Bug 75646 - Endless loop? in DefaultCommentMapper.storeLeadingComments
Summary: Endless loop? in DefaultCommentMapper.storeLeadingComments
Status: RESOLVED DUPLICATE of bug 75632
Alias: None
Product: JDT
Classification: Eclipse Project
Component: Core (show other bugs)
Version: 3.0   Edit
Hardware: PC Windows XP
: P3 critical (vote)
Target Milestone: 3.1 M3   Edit
Assignee: Frederic Fusier CLA
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2004-10-05 10:46 EDT by Michael Valenta CLA
Modified: 2004-10-06 15:51 EDT (History)
2 users (show)

See Also:


Attachments
VM dump (690.66 KB, text/plain)
2004-10-05 10:47 EDT, Michael Valenta CLA
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Michael Valenta CLA 2004-10-05 10:46:36 EDT
I am running on todays integration build and it is bordering on unusable due 
to 3-5 second delays that happen frequently. On taking a VM snapshot, one of 
the things I noticed was a large number of SelectionListenerWithASTManager 
jobs. Should there really be that many of these jobs? I have the entire 
Eclipse base a s source which may be a trigger. Marking as major due to the 
response times I am seeing (although this may or may not be the cause of 
that). I will attach the VM trace.
Comment 1 Michael Valenta CLA 2004-10-05 10:47:43 EDT
Created attachment 15006 [details]
VM dump
Comment 2 Michael Valenta CLA 2004-10-05 11:56:24 EDT
Changing to critical. The build is unusable for me.
Comment 3 Dani Megert CLA 2004-10-06 08:28:22 EDT
*** Bug 75651 has been marked as a duplicate of this bug. ***
Comment 4 Dani Megert CLA 2004-10-06 11:53:42 EDT
I just run into this as well (editing ASTProvider.java), the CPU stays at 100%
and the stack trace is always at the same place (while creating an AST). Endless
loop?
org/eclipse/jdt/core/dom/DefaultCommentMapper.storeLeadingComments(DefaultCommentMapper.java:321)



This bug uncovered some other minor issues that need to be improved:
- instead of n waiting SelectionListenerWithASTManager jobs of which n-1 will be
cancelled immediately when invoked only create one waiting job (filed bug 75710)
- if the SelectionListenerWithASTManager waits for an AST and the editor closes
we should not create an AST for that one but return null (filed bug 75711)
Comment 5 Michael Valenta CLA 2004-10-06 11:59:03 EDT
Is there some way to disable this background job? I am still trying to use 
yesterdays build but too many operations are adversly affected. For example, 
performing an Open Types (CTRL-SHIFT-T) shows a progress dialog for an 
extremely long time before opening the types dialog. Others don't seem to have 
the same issue so I'm wondering if they have some preference configuration 
that bypasses this problem.
Comment 6 Dani Megert CLA 2004-10-06 13:08:06 EDT
Nothing you can do: the AST gets created during reconcile and this is
synchronized with the compilation unit. Every operation that synchronizes on
that same CU has to wait.
Comment 7 Frederic Fusier CLA 2004-10-06 15:51:23 EDT

*** This bug has been marked as a duplicate of 75632 ***