This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
Summary: | Editor frozen, CPU at 100% | ||
---|---|---|---|
Product: | apisupport | Reporter: | Jan Becicka <jbecicka> |
Component: | Project | Assignee: | Jesse Glick <jglick> |
Status: | RESOLVED WORKSFORME | ||
Severity: | blocker | CC: | dbalek, issues, issues, lttldrgn, mmirilovic, pribyl, pzajac |
Priority: | P3 | Keywords: | PERFORMANCE, RANDOM, THREAD |
Version: | 6.x | ||
Hardware: | All | ||
OS: | Windows XP | ||
Issue Type: | DEFECT | Exception Reporter: | |
Attachments: |
Bunch of FTDs
Messages.log just for record. No exceptions there. messages.log from editor freeze Thread dump while netbeans locked up Thread dump |
Description
Jan Becicka
2007-08-08 10:09:56 UTC
Created attachment 46310 [details]
Bunch of FTDs
Created attachment 46311 [details]
Messages.log just for record. No exceptions there.
From the thread dumps, it seems that the SourceForBinaryQuery.findSourceRoots takes a very long time to answer on apisupport/projects. Yes, all queries on NBM projects will take a while if it needs to scan for project metadata the first time. Some day queries will respond faster. (Not clear from description whether this metadata should have already been loaded or not.) But the proximate cause of the freeze is that JavaCodeTemplateProcessor.initParsing is blocking EQ running a compilation task, which I understand is forbidden. Check the thread dump for EQ and the log: WARNING [org.netbeans.api.java.source.JavaSource]: JavaSource.runUserActionTask called in AWT event thread by: org.netbeans.modules.editor.java.JavaCodeTemplateProcessor.initParsing(JavaCodeTemplateProcessor.java:553) As Honza B. stated, he had to kill&restart the IDE. The CPU was hogged for ~5mins. Even if the AWT Thread would not wait for the code completion to finish, there would be no Java support (no hyperlinks, no CC, no sem. coloring, no mark occurrences) until the currently running CC query would finish - it is unacceptable to wait for ~5mins, IMO. For the code templates running in AWT, it is question to Dusan how much can be done about this. Have not heard of such long delays. Is the D drive a local disk? Which project was being worked on? Anyone else had a similar problem? Nothing that I know of has changed recently in apisupport. I just ran 070808 (Basic) with JDK 6 on a fresh userdir and opened java/sourceui. Took maybe 5-10 seconds to open with the progress dialog. Java scanning started a few seconds later, took maybe a minute or less incl. JRE (5) scanning. After that, editor on random source files in the project was responsive for code completion at least. No freezes or unreasonable delays encountered. What information is needed now? Not clear why is this bug INCOMPLETE. Whether the problem is reproducible (on your normal development machine or on other machines), whether anyone else is seeing similar problems, whether the problem began happening recently or has been a problem for a long time, etc. I cannot reproduce any delays on the same order of magnitude as what you describe, so it is unlikely I am going to be able to improve anything. I cannot reproduce it again. It happened once. I propose to lower priority and remove INCOMPLETE kw. Good to know it is not always like this, anyway. I can try to track I/O usage opening or editing a typical module since I suspect it was just a bad day for I/O for some reason, and apisupport certainly uses a lot of I/O. Should be fixed for 6.0 if possible (and if reproducible). TM=6.0 for now. Not known to be reproducible, and anyway likely affects only developers of nb.org modules (rather than general apisupport users). I have experienced this a handful of times as well. I just d/l NB6.0rc1 and it still seems to happen when I work with large projects (10000+ files). While working, performance seems to degrade over time and sometimes it will lock up and not return. I've been running jconsole to monitor NB and some other performance issues related to the Navigator window when switching source tabs (see issue 117602). Basically when the heap limit is reached it gets stuck in a continual garbage collection and obviously can not recover. In the past I did not get any errors, but I let it run a while longer (30+ minutes) and finally got a OutOfMemoryError exception in the messages.log that may help point to what is causing this error. I will attach that along with a thread dump in case it helps. Created attachment 53152 [details]
messages.log from editor freeze
Created attachment 53153 [details]
Thread dump while netbeans locked up
Product Version: NetBeans IDE 6.1 RC2 (Build 200804170002) Java: 1.6.0_10-beta; Java HotSpot(TM) Client VM 11.0-b11 System: Windows XP version 5.1 running on x86; Cp1250; cs_CZ (nb) Well, this happened to me too. The IDE was frozen, CPU 100% I happened after working with code completion in a really small sample project ( I created new one, added just one class. No additional libraries). I started to type some word, like sys| , togled code completion and deleted the "sys" (letter after letter, with the CC still on). The CC was changing according to the current word. Then the IDE freezed. It happened again, after restarting of the IDE, this time I did not deleted the letters in "sys" word, I just set the cursor to the start of the word by arrow keys. I am attaching the thread dump (But I am sorry, but it is probably not complete, although I copied all that was in the console output) Created attachment 60682 [details]
Thread dump
But it seems to be really random. I just started with new userdir, created the same project, tried the same approach and it works smooth and fast :( The newly attached thread dumps do not even mention apisupport, so if you can reproduce even sometimes, probably better to file a separate bug. The last thread dump is rather an #132864 incarnation. That should be fixed in the current dev build. Issue #132864 "Content of Copy to Folder editbox..." in php/project? Typo? Though I was still seeing the original problem in 6.1beta, I stopped seeing the issue with 6.1 RC2. In fact the latest release is fantastic. No more out of memory exceptions. Good jobs NB developers. OOPS - #132863 is the correct issue number. Sorry. No clearly identifiable problem here. If future freezes occur, as usual file them with thread dumps and any steps to reproduce that might be available. |