[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
[jdt-debug-dev] RE: [platform-debug-dev] Ideas for Bug 30816?
|
Jared,
Thanks for the response. Although I'm subscribed to the mailing list(s) at work, I'm actually doing this work at home. Thus, I was grubbing around in the JDI code during my vacation.
The KVM is happier with class prepare events to the point that line numbers and stepping work with the prepare requests and don't work otherwise. I just "cloned" the implementation of a similar method to set up the class prepare request if the KVM was on the other end of the connection. You are correct that that method does not exist outside of my workspace.
I'll have to go revisit the concurrency problem yet again. Although I wrote it up as being from two different threads, I believe that later I saw that was not true. Strangely, I can place a Thread.sleep in the resume method of any duration and it won't make a difference. But stopping at a breakpoint in that method seems to ALWAYS work. I guess I'm unclear on how much threading is going on in the debug implementation. I just keep wondering if somehow the KVM is getting bogged down and somehow breaking in the resume method helps it catch up. Not sure really. If you guys have thoughts on what I should be looking for in this issue I will gladly do what I can. This is really the primary concern, as the KVM crashes with this problem... Although the class prepare issue would soon follow if we can fix resume. Any direction for providing useful information would be most appreciated.
Thanks,
Craig
-----Original Message-----
From: Jared Burns [mailto:jaredburns@xxxxxxx]
Sent: Wednesday, January 07, 2004 12:37 PM
To: platform-debug-dev@xxxxxxxxxxx; jdt-debug-dev@xxxxxxxxxxx
Subject: Re: [platform-debug-dev] Ideas for Bug 30816?
Sorry about not responding to your last comment in the bug report. You posted
that update while we were all away for the holiday. :)
In the bug report, you say you're able to make the KVM happier by enabling
class prepare requests. The specific method you say you're calling doesn't
exist. I assume you're creating this method and making it create a
ClassPrepareRequest for "*"? When you do this, how much happier does the
KVM's debug support become? Does it just fix line numbers?
About the concurrency problem you mentioned... You'll just have to look at
that one more closely, preferrably sharing more information with us as you
go. You said you see two resume events to our VirtualMachineImpl coming from
separate threads. Which threads are they and what do the stack traces look
like? Do you see any obvious problems?
Thanks for your help on this problem. We don't have cycles to spend exploring
this ourselves, so your help is appreciated.
- Jared