Trivial vfs-3.0.x manifest access fails

[thomas.diesler]

When trying to access the manifest from a jar, I get

Caused by: java.io.FileNotFoundException: /home/tdiesler/svn/jboss-osgi/trunk/testsuite/example/target/test-libs/example-simple.jar/META-INF/MANIFEST.MF (Not a directory)
    at java.io.FileInputStream.open(Native Method)
    at java.io.FileInputStream.<init>(FileInputStream.java:106)
    at org.jboss.vfs.spi.RealFileSystem.openInputStream(RealFileSystem.java:67)
    at org.jboss.vfs.VirtualFile.openStream(VirtualFile.java:225)
    at org.jboss.osgi.vfs30.VirtualFileAdaptor30.openStream(VirtualFileAdaptor30.java:106)
    at org.jboss.osgi.vfs.VFSUtils.getManifest(VFSUtils.java:49)
    at org.jboss.osgi.spi.util.BundleInfo.<init>(BundleInfo.java:109)

Essentially I do

VirtualFile root = VFS.getChild(url);
VirtualFile manifest = root.getChild(JarFile.MANIFEST_NAME);
InputStream stream = manifest.openStream();

In vfs21 I do

VirtualFile root = VFS.getRoot(url);
VirtualFile manifest = root.getChild(JarFile.MANIFEST_NAME);
InputStream stream = manifest.openStream();

which works.

[alesj]

VFS3 requires explicit mount of non-plain files; e.g. zip, gzip, ...

See AbstractVFSArchiveStructureDeployer in Deployers on how this is done, or VFS tests.

[thomas.diesler]

I added this

      MountOwner owner = new VirtualFileOwner(root);
      Automounter.mount(owner, root);

which seems to work.

Is this right?

Is there some cleanup needed?

Please document the correct usage of the API.

Good javadoc with code examples on how to use the API is the minimum.

Alternatively, you could provide a wiki or docbook.

[thomas.diesler]

No, I do not want to see some other bit of undocumented code.

Instead, I want to be directed to the jboss-vfs documentation.

Perhaps you like to confirm that this is indeed on the roadmap.

https://jira.jboss.org/jira/browse/JBVFS-113

[alesj]

I added this

      MountOwner owner = new VirtualFileOwner(root);
      Automounter.mount(owner, root);

which seems to work.

Is this right?

Is there some cleanup needed?

No need for MountOwner, Automounter::mount takes Object, and wraps it -- same as you did explicitly.

Yes, you need to preform explicit cleanup as well -- Automounter::mount returns Closeable cleanup handle.

[dmlloyd]

But, don't use automounter unless you have no other choice.  It's not intended as a consumable API; it's only there until we can work out a good way to get mounting into the deployment lifecycle/dependency graph.

In fact maybe I can convince John to deprecate it right off...

I think it'd be better if we have a deployed MC bean which represents the mount of the archive, and then everything using that mount will have a dependency on that.

In the standalone case (e.g. testing) you can just mount the archive and unmount it again in a finally block.

Note that the normal mount API is well-documented.

[asoldano]

[e-mail] ha scritto:

I think it'd be better if we have a deployed MC bean which represents the mount of the archive, and then everything using that mount will have a dependency on that.

+1