Community
Participate
Working Groups
To reproduce: # copy the attached stack trace to the clipboard # open any Bugzilla task in the Mylyn task editor (markup enabled) # paste into the new comment part # observe that Eclipse hangs and never comes back The stack trace of the hung thread is also attached. Might be related to bug 264345.
Created attachment 141700 [details] trace of the hung thread, from the debug view
Created attachment 141701 [details] stack trace that, when pasted, causes the hang
The line of text being processed at JavaStackTraceBlock.java:50 is 'org.eclipse.ui.internal.PerspectiveBarContributionItem.select(PerspectiveBarContributionItem.java:124)'
Created attachment 141702 [details] JUnit test case that reproduces the problem.
Created attachment 141703 [details] mylyn/context/zip
Created attachment 141726 [details] patch that fixes the issue
Created attachment 141727 [details] mylyn/context/zip
Fixed for 3.2.1
I was going through my queries this morning, when my workspace locked up trying to view a bug. Running 'kill -3' produced a stack for thread-0 that looks a lot like the one in the first attachment. It appears that I can reliably reproduce the hang trying to open bug 284002
I should note that I'm up to do with the latest sources.
*up to date* with the latest sources, sorry.
I can confirm that this is still happening.
Created attachment 142060 [details] patch that disables parsing of stack traces Since the effects of this bug are severe I would like to get this fixed for the 3.2.1 release. Unfortunately this bug report does not have any information about the cause of the problem which had already been partially fixed. Therefore I am not able to debug and fix this properly in time for the release but have attached a patch to disable parsing of stack traces as a stopgap solution. David, please let me know if you have time to work on a fix until tomorrow. I will commit my patch otherwise.
I have committed the patch which disables decoration of stack traces and avoids the busy loop. Leo, can you verify?
Yes, that relieves the blocker status for me.
Created attachment 142255 [details] patch that disables test cases
Created attachment 144325 [details] patch that fixes the hang and enables JUnit tests I've got a candidate fix for this issue that causes WikiText to detect stack traces without the dreaded hang. Anyone that's interested should try the patch and open bugs containing stack traces to determine if they are correctly detected and if the UI remains responsive.
Created attachment 144326 [details] mylyn/context/zip
I have applied the patch to my workspace and will see how it goes. Shawn, Rob, Leo, can you do the same? David, do you have any better clue to why this is happening at all?
(In reply to comment #19) > David, do you have any better clue to why this is happening at all? Looks to me like the problem part of the regex is as follows: @([A-Za-z][a-z0-9_$]*)+@ Notice how two very similar repeating segments are nested? Though it _shouldn't_ loop infinitely, you can see why it may involve many, many computations (exponentially more expensive, depending on the string it's trying to match). The patch changes the problem part as follows: @[A-Za-z][a-zA-Z0-9_$]*@ So while I'd like to blame the regex runtime, I can't really do it with a straight face... looks to me like a lesson in how to write a regex. I may have to blog about this one...
Ack... should have previewed the markup before displaying it. Here are the two regex segments: old: ==<code>([A-Za-z][a-z0-9_$]*)+</code>== new: ==<code>[A-Za-z][a-zA-Z0-9_$]*</code>==
patch applied to head reassigning to Steffen to apply the patch to the branch
(In reply to comment #19) > I have applied the patch to my workspace and will see how it goes. Shawn, Rob, > Leo, can you do the same? > > David, do you have any better clue to why this is happening at all? This patch fixes the problem I was running into. Great!
Patch applied to m_3_2_x branch. I'll trigger a new weekly build.
(In reply to comment #23) > This patch fixes the problem I was running into. Great! (In reply to comment #24) > Patch applied to m_3_2_x branch. I'll trigger a new weekly build. resolved as per above comments
*** Bug 284210 has been marked as a duplicate of this bug. ***
*** Bug 284154 has been marked as a duplicate of this bug. ***
