GRECLIPSE

out of space in CodeCache for adapters

Details

  • Type: Bug Bug
  • Status: Resolved Resolved
  • Priority: Major Major
  • Resolution: Cannot Reproduce
  • Affects Version/s: 2.5.0Release
  • Fix Version/s: 2.5.2.Release
  • Component/s: Editor
  • Labels:
    None
  • Environment:
  • Number of attachments :
    3

Description

I had this issue occur again. The issue occurred immediately after adding a groovy source folder (using the Eclipse build path editor) and then opening a Groovy class within that class. I was not able to get a lot of information because Eclipse shutdown (I believe I accidentally selected the button that is shown with this error to restart eclipse). After Eclipse is restarted everything works again as expected.

My environment has been changed quite a bit. I have attached a new file with the environment. A few highlights are:

It really seems like this issue occurs when a lot of new/updated Groovy code is introduced. Previously I think this may have been related to when I switched branches as mentioned in comment-271504. One other thing that may be relevant is that I use gmaven and m2e integration. I really do not know if that information helps, but thought I would make that as evident as possible.

Thanks again for your help and sorry for not being able to provide more details on this issue.

  1. Text File
    greclipe-stacks.txt
    19/Aug/11 1:01 PM
    10 kB
    Andrew Eisenberg
  2. Text File
    GRECLIPSE-1074-updatedeclipseerror.txt
    19/Aug/11 1:01 PM
    6 kB
    Andrew Eisenberg
  3. Text File
    GRECLIPSE-1074-updateenvironment.txt
    19/Aug/11 1:01 PM
    115 kB
    Andrew Eisenberg

Issue Links

relates to

Bug - A problem which impairs or prevents the functions of the product. GRECLIPSE-1074 Opening Groovy File causes VerifyError: class: ArrayUtil, method: <clinit> signature: ()V) Illegal type in constant pool

  • Major - Major loss of function.
  • Resolved - A resolution has been taken, and it is awaiting verification by reporter. From here issues are either reopened, or are closed.

Activity

Ascending order - Click to sort in descending order
Hide
Permalink
Andrew Eisenberg added a comment - - edited

Cloned from GRECLIPSE-1074.

Unfortunately, it doesn't look like comments were copied over.

Show
Andrew Eisenberg added a comment - - edited Cloned from GRECLIPSE-1074 . Unfortunately, it doesn't look like comments were copied over.
Hide
Permalink
Andrew Eisenberg added a comment -

It's not out of the question that gmaven could introduce code cache problems since running with m2e would mean that gmaven is partially executed in the same jvm as the workbench. I have no way of knowing this unless I can get a memory snapshot.

Just make sure that you are using at least the following settings in your STS.ini:

-Xmx1024m 
-Xms256m 
-XX:PermSize=64m 
-XX:MaxPermSize=256m 
-XX:ReservedCodeCacheSize=64m

A yourkit snapshot of the system after you see this problem again would be helpful. Although that being said, the last snapshot you sent me didn't show me anything too useful.

Show
Andrew Eisenberg added a comment - It's not out of the question that gmaven could introduce code cache problems since running with m2e would mean that gmaven is partially executed in the same jvm as the workbench. I have no way of knowing this unless I can get a memory snapshot. Just make sure that you are using at least the following settings in your STS.ini: -Xmx1024m -Xms256m -XX:PermSize=64m -XX:MaxPermSize=256m -XX:ReservedCodeCacheSize=64m A yourkit snapshot of the system after you see this problem again would be helpful. Although that being said, the last snapshot you sent me didn't show me anything too useful.
Hide
Permalink
Andrew Eisenberg added a comment -

I am searching for any mention of a leak in CodeCache in groovy, but can't find anything. It does seem like someone else had a vaguely similar problem:

http://stackoverflow.com/questions/6015873/eclipse-dies-every-few-minutes-with-latest-jdk-saying-out-of-space-in-codecache

and here:

http://www.apacheserver.net/trunk-build-problem-out-of-space-in-CodeCache-for-adapters-at1027069.htm

The solution in the second link was to set the MAVEN_OPTS to be this:

-Xmx1024m -Xms256m -XX:PermSize=64m -XX:MaxPermSize=256m -XX:ReservedCodeCacheSize=64m

I am not sure how this would affect m2e since m2e is running inside of Eclipse/STS and would use the memory of Eclipse's JVM. But it is perhaps worth a try.

Show
Andrew Eisenberg added a comment - I am searching for any mention of a leak in CodeCache in groovy, but can't find anything. It does seem like someone else had a vaguely similar problem: http://stackoverflow.com/questions/6015873/eclipse-dies-every-few-minutes-with-latest-jdk-saying-out-of-space-in-codecache and here: http://www.apacheserver.net/trunk-build-problem-out-of-space-in-CodeCache-for-adapters-at1027069.htm The solution in the second link was to set the MAVEN_OPTS to be this: -Xmx1024m -Xms256m -XX:PermSize=64m -XX:MaxPermSize=256m -XX:ReservedCodeCacheSize=64m I am not sure how this would affect m2e since m2e is running inside of Eclipse/STS and would use the memory of Eclipse's JVM. But it is perhaps worth a try.
Hide
Permalink
Rob Winch added a comment -

Thanks for the fast response on this. I have updated my .ini settings per your recommendation.

I'm not really particular about where this issue is tracked, but out of curiosity what makes you think the original issue was fixed? I realize that this error differs from the initial posting, but it aligns with the first error I got when opening the editor (see the greclipse-stacks.txt and this comment). Perhaps you noticed something I didn't or it has gotten confusing since I did not give all the information up front (sorry about that).

I will try and get another yourkit snapshot next time. I know this is a poor excuse, but I believe I am mistakenly restarting Eclipse before I realize what has happened (this is an unfortunate side effect of using a lot of key bindings, the error dialog having a restart option on it, and my fingers thinking they are smarter than my brain).

A little more information on the gmaven is that I had some difficulty getting the m2e 1.0 groovy integration to work. I kept getting a maven error "Plugin execution not covered by lifecycle configuration". I resolved this by adding the lifecycle-mapping plugin set to execute. The problems I was having are probably more appropriate for GRECLIPSE-1068, but I thought it would be beneficial to know I was using the lifecycle-mapping since gmaven may have a part in this and it impacts how things work behind the scenes. I'm really beginning to think this issue is occurring because gmaven is being invoked by Eclipse. After all, these sort of issues were the reason that the m2e team decided not to directly support running maven plugins in Eclipse.

Thanks again for all your help!

PS: I do plan on trying to give feedback on the new m2e integration with the troubles I have had but at the moment I need to focus on "real work" as much as possible.

Show
Rob Winch added a comment - Thanks for the fast response on this. I have updated my .ini settings per your recommendation. I'm not really particular about where this issue is tracked, but out of curiosity what makes you think the original issue was fixed? I realize that this error differs from the initial posting, but it aligns with the first error I got when opening the editor (see the greclipse-stacks.txt and this comment ). Perhaps you noticed something I didn't or it has gotten confusing since I did not give all the information up front (sorry about that). I will try and get another yourkit snapshot next time. I know this is a poor excuse, but I believe I am mistakenly restarting Eclipse before I realize what has happened (this is an unfortunate side effect of using a lot of key bindings, the error dialog having a restart option on it, and my fingers thinking they are smarter than my brain). A little more information on the gmaven is that I had some difficulty getting the m2e 1.0 groovy integration to work. I kept getting a maven error "Plugin execution not covered by lifecycle configuration". I resolved this by adding the lifecycle-mapping plugin set to execute. The problems I was having are probably more appropriate for GRECLIPSE-1068 , but I thought it would be beneficial to know I was using the lifecycle-mapping since gmaven may have a part in this and it impacts how things work behind the scenes. I'm really beginning to think this issue is occurring because gmaven is being invoked by Eclipse. After all, these sort of issues were the reason that the m2e team decided not to directly support running maven plugins in Eclipse. Thanks again for all your help! PS: I do plan on trying to give feedback on the new m2e integration with the troubles I have had but at the moment I need to focus on "real work" as much as possible.
Hide
Permalink
Rob Winch added a comment -

I can go ahead and give the MAVEN_OPTS a try, but I am doubting that this is the fix I need. The reason I doubt this is because we have a Jenkins build that has never ran into this issue. I verified and the jekins build does not define the ReservedCodeCacheSize.

Show
Rob Winch added a comment - I can go ahead and give the MAVEN_OPTS a try, but I am doubting that this is the fix I need. The reason I doubt this is because we have a Jenkins build that has never ran into this issue. I verified and the jekins build does not define the ReservedCodeCacheSize.
Hide
Permalink
Andrew Eisenberg added a comment -

I agree that the MAVEN_OPTS is unlikely to work, but I don't think it would hurt anything.

I currently, Groovy-Eclipse m2e support for 1.0 is not really working. I hope to have a new version out later today. Once this is available, you should no longer have the "Plugin execution not covered by lifecycle configuration" errors appearing. And this may (just may) in turn affect your codecache problems.

When the m2e support is available, I'll post a comment on this bug.

I closed the first bug because you are no longer getting a verify error (unless I missed something). The verify errors were happening for known reasons and all of that has been addressed already. The codecache problem is something different, however.

Show
Andrew Eisenberg added a comment - I agree that the MAVEN_OPTS is unlikely to work, but I don't think it would hurt anything. I currently, Groovy-Eclipse m2e support for 1.0 is not really working. I hope to have a new version out later today. Once this is available, you should no longer have the "Plugin execution not covered by lifecycle configuration" errors appearing. And this may (just may) in turn affect your codecache problems. When the m2e support is available, I'll post a comment on this bug. I closed the first bug because you are no longer getting a verify error (unless I missed something). The verify errors were happening for known reasons and all of that has been addressed already. The codecache problem is something different, however.
Hide
Permalink
Rob Winch added a comment -

>> When the m2e support is available, I'll post a comment on this bug.

Thank you

>> I closed the first bug because you are no longer getting a verify error (unless I missed something). The verify errors were happening for known reasons and all of
>> that has been addressed already. The codecache problem is something different, however.

In the previous issue I posted the the Verify Error on the initial bug description. However, it turned out (as mentioned in this comment) that there was a Code Cache error that happened first and then subsequent openings caused the editor to give the Verify Error. You can see both the Verify and Code Cache errors in the greclipe-stacks.txt I did not have the opportunity to the editor again because I clumsily selected to restart eclipse from the error dialog. I suspect if I had been able to open it again I would have gotten the Verify error. Please note I am not trying to be rude. I just wanted to ensure that you knew that I still believe it is the same issue as this information might be useful.

Show
Rob Winch added a comment - >> When the m2e support is available, I'll post a comment on this bug. Thank you >> I closed the first bug because you are no longer getting a verify error (unless I missed something). The verify errors were happening for known reasons and all of >> that has been addressed already. The codecache problem is something different, however. In the previous issue I posted the the Verify Error on the initial bug description. However, it turned out (as mentioned in this comment ) that there was a Code Cache error that happened first and then subsequent openings caused the editor to give the Verify Error. You can see both the Verify and Code Cache errors in the greclipe-stacks.txt I did not have the opportunity to the editor again because I clumsily selected to restart eclipse from the error dialog. I suspect if I had been able to open it again I would have gotten the Verify error. Please note I am not trying to be rude. I just wanted to ensure that you knew that I still believe it is the same issue as this information might be useful.
Hide
Permalink
Andrew Eisenberg added a comment -

I understand your confusion. And maybe you are right, that the code cache problem caused the verify error.

On the one hand, because you were getting a verify error and now you are not, this implies that a build that you picked up has fixed it. Furthermore, I don't see how a code cache problem would be causing a verify error (it would cause lots of other problems, but probably not that).

On the other hand, even if this is the same problem, I'm not too fussed about opening a new issue for it. The summary now correctly encapsulates the problem and helps focus on the main issues.

Show
Andrew Eisenberg added a comment - I understand your confusion. And maybe you are right, that the code cache problem caused the verify error. On the one hand, because you were getting a verify error and now you are not, this implies that a build that you picked up has fixed it. Furthermore, I don't see how a code cache problem would be causing a verify error (it would cause lots of other problems, but probably not that). On the other hand, even if this is the same problem, I'm not too fussed about opening a new issue for it. The summary now correctly encapsulates the problem and helps focus on the main issues.
Hide
Permalink
Rob Winch added a comment -

No problem. Just wasn't sure if some of the information on the other issue might be helpful in figuring out this issue.

Show
Rob Winch added a comment - No problem. Just wasn't sure if some of the information on the other issue might be helpful in figuring out this issue.
Hide
Permalink
Rob Winch added a comment - - edited

For better or worse I am fairly convinced this issue has resolved itself. I believe increasing the code cache helps an issue that is happening in the gmaven plugin. If you like you can probably close the issue. If happens again I can either have it reopened or log a new JIRA.

Show
Rob Winch added a comment - - edited For better or worse I am fairly convinced this issue has resolved itself. I believe increasing the code cache helps an issue that is happening in the gmaven plugin. If you like you can probably close the issue. If happens again I can either have it reopened or log a new JIRA.
Hide
Permalink
Andrew Eisenberg added a comment -

Thanks for keeping up with this bug. I'll take your word for it that it is fixed. At least we now know that gmaven may be leaking the code cache and that a groovy-eclipse configurator for m2e 1.0 will likely avoid this problem.

Please re-open or create a new issue if you see the problem again.

Show
Andrew Eisenberg added a comment - Thanks for keeping up with this bug. I'll take your word for it that it is fixed. At least we now know that gmaven may be leaking the code cache and that a groovy-eclipse configurator for m2e 1.0 will likely avoid this problem. Please re-open or create a new issue if you see the problem again.

People

  • Assignee:
    Andrew Eisenberg
    Reporter:
    Rob Winch
Vote (0)
Watch (0)

Dates

  • Created:
    Updated:
    Resolved: