[prog] This segmetation fault makes no sence to me.

[Sue Stones]

On Sat, 31 May 2003 05:34, Jimen Ching wrote:
> On Fri, 30 May 2003, Sue Stones wrote:
> >anual salary is 350000
>
> Is this a software engineer's salary?  ;-)

Yes but note it doesn't say what the currency is!  
$US 1.00 =  118.350 JPY (Japan Yen)
$US 1.00 =   1,637.86 ITL (Italy Lire)
$US 1.00 =   8,254.38 IDR (Indonesia Rupiahs)
	:-)

<cut>
> The same goes for andCo.  Note also, I didn't see an assignment operator
> defined below, so one is generated for you.  And it just does a byte by
> byte copy.  I'll mention why this is significant later.
<cut>

I do have one written, I just didn't copy it here because I didn't think it 
was relevent.  (wrongly).

> Does Manager inheret from Employee?  If so, you need a mem-initializer
> like so:
<cut>

Yes it does.  I had thought that the mem-initialiser and calling the 
constuctor of the base class were the same thing.  I have re-erad that 
section of the book and I understand it now.  Thanks.  (Now I just have to 
remember it!)

<cut>
> The output looks like andCo.add() succeeded and returned.  When 'return 0'
> is executed, the destructor comes into play.  This is most likely where it
> segfaulted.  As I mentioned above, an assignment operator function was
> probably generated (if you didn't define one).  This will perform a byte
> copy of the contents of Manager.  

Yes.  When I ran the debugger it showed that the destructor for Employee is 
called as soon as a Manager has been created.  But fixing these things 
dosen't solve the problem.  If fact I can still use the derived class fine up 
untill the end of the program.

>If Manager inherets from Employee and
> Employee has a 'char *' data member, then 'smith' is probably pointing to
> freed memory, since the temporary used in
>
<cut>

This seems to be the one mistake I didn't make.  


> This is just a guess, since I don't know the class definition of Manager
> nor it's destructor definition.  But the segfault is most likely due to
> some memory de-allocation problem.
>
> With gdb, if you compiled all of the code with -g3 -O0 (enable debugging
> with macro symbol and disable optimization), then you can put a break
> point in the destructor and walk through it.  I.e.
<cut>

Yes I agree, but I still haven't found it!!!

Thanks for the instructions for gdb, it has helped a lot.  I will try and 
figgure out Kdevelop very soon now.  (I didn't know what it was for untill 
Meredydd mentioned it.)


> P.S.  I just looked at the main function again, and I noticed 'smith' is a
> local variable.  Are you sure you want to add a variable on the stack to
> some container in andCo?  That doesn't look right.  This could also be the
> problem.  I recommend going through the process of running gdb to
> determine the cause of the segfault.  This will tell you both the cause of
> the segfault and give you experience with using gdb.

Ah yes annother confusion set straight in my brain.  Who knows what language 
I was reverting to there!   I had to look this up in the text book too!

Thanks for your help.  It hasn't solved the problem whit that program, but I 
have learnt a lot.  Since I can still continue to write the program, (which 
doesn't fit with my analysis of where the problem stems from) I will possibly 
just do that and try and  come back to the SEG Fault later.

sue