Details
-
Type:
Bug
-
Status:
Closed
-
Priority:
Major
-
Resolution: Fixed
-
Affects Version/s: None
-
Fix Version/s: 0.2
-
Component/s: None
-
Labels:None
-
Number of attachments :
Description
DelegatingInvoker raises a false Cyclic dependency exception if after an execption occured in the delegate.
A trivial fix is to wrap the invokation of the delegate with a try-finally block resetting the "executing" flag.
Activity
This fixes a two-fold issue with DelegatingInvoker.
1. Adds a try/finally to make sure the executing flag is switched off no matter what happens during method invocation.
2. changed the excuting flag to a Set that locks execution on a per-method basis instead of per-instance basis.
Scott T Weaver
added a comment - This fixes a two-fold issue with DelegatingInvoker.
1. Adds a try/finally to make sure the executing flag is switched off no matter what happens during method invocation.
2. changed the excuting flag to a Set that locks execution on a per-method basis instead of per-instance basis. Scott T Weaver
made changes -
| Field | Original Value | New Value |
|---|---|---|
| Attachment | DelegatingInvoker.patch [ 12472 ] |
Modification to DelegatingTest to test changes
Scott T Weaver
added a comment - Modification to DelegatingTest to test changes Scott T Weaver
made changes -
| Attachment | DelegatingTest.patch [ 12473 ] |
Hi Scott.
Can you attach the patched files please? I can't get these patches to work (I get a runtime error under win32 and an error that patch #3 doesn't match under linux).
Thanks,
Dan
I am sure that the patches I posted way back in July are far to old to apply correctly to the current CVS head. I am not sure I still have the proxytoys code box anymore. I will see if there is a way I can easily regen the patches but I can't promise anything.
Scott T Weaver
added a comment - I am sure that the patches I posted way back in July are far to old to apply correctly to the current CVS head. I am not sure I still have the proxytoys code box anymore. I will see if there is a way I can easily regen the patches but I can't promise anything. Fixed this now. But why do we check for cycles at all? If I have a method that is calling itself ... fine, e.g. I could calculate a Fibonacci number. If we remove the check, we can avoid the expensive ThreadLocal and set altogether. WDYT?
Jörg Schaible
added a comment - Fixed this now. But why do we check for cycles at all? If I have a method that is calling itself ... fine, e.g. I could calculate a Fibonacci number. If we remove the check, we can avoid the expensive ThreadLocal and set altogether. WDYT? Finally fixed with PTOYS-8. Cannot close this issue though due to JIRA rights.
Jörg Schaible
added a comment - Finally fixed with PTOYS-8 . Cannot close this issue though due to JIRA rights. Jörg Schaible
made changes -
| Fix Version/s | 0.2 [ 11692 ] | |
| Resolution | Fixed [ 1 ] | |
| Assignee | Joerg Schaible [ joehni ] | |
| Status | Open [ 1 ] | Closed [ 6 ] |
Can you provide also a unit test? There might be more to consider, than just eliminating the check.