Re: variables in scopes

[Vara Prasad <prasadav at us dot ibm dot com>]

Ulrich Drepper wrote:

> Vara Prasad wrote:
> ....
>  
>
> I'm talking about either inlined functions or code like this:
>
> int foo (int a, int b) {
>  int c = a;
>  if (c != 0) {
>    int a = 42 + b;
>    bar(&a);
>    c += a;
>  }
>  return c;
> }
>
> Here 'a' is shadowed.  You can easily construct a similar case with
> inlining.  And both variables could be parameters, of local variables,
> or any combination.
>  
O.K, now i understand what you mean by shadow variables and it makes sense.

> In DWARF the if block is encoded with DW_TAG_lexical_block.  This is one
> level down from the block of the function.  So the variable could
> theoretically be accessed for a probe at the call to 'bar' with
>
>  @a@
>
> and the parameter with
>
>  @a\1@
>
> meaning, one level up.
>
>  
I still prefer function syntax to access these variables than language 
constructs.

>  
>
> > Can you give me some example usage for this feature in a systemtap script?
> >    
>
> I think you see that the misunderstanding has prevented you from seeing
> an application.  A probe at the call to 'bar' certainly might want to
> look at the parameter 'a'.
>
>  
I still go back to my original point that we should let probe handler 
have access to only the local variables that it will have access if the 
handler was written inline in the function being probed. In this case 
since inline code will not have access to parameter "a" i am thinking 
probe handler should not have, may be original author had a good reason 
why it was shadowed.