On Tue, Jul 26, 2016 at 04:16:04PM +0200, Igor Mammedov wrote: > On Tue, 26 Jul 2016 18:41:22 +0530 > Amit Shah <amit.s...@redhat.com> wrote: > > > On (Tue) 26 Jul 2016 [14:58:39], Igor Mammedov wrote: > > > > This was flagged by a nightly run of the static checker when this > > > > series was pulled. On a 'before' tree, ie one w/o the patches, do > > > > this: > > > > > > > > qemu -dump-vmstate before.json > > > > > > > > and for after: > > > > > > > > qemu -dump-vmstate after.json > > > > > > > > then, > > > > > > > > python ./scripts/vmstate-static-checker.py -s before.json -d after.json > > > > > > > I don't think it is valid comparison though, as it compares default PC > > > machines. > > > In this case it's pc-i440fx-2.6 and pc-i440fx-2.7, you see the difference > > > which is expected due to instance_id change. > > > > > > You shouldn't see it when comparing same machine types. > > > > No, this is comparing the git tree just before and after the series is > > applied. > I've checked dump_vmstate_json_to_file() implementation and it looks like > it dumps only dc->vmsd enabled devices. > > In this patch vmstate registration has been moved to to apic_comon_realize() > that's why dump_vmstate_json_to_file() doesn't dump apics anymore and you see > the change
So, can anything break because of this difference, or is this just a limitation of -dump-vmstate? -- Eduardo
