Bug 3500

Summary: Inferior process quits with 'Trace/breakpoint trap'
Product: frysk Reporter: Mike Cvet <mcvet>
Component: generalAssignee: Mark Wielaard <mark>
Status: RESOLVED FIXED    
Severity: normal CC: cagney, mark, npremji
Priority: P2    
Version: unspecified   
Target Milestone: ---   
Host: Target:
Build: Last reconfirmed:
Bug Depends on: 3571, 3676    
Bug Blocks: 1522, 1633, 3346    

Description Mike Cvet 2006-11-10 14:56:36 UTC 
When opening the SourceWindow and doing some stepping; if one quits frysk, or 
'runs' the process (deleting the observer in either case) the inferior process 
quits with the above error.
Comment 1 Mike Cvet 2006-11-10 17:48:28 UTC 
I should not that this is on an x86_64 dual core intel machine on FC6
Comment 2 Mike Cvet 2006-11-13 21:49:10 UTC 
... And doesn't appear to occur on FC5 i386.
Comment 3 Mike Cvet 2006-11-16 21:31:22 UTC 
I had a kernel update this week and this appears to have gone away. I'm unsure 
if its a kernel problem though.

The kernel which the problem appeared on was:

2.6.18-1.2798.

I'm now on 

2.6.18-1.2849.
Comment 4 Mike Cvet 2006-11-22 18:28:21 UTC 
This original bug occured with just having TaskObserver.Instruction blocking the 
Task. With that latest kernel update it seemed to dissapear; but now when adding 
an additional Observer from TaskObserver.Code, the problem has reappeared. Not 
sure what the correllation is.
Comment 5 Mark Wielaard 2006-12-06 14:27:47 UTC 
Testcase committed.

frysk-core/frysk/proc/TestTaskObserverInstructionAndCode.java

This crashes the inferior after reinstalling a Code observer and trying to
resume the Proc. The test is disabled in CVS.
Comment 6 Mark Wielaard 2006-12-07 15:23:58 UTC 
The testcase in comment #5 seems to be a different problem. It has now been
reported as bug #3676
Comment 7 Andrew Cagney 2006-12-07 20:45:16 UTC 
Consider:

01: jump 5 <- PC


...
04: nop    <- breakpoint
05: 

And a single step
Comment 8 Andrew Cagney 2006-12-07 20:46:43 UTC 
Consider:

Consider:

01: jump 5 <- PC


...
04: nop    <- breakpoint
05: nop    <- breakpoint

And a continue
Comment 9 Andrew Cagney 2006-12-07 20:49:05 UTC 
Consider:

04: nop    <- breakpoint <- PC
05: nop    <- breakpoint

And a single step
Comment 10 Andrew Cagney 2006-12-07 20:50:18 UTC 
Consider:

04: nop    <- breakpoint
05: nop    <- breakpoint  <- PC
o6: nop    <- breakpoint

And a continue
Comment 11 Mark Wielaard 2007-02-12 17:32:34 UTC 
Does this still happen?
I played with stepping in the gui and it seems to work nicely.
Stopping, running and quiting keep the program going just happily for me.
Comment 12 Mike Cvet 2007-02-13 14:27:35 UTC 
I can no longer reproduce the original behaviour for this bug. Nice work!
Comment 13 Mark Wielaard 2007-02-13 14:39:14 UTC 
Reported as fixed. Yeah!