[Dwarf-Discuss] target address dependency of the FDE header in DWARF CFI

[David Anderson]

On 12/10/2011 10:04 AM, Eli Bendersky wrote:
>>> The CU specific address_size is used for that CU's  CIE/FDE set.
> David, could you elaborate - what to you mean by "that CU's CIE...".
> How do I know to which CU a CIE belongs?

Each FDE has an initial_location and  address_range indicating what 
instruction area
is applicable to the FDE.     That FDE refers to a specific CIE.     A 
CU also applies
to a range of addresses (some CU  DIEs have DW_AT_low_pc or 
DW_AT_ranges, but
certainly the functions in a CU will have addresses or address ranges).

So using an address as the key, find the CU applicable and the FDE 
applicable (which determines
the CIE applicable).   Hence finding a CU that 'owns' (speaking 
loosely)  a particular CIE.
In other words, the ownership is implicit in the DWARF data, not explicit.

It's normal for a single CU to have a single CIE, but that is a detail, 
and CIEs can
be split or merged without changing the notions here  (sharing a CIE 
across multiple
CUs could make things one's dwarf-reading code a little messier).

Normally the address-association here is really simple, and debuggers 
have all the lookup
tools in hand to do the association easily.

Hope this makes sense.  Always possible I'm just wrong :-)
DavidA.

-- 
It's important to ask yourself, when you're considering how much greener the grass is on the other side, if you've even been watering yours.  -- Unknown