[Dwarf-Discuss] address values in constant forms
roland at redhat.com
Fri Dec 11 16:13:13 PST 2009
> If you tell consumers they are supposed to modify DW_OP_addr
> operations, in an unspecified and platform-specific way, I think
> you're going to create a pretty big headache for e.g. Cell
> and Harvard architecture systems.
> Maybe you get lucky and the results are never worse than today,
> but I think you're generating a lot of complexity.
I am not describing some entirely new scheme. This is the status quo for
all cases where DW_FORM_addr and DW_OP_addr are used today. If it were
not, no ELF/DWARF system's debuggers would today make any sense of address
constants in position-independent DSOs without those extra ELF relocs that
we do not use today. My suggestion was just to clarify and recognize that
reality in the DWARF spec and to explicitly encourage using these forms
consistently for all cases with the same address-constant semantics.
More information about the Dwarf-Discuss