Re: [PATCH v2 18/25] hw/sd: ssi-sd: Bump up version ids of VMStateDescription

[Bin Meng]

On Mon, Jan 25, 2021 at 6:41 PM Dr. David Alan Gilbert
<dgilbert@redhat.com> wrote:
>
> * Bin Meng (bmeng.cn@gmail.com) wrote:
> > On Mon, Jan 25, 2021 at 12:59 AM Philippe Mathieu-Daudé <f4bug@amsat.org> 
> > wrote:
> > >
> > > On 1/23/21 11:40 AM, Bin Meng wrote:
> > > > From: Bin Meng <bin.meng@windriver.com>
> > > >
> > > > With all these fixes and improvements, there is no way for the
> > > > VMStateDescription to keep backward compatibility. We will have
> > > > to bump up version ids.
> > >
> > > Unfortunately this breaks bisectability (think about downstream
> > > distributions cherry-picking patches individually).
> > >
> > > I don't think there is a problem increasing 2 -> 3 -> 4 -> 5
> > > (Cc'ed David in case). Could you respin increasing the version
> > > on each VMState change?
> > >
> >
> > I definitely could be wrong, the reason I posted a single patch to
> > upreve the version is that, I was under an impression that in each big
> > release (like here 5.2.0 -> 6.0.0), the incompatibility version id
> > should be bumped up once.
> > It does not look correct to me that in a big release we bump up the
> > version id for 10 times.
>
> I think I agree; I don't think we've ever done it incrementally like
> that before.
>

Thanks David.

> It would only break bisectability if you were cross-version migrating
> during the bisect which is rare.
>
> > Since this is a series to fix issues in the ssi-sd, I don't think it's
> > practical for downstream to just cherry-pick some commits while
> > leaving some other commits there.
>
> Never underestimate downstream :-)
> However, please add a comment when you're doing incrimentals like this -
> e.g. a TODO or something showing that it's unfinished and you need the
> remaining patches so people don't do it accidentally.
>

Sure, next time :)

Philippe, I guess we will need to hold on your PR?
http://patchwork.ozlabs.org/project/qemu-devel/list/?series=226135

Regards,
Bin