Hi,

Both Petr and me were in Toronto last week. And he pointed out that his
bug report #4895 about "updateHit gets called for each task" was still
open and that he had to work around that for fltrace. Nothing is more
motivating than meeting your users in person, so here is a fix and an
explicit test case for that issue:
    
    frysk-core/frysk/proc/ChangeLog
    2007-10-29  Mark Wielaard  <mwielaard@redhat.com>
    
           * Observable.java (observers): Now a HashMap.
           (add): Keep count.
           (delete): Likewise.
           (contains): New method.
           (iterator): Use HashMap.keySet().
           (removeAllObservers): Likewise.
           * Task.java (notifyCodeBreakpoint): Check whether observer is
           contained in the codeObservers of this Task.
           * TestTaskObserverCode.java (testMultiTaskUpdate): New test.
           (testMultiTaskUpdateUnblockDelete): Likewise.
           (requestDummyRun): New variant that takes a Task.
    
    frysk-core/frysk/proc/live/ChangeLog
    2007-10-29  Mark Wielaard  <mwielaard@redhat.com>
    
           * LinuxTaskState.java (sendContinue): Add logging.

The actual fix should have been only a simple one-liner in
Task.notifyCodeBreakpoint() to check whether the Task was actually
monitoring this particular breakpoint address. But it was a little bit
more involved than that.

Since Observers, especially Code Observers, can be added more than once
the ltrace testcase Petr added showed a latent bug where removing a Code
observer from one location would completely remove it from that task.
That is why Observable now keeps track of how many times an Observer is
added.

Writing the testMultiTaskUpdate showed another issue when you unblock
and immediately delete a Code observer (while the Task is resumed and in
the process of stepping over the breakpoint in question). I rewrote it
into its own testcase (testMultiTaskUpdateUnblockDelete) and filed a bug
report #5229. I also added some logging to LinuxTaskState.sendContinue()
to make it easier to track these issues down.

Cheers,

Mark