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.
I attempted to use remote profiling (of memory usage) using milestone 5 and the JVM core dumped. Remote JVM was running on Solaris (SunOS 5.8), NetBeans 4.1 beta 1 was running on Windows 2000. The remote application was WebLogic 8.1. The core dump occured when starting the application - I think when still instrumenting classes. I'll attach the hs_err log file generated when the JVM died. Let me know if there is anything else that I can provide that might help track down the problem.
Created attachment 20664 [details] log file generated when JVM died
Thank you very much for the feedback. This problem looks pretty unusual - the error log suggests it happens in pretty much the least place I would expect something to happen. One question: is it reproducible with your application?
Created attachment 20707 [details] Error reported when attempting CPU profiling
Yes, it is quite reproducable. Every time I start the app with memory profiling I get this core dump. I just tried it selecting "Application Performance" -> "Entire Application" and got a different failure - application dies with a NullPointerException. I'll attach a log. The IllegalAccessException at the start of the log doesn't happen when executing normally.
Ok, again many thanks for the feedback. To summarise, there are three separate issues here: 1. VM crash when getting object size on start of memory profiling. Likely caused by null object passed to memory profiling instrumentation code (which shouldn't happen). 2. NPE when doing CPU profiling. Apparently caused by the system classloader unable to retrieve class file bytes for a class that was supposedly loaded by either system or bootstrap classloader. Again shouldn't happen. 3. A reflection call failure. This is a known issue http://www.netbeans.org/issues/show_bug.cgi?id=55580 Has not been investigated yet. Number 3 is less acute, and will be investigated later. Number 1 and 2 I am going to look at immediately. As a first step, I am going to send the user a "patch" - several updated JARs from JFluid distribution, that contain fixes for some bugs we previously found, plus diagnostics in the places where these reported bugs occur. I will explain how to use it in the cover e-mail, but PLEASE ACKNOWLEDGE THE RECEIPT OF THIS PATCH! That's because some e-mail systems can swallow "suspicious" attachments without notice. Furthermore, since Issuezilla is going to be unavailable for the next 2 days, please communicate with me (Mikhail point Dmitriev at sun point com) directly. Regards, Misha
Update: - Problem 3 above has been fixed, see link to the relevant issue. - Problem 1 above has been identified as a likely VM bug; a workaround is implemented in the tool; we are awaiting confirmation that it works from two other users who reported this problem and to whom we sent a patch. - Problem 2 still needs investigation. Will probably need to open a separate issue for it. Last time we sent a patch with just JFluid engine JARs to the user who originally reported the problem, it didn't work for him for some reason. If the user can point us at some ftp site, we can upload the whole .NBM of the preview version of the next JFluid release, that contains the above fixes.
Thanks Misha. Unfortunately I don't have access to a public ftp site, otherwise I'd definitely be keen to try the fixes. Regarding problem number 2, I've just tried profiling locally (downloaded NetBeans and profiler for Solaris) and the problem didn't occur. (This was with the original milestone 5 jar files). i.e. it is specific to remote profiling.
Update: problem 1 above appeared to be a bug in the bytecode instrumentation and has just been fixed. So, what remains is problem 2, and I am going to open a separate issue for it.
Opened a separate issue for problem 2: http://www.netbeans.org/issues/show_bug.cgi?id=57231. Closing this one now.
Verification of old issues.
Closing old issues.