Bhaskar Maddala wrote:
I frequently (seem to) run into issues that result in Class loading
related exceptions, doing a manual inspection of the manifest headers,
I often come to the conclusion that the class should have been in the
class space of the bundle that is causing the exception, at this state
it would be great to have a tool to visually inspect the classes that
are available in the bundle class space along with the jars from which
the classes are being loaded, this diagnostic information would, I
think, help in determining the error in the Manifest headers sooner
and lessen my pain attempting to understand OSGi.
Felix tries to give some of this diagnostic information when a CNFE
occurs...it will tell you if the package is available and who is
providing, for example.
Having understood the scope of the work that needs to be implemented I
think I have decided to settle for option 2
-> user does a search by class name and a bundle id
I intend to provide class meta information in the event that the class
is available in the bundle class space, else use
PackageAdmin.getBundle(java.lang.Class clazz) to retreive the bundle
and display its information.
From Karl's and Niclas responses above, my initial though was that
Niclas is correct, in that
I want to know where a particular class comes from for a specific
bundle class space. Based on what I understand, a class may be loaded
from 2 separate locations, in different class spaces, so I think I
need step 3.
1) Get the package name from the class name.
2) Locate all bundles exporting that package.
3) Check which Bundle Classloader equals the class classloader.
From Karl's response I took a look at the Javadoc for PackageAdmin
___
public Bundle getBundle(java.lang.Class clazz)
Returns the bundle from which the specified class is loaded. The class
loader of the returned bundle must have been used to load the
specified class. If the class was not loaded by a bundle class loader
then null is returned.
__
This seems to imply that my understanding of class being (potentially)
loaded from multiple bundles in different class spaces is incorrect,
what exactly am I missing?
No, your assumption is correct. Any given class can only be loaded by
one class loader (and thus one bundle), so the above API returns the
bundle that loaded the class instance you passed in. It is possible that
Foo.class was loaded from two different bundles, so in that case there
really are two different class instances, but you only pass in one class
instance to the above method, thus only one bundle will be returned --
the bundle that loaded that class. (Don't be confused when I say class
instance, I am not talking about instances of Foo, but rather instances
of Class.)
-> richard
My approach to try to implement this is as follows
Try and fix
https://issues.apache.org/jira/browse/FELIX-574
https://issues.apache.org/jira/browse/FELIX-566
to get an understanding of how the web console code is organized, I
took a first stab at 574 and know what to do to fix this. On 566, I
think I should be able to address it, once I have 574 addressed.
Only after then do I intend to try implementing the additional servlet
for performing the class and bundle id based search.
Once again thank you all for chiming in, greatly appreciated as it
makes my understand a little better each time.
Thanks
Bhaskar
On Fri, Jun 6, 2008 at 9:40 AM, Richard S. Hall <[EMAIL PROTECTED]> wrote:
Niclas Hedhman wrote:
On Friday 06 June 2008 01:47, Richard S. Hall wrote:
In some
cases, this won't be so easy, because you might have to look into the
contents of embedded JAR files...
Well, if he has the class (which I believe was the initial question), then
I think the process would be much easier.
1) Get the package name from the class name.
2) Locate all bundles exporting that package.
3) Check which Bundle Classloader equals the class' classloader.
No?
Yes, if he has the class. However, I thought his original question on users@
said that he wanted to be able to list all class available in a class space.
I could be remembering incorrectly. Nevertheless, you are correct, and as
Karl points out, if that is all he wants to do, it is really easy.
-> richard
Cheers
