On 7 August 2018 at 14:21, Peter Maydell <peter.mayd...@linaro.org> wrote: > If you want to make it a bug to > not have a needed function, you need to have code that checks > for this in device init and aborts, so that device authors > don't fall into this trap.
We should also sanity-check the subsection names at device init, because there's a hidden requirement that they start with the name of the parent VMSD, and if you get that wrong then nothing complains except that trying to reload the vm state silently doesn't work right. thanks -- PMM
