On Tue, 2017-01-03 at 14:48 +0100, Martin Kletzander wrote:
I lean towards merging this or a comparable solution. It's true that we aren't currently hitting this on our main targets, but relying on undefined behavior is definitely something we want to avoid, plus I don't see any real drawbacks in changing this to a macro. Feel free to have a look at the other approaches (and whole threads) and see what you like: - https://www.redhat.com/archives/libvir-list/2016-June/msg02173.html - https://www.redhat.com/archives/libvir-list/2016-December/msg00379.html Yeah, it goes a long way back, and I know about even longer standing clang problems that we're just not dealing with.
I like my own solution better than either yours or Jano's :) And given Dan's okay with it as well, I will push the patch in a while unless someone disagrees. Jano, does the other stuff you fixed with your series back in June still apply? Going through the thread, it looks like I tried and failed to reproduce the build failure on my own setup at the time. -- Andrea Bolognani / Red Hat / Virtualization