Comment 1 Markus Keller 2005-06-24 12:55:17 EDT

Created attachment 23953 [details]
possible fix

Comment 2 Markus Keller 2005-08-08 09:43:00 EDT

Created attachment 25830 [details]
Better fix

This is a better fix that also cancels a pending tree node expansion in tree
depth > 2. Patch released to HEAD.

Here's a class for manual testing. The first fix did not cancel the search when
expanding toString in this caller hierarchy:
take(Object)
+ run()
- run2()
  - toString()
      ...
  + hashCode()


public class CallHierarchyTest {
	public void take(Object o) { //start CH here
	}
	
	public void run() {
		take("A");
	}
	
	void run2() {
		take("B");
	}
	
	public String toString() {
		run();
		run2();
		return super.toString();
	}
	public int hashCode() {
		run();
		run2();
		return super.hashCode();
	}
}

Comment 3 Dirk Baeumer 2005-08-08 10:31:19 EDT

Markus, can you please step by an explain the patch to me. I don't see what the
advantage is of creating the wrapper via an extension point now. Code wise it
doesn't look too different.

Comment 4 Markus Keller 2005-08-10 12:14:30 EDT

Created attachment 25980 [details]
Fix without extension point for adapter

Comment 5 Dirk Baeumer 2005-08-11 06:25:38 EDT

Patch looks good for me. One little thing: shouldn't we assist in
MethodWrapperWorkbenchAdapter that the methodWrapper is not null.

Martin, can you please have a look as well ?

Comment 6 Markus Keller 2005-08-11 10:47:08 EDT

Yes, we could add an assert to check that the MethodWrapper is not null.
However, this is currently guaranteed by construction and we also didn't have it
in DeferredMethodWrapper. I'll add it when I release to HEAD.

Explanation for Martin: The MethodWrapperWorkbenchAdapter is necessary because
of the way the DeferredTreeContentManager determines the Jobs to cancel:
- We request cancellation of the tree root MethodWrapper
- The DeferredTreeContentManager asks all running jobs whether thy run on a
child of the element to cancel. To determine the parent of a tree element, it
tries to adapt each parent element to IWorkbenchAdapter and then asks the
adapter for its parent.

Comment 7 Martin Aeschlimann 2005-08-11 10:59:49 EDT

patch reviewed, ok to release.

Comment 8 Markus Keller 2005-08-11 12:13:44 EDT

Released to 3.1.1.
Released to HEAD with assert in MethodWrapperWorkbenchAdapter.

Comment 9 Dirk Baeumer 2005-09-02 11:43:17 EDT

start verifying...

Comment 10 Dirk Baeumer 2005-09-02 11:58:40 EDT

Verified that the first top level search any any search caused be expanding a
node can be canceled.

Verified on M20050831-1200 + plug-in export.