Re: User Defined Types and the Debugger

[Keith Seitz <keiths@redhat.com>]

On Mon, 14 Oct 2002, Matteucci, Cheryl A (N-B&M Assoicates) wrote:

> Support,

Please do not call me "support": I am not a support engineer, you don't 
have a support contract with me, no one has a support contract with me, no 
one pays me to work on Insight, and I get very, very little help from 
others.

> A = B / C     where A = floating point class attribute (not static or const,
> class scope)
>                             B = locally defined floating point variable (not
> static or const)
>                             C = user defined floating point type that is
> static and const (class scope).
> 
> All of the above is in the same class.

It would be much better to have an actual testcase or some lines of code 
to inspect. There has been some churn on c++ inside GDB over the past few 
months, and it is difficult to determine exactly what is going on without 
something more concrete.

> The debugger shows me a value of 0 for C, but it is statically initialized
> to 70.0.  I set a breakpoint at this line of code and when I step through
> this line of code, the debugger gives me a segmentation violation (I assume
> because it thinks I am dividing by zero).  But if I set a breakpoint at some
> point after this line of code, the debugger successfully executes it.  When
> I run the program outside of the debugger, it is successfully executed.

The debugger does not generate segmentation faults in programs: the 
programs do that all by themselves. There could be many reasons why your 
statically initialized variable does not appear initialized.

The best thing to do right now is to post a short testcase demonstrating 
the problem. What version of GDB/Insight are you using? What version of 
gcc? If you suppress Insight and use command-line GDB (gdb -nw/insight 
-nw), does the problem also exist?

Keith