On Mon, 2 Mar 2020 at 16:55, Igor Mammedov <[email protected]> wrote: > > On Mon, 2 Mar 2020 15:41:13 +0000 > Peter Maydell <[email protected]> wrote: > > Hi Igor, I just noticed this, and I don't think it's the > > right thing. The board model should have its own state > > structure which contains any objects it creates. Just > > because there happens currently to be only a single > > object in this case doesn't mean we want to lose the > > structure. In particular, we now just leak the > > pointer to the TYPE_AW_A10 object, rather than having > > it be tracked by being pointed to from the MachineState. > > Being able to avoid just leaking pointers to objects like > > that is one of the reasons I like having a MachineState now. > The reason why this structure was removed was that it wasn't > MachineState object but a random structure which was a common > pattern in pre-QOM qemu. > > Code was allocating 's', assigning pointer to s->a10 member and > then happily loosing both pointers in the end.
Oh, so it was -- I misread the code. > I can convert it to MachineState derived board as > an example to follow. No, you don't need to do that, I hadn't realized it wasn't already MachineState derived, because it looked superficially like it was. Apologies for the false alarm. thanks -- PMM
