"Kővágó, Zoltán" <dirty.ice...@gmail.com> writes: > As discussed here[1], I'm splitting the qapi related patches from my > previous -audiodev patch series. These are patches 2--7 and 9 from my > previous patches. (Patch 1 was merged into -trivial in the meanwhile.) > > Please review. > > [1]: http://lists.nongnu.org/archive/html/qemu-devel/2015-09/msg01505.html
Ongoing QAPI work has shown that the visitors code is brittle, and we need to retrofit a proper contract[*]. As discussed before, you're proposing dropping a design assumption of OptsVisitor. Messing with design assumptions is always scary, but it's even scarier when the whole edifice built on it has been found brittle already. I'm afraid we need retrofit a contract *first*, both to build understanding on how this sucker is supposed to work (right now nobody knows), and to hopefully make it more robust. Your series adds urgency to the job. [*] http://lists.nongnu.org/archive/html/qemu-devel/2015-09/msg03363.html
