Message #786

From: Melinda Green <melinda@superliminal.com>
Subject: Re: [MC4D] The final build?
Date: Tue, 17 Nov 2009 00:30:11 -0800

Brandon Enright wrote:
>> n-jdk-1.6.0.17 on Linux amd64/x86_64
>>
>> I’m pretty adept with GDB and troubleshooting compiled applications
>> but I don’t know enough about Java to be useful. If you can give me
>> the steps to force Java to print a stack trace when it is hung I’d be
>> happy to do so. I’m also up for running a special build if you have
>> some patches in mind. I may need my hand held through the build phase
>> though.
>>
>
> I ended up figuring out how to take a stacktrace (jstack -l <pid>). It
> looks like it could be a deadlock. Almost all of the threads are
> waiting on various locks.
>
> Rather than clutter the list, I have put the stacktrace here:
>
> http://noh.ucsd.edu/~bmenrigh/jstack_mc4d_4_0.txt
>
> This was taken with the application hung after pressing ^F
Hey, that’s cool. Thanks for looking into this, Brandon; that’s a very
helpful tool! I don’t know what all the SyntheticImageGenerator threads
are doing, but the most important thread is always the AWT EventQueue
thread in which all the application UI work happens (or should happen!)
It’s especially great to have someone actively testing on Linux. Now if
only we could find a similarly motivated Mac user, we’d have the bases
covered.

As you now know, you weren’t using the latest build (r164). That’s
entirely my fault because I uploaded and deleted several versions today
as I rapidly found problems, fixed them, and created more problems. At
this point I seem to have repaired the worst of them. There are still
some annoying jumping around of the panels when you rebuild puzzles, but
everything *appears* to be functionally solid to me. Please retest as
much as you have patience for and let us know how it goes. With a bit of
luck tomorrow could be the day!

Thanks again,
-Melinda