[Dwarf-Discuss] Fission + cross-CU references (ref_addr)
Fri May 5 19:06:20 GMT 2017
On Fri, May 5, 2017 at 11:50 AM Robinson, Paul <paul.robinson at sony.com>
> What guarantees that the information referred to is the same across all
> "duplicate" TUs or otherwise does not affect the semantic content of the TU?
> The same guarantees that exist without TU->CU references, I would think?
> It's up to the implementation to ensure that any two TUs are semantically
> No it's not. DWARF specifies what makes TU's semantically equivalent, by
> specifying what feeds into the signature computation, and matching
> signatures provide the guarantee. References use forms that are not
> included in the signature and therefore the referenced information is not
> included in the guarantee.
Fair enough, I though maybe that'd changed/become more open-ended - LLVM
goes off-spec here and uses a hash of the mangled name of the type. This
ensures more (semantically equivalent) definitions can be deduplicated and
reduces the overhead/complexity of the implementation. (it does risk LLVM's
hashes colliding with another implementation's hashes in
surprising/incorrect ways, though).
In any case that's somewhat of a tangent to my other/broader
point/curiosity about the structural representation/implications of having
a concept of "contributions" that's not "the whole section" or "the unit"
in the DWO file.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Dwarf-discuss