failure (FolderDetectionTest.testMoveLibsToFoo) was caused by a
>> Jackrabbit bug, for which I implemented a workaround.
> Great, thanks Justin!
np.
>
>>
>> The second failure (FolderDetectionTest.testMoveLibsToApps) was caused
>> by the fact that when session.move("/src", "/dest") is called, events
>> are sent only for NodeAdded /src and NodeRemoved /dest. The code in
>> RootFolderListener expected to get a NodeAdded event for each node in
>> the moved subtree.
>>
>> Looking at the JCR spec, Jackrabbit seems to be doing the right thing
>> here, so I ended up removing the path checking code in r988136. The net
>> result is that the watched subtrees (i.e. /apps and /libs) are scanned
>> whenever any change is done inside of them. I'm certainly open to other
>> alternatives for this.
> Sounds good to me. I've one question :) Do we need the dependency to JCR
> 2 for observation of the move event? As far as I can see this is
> currently the only place where jcr 2 api is used; otherwise the
> jcrinstaller would still work with jcr 1. I'm not against requiring JCR
> 2 for the installer but if it is just for this event, we might reduce
> the requirement to jcr 1
The NODE_MOVED event shouldn't be there. I'll remove it. I inserted it
as an attempt to fix the problem, but it didn't work and I forgot to
take it out.
We shouldn't need JCR 2.
To be clear, this test (FolderDetectionTest.testMoveLibsToApps) would
fail as it is currently written under JR 1 as well as JR 2. The only
reason it passed before was because the move resulted in a NodeAdded
event being fired for /apps, which caused /apps to be rescanned. The bug
was that moving a subtree to /apps/foo didn't cause /apps to be rescanned.
NodeAdded /apps - /apps get rescanned
NodeAdded /apps/foo/install - /apps/foo/install gets rescanned
NodeAdded /apps/foo - nothing happens
So, in other words, making the test work with Jackrabbit 2 exposed a bug
which had been there all along.
>
>>
>> In any case, all the tests are passing for me now. Created SLING-1681 to
>> merge it into the reactor and SLING-1682 to add jcr and osgi install to
>> launchpad.
> Great
>
> I've committed my last changes, so the stuff should be ready for a
> release (there are still bugs I would like to get fixed, but as those
> are there for a long time, we can release SLING-6 with them).
>
> And I finally fixed a bug in the IT tests (the
> ConfigInstallTest#testDeferredConfigInstall) - at least in my
> environment there are two config admin bundles installed; as only one is
> stopped, the config admin is still there, making the test fail.
> I changed this to simply stop all config admins :)
That's odd; never saw that one.
Justin
>
> Regards
> Carsten
