[llvm-dev] Weak symbol/alias semantics

Rafael Avila de Espindola via llvm-dev llvm-dev at lists.llvm.org
Wed Jan 18 08:49:25 PST 2017


> I suppose I could do this by transforming aliases to the proposed canonical
> form just before we do the weak resolution in the ThinLTO backends. Longer
> term, how would such a change be implemented? I.e. would the current IR
> continue to be accepted (and generated by clang), and the transformation
> occur within LLVM early in the pipeline?

Longer term I would love to

* Reject the old form in the verifier.
* Upgrade every producer.
* Upgrade old bitcode files when reading.

Cheers,
Rafael


More information about the llvm-dev mailing list