[ImageJ-bugs] [Bug 1013] New: Apparent deadlock freezes Fiji

[bugzilla at fiji.sc]

http://fiji.sc/bugzilla/show_bug.cgi?id=1013

            Bug ID: 1013
           Summary: Apparent deadlock freezes Fiji
           Product: Fiji
           Version: unspecified
          Hardware: PC
                OS: All
            Status: NEW
          Severity: blocker
          Priority: P5
         Component: Generic
          Assignee: imagej-bugs at imagej.net
          Reporter: ilan.tal at gmail.com

In 2 of my dialog boxes I have a button which calls Close All

IJ.runPlugIn("ij.plugin.Commands", "close-all");

I run ./ImageJ-linux64 in a terminal to see any possible error messages, but
there are no errors. I also see the problem on a Mac, which is the first place
it was reported.

The program I use to demonstrate the problem is Run from CD from the Pet-Ct
package. Any dialog which calls the above runPlugIn will do just as well. Read
BI studies, which needs a database also shows the problem.

To demonstrate the problem we need at least 2 cines running at a high frame
rate. A frame rate of 1 will work, a frame rate of 10 will hang. What I do is
bring up 2 series and run both of them in cine with a frame rate of 10.

Then I press my button. The first cine will close, but the second won't. The
cine will freeze, as will everything else in Fiji. The only way to solve the
problem is to force quit Fiji.

Since there are no error messages, it looks like a deadlock. There was a bug in
ImageJ which I reported to Wayne Rasband and he fixed in the daily build, so
you need to use the latest ImageJ to avoid the bug which Wayne fixed.

An interesting point is that Close All in the File menu works OK. Likewise
ImageJ works OK even with the button. Only Fiji hangs.

I am guessing that there is a deadlock somewhere. Because of the high frame
rate there are many events. The first cine gets closed and to continue some
internal loop for the second cine, an event telling that the first cine has
closed needs to arrive. That event is apparently lost so the system patiently
waits forever.
This is obviously a guess on my part. My Netbeans uses ImageJ for debugging and
ImageJ works no matter what I do. I don't have the full source code for Fiji on
my machine so it is difficult to pick up where the deadlock might be. This is
without getting into the fact that I don't really know what is going on inside
the system code. In short, it is over my head.

I will be happy to help if I can do anything to help with this bug.

Ilan

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://imagej.net/pipermail/imagej-bugs/attachments/20150301/8bf44baa/attachment.html>