Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.

Still can't stop at crash point

paramont
Beginner
1,765 Views
I have choose all the items in debug->exceptions->stop always->change,I still get the error message as follows:
The thread 0x498 has exited with code 3 (0x3).
The program 'C:francf2dDebugf2d.exe' has exited with code 3 (0x3).
I have even added 0x3 into the exceptions list, it still didn't work.

Any idea about this?
0 Kudos
11 Replies
Steven_L_Intel1
Employee
1,765 Views
Doesn't look as if it crashed - looks as if the thread exited "normally" (although it supplied a return status of 3). Do you see any other messages about exceptions?

Steve
0 Kudos
dmoog4
Beginner
1,765 Views
I also continue to be plagued by this problem. I am running CVF 6.6.0 under Win 2000 Pro, and nearly every program error on the debugger crashes without stopping; that is, the disassembly window opens with a message "User breakpoint called from code at ..." and I can't examine variables or execute debugger commands, and have no indication where the error occurred. I have "Stop always" set for all exceptions. I have been finding crash points the hard way for a long time but would very much like to get the debugger to stop on errors.

Also I should note that I have the problem with crashing upon starting the debugger (i.e. F5), which I have been avoiding by setting a break point at the first command, but this is far from ideal in my case.

Any help would be extremely valuable.
0 Kudos
Steven_L_Intel1
Employee
1,765 Views
dmoog4, I'd suggest not using "stop on all exceptions" and let the Fortran error handler take care of it. I think you're getting an I/O error - the "user breakpoint" is how the run-time library signals these.

An alternative is to use the "context" menu in the debugger to set the context back in the call stack far enough to show your code.

If you continue to have problems, please send us a complete example in a ZIP file to vf-support@compaq.com.

For the debugger crashing problem, I suggest deleting the file with a .opt file type in your project folder.

Steve
0 Kudos
paramont
Beginner
1,765 Views
Do you mean that this is because of my project is designed to exit at this point? not because of CVF5D?

I didn't see other messages for this exit.
0 Kudos
Steven_L_Intel1
Employee
1,765 Views
paramont, I don't know exactly what your program is doing. In DVF 5.0, any run-time error would just display an error message and then terminate the program. Do you see any error messages? They may be in another window.

Steve
0 Kudos
paramont
Beginner
1,765 Views
This is a finite element analysis code. When the code crashed, the dos command window will be closed too. The only messege I saw was the two lines I listed in posts above.

I may email the whole code if necessary.

appeciate your help!
0 Kudos
Steven_L_Intel1
Employee
1,765 Views
paramont, since you are still using DVF 5.0, we won't be able to help you at vf-support. If you'd like formal support, please upgrade to version 6.6. Your 5.0D is now five years old.

Steve
0 Kudos
Jugoslav_Dujic
Valued Contributor II
1,765 Views
As Steve said, "User breakpoint called at..." usually indicates an I/O error; another possibility is (de)allocation error. Debugger stops in assembly code of a system dll. However, you can get the offending line from your code by selecting "Call Stack" debugger window and double-clicking your first routine from the list.

As far as I know, it is not possible to track down math library errors, such as sqrt(-1.) or acos(2.) down to offending line. Am I right Steve?

Jugoslav
0 Kudos
Steven_L_Intel1
Employee
1,765 Views
Jugoslav, it has been so long I don't remember exactly how math errors were handled in 5.0. I suspect you are correct. I do know that in 6.0, the debugger will show the source line for all errors.

Steve
0 Kudos
paramont
Beginner
1,765 Views
Thanks for all the replies.

I ran this code in a separate DOS command window, it will crash and give me an error message of:
run-time error M6201: MATH
- sqrt: DOMAIN error

Thanks.
0 Kudos
Steven_L_Intel1
Employee
1,765 Views
I don't think there's a straightforward way of catching SQRT errors in 5.0.

Steve
0 Kudos
Reply