how --builtin--return--address() works??

John Fine
Mon Sep 15 12:09:00 GMT 2008

I agree with Andrew's answer.  I just want to add a few details where I 
think it might have been misleading.

Andrew Haley wrote:
>> 3) The location where that return address is stored on stack may vary 
>> depending upon  number  of arguments  passed to the function, or number of
>> local variables inside the function, or whether the function is returning
>> a value or not........
> Not usually, no.  On those machines that pass the return address on the stack,
> that address is on the top.
 From the point of view of code running inside a function, the location 
of the return address typically does vary with the number of local 
variables.  It does not vary with the number of arguments.
>> How __builtin_return_address() function knows where exactly the return
>> address is stored on stack????
> The compiler knows where everything is: the return address is nothing
> special.
Often the compiler can't know where "everything" (relevant to the 
current stack frame) is, because it often can't know how many arguments 
were passed.

But (as you indicated) it always knows where the return address is.  If 
it didn't, then it couldn't even generate correct code for the function 
to return.

Because of the flexibility the compiler has with register use, it 
typically would be impossible for a something like 
__builtin_return_address() to be fully defined by a header file.  The 
compiler can know all about the stack usage for saved registers and 
local variables, etc. so at any point in a function it knows where the 
return address is.  But code in a header file can't know that much.

More information about the Gcc-help mailing list