For patch 17 on onwards it was just seeing what the actual benefit of the derived class was - I think I get it later on but you should mention it up front.
I do think we need to address the ordering constraint in 21 - are we introducing one or just formalising what has been created? If we are introducing one then can we a) do it a better way with the structuring of QOM or b) enforce it so new models don't run into unexpected bugs. On Wed, 3 Feb 2021 at 17:10, Claudio Fontana <cfont...@suse.de> wrote: > > Hi Alex, > > thanks for your review, > > On 2/3/21 5:57 PM, Alex Bennée wrote: > > > > Claudio Fontana <cfont...@suse.de> writes: > > > > <snip> > > > > Final comments. I think overall this series is looking pretty good > > although I got a bit lost at the end when we started expanding on the > > AccelClass. > > The main yuck was the start-up ordering constraint which > > To be sure, are you referring to tcg_accel_ops_init(), ie your comments > towards the end of PATCH 17? > > Ciao, > > Claudio > > > would be nice to remove or failing that catch with some asserts so weird > > bugs don't get introduced. > > > > Paolo, is it worth picking up some of the early patches to reduce the > > patch delta going forward? > > > -- Alex Bennée KVM/QEMU Hacker for Linaro
