[libvirt PATCH v4 1/3] add QEMU_CAPS_TCG_DISABLED and probe conditionally

[Peter Krempa]

On Tue, Apr 28, 2020 at 15:27:18 +0200, Michal Privoznik wrote:
> On 4/28/20 1:19 PM, Peter Krempa wrote:
> > On Tue, Apr 28, 2020 at 13:13:32 +0200, Ján Tomko wrote:
> > > On a Tuesday in 2020, Michal Privoznik wrote:
> > > > On 4/27/20 10:22 PM, tobin wrote:
> > > > 
> > > > > Yeah fine with me. Thank You.
> > > > > 
> > > > > When it's a positive capability, you don't even need
> > > > > virQEMUCapsProbeQMPTCGState,
> > > > > you can just add the capability to virQEMUCapsObjectTypes.
> > > > > 
> > > > 
> > > > Yep. I've went with that. This is now pushed.
> > > > 
> > > 
> > > Umm, how do you know then if the capability is not missing because the
> > > QEMU is too old to support it?
> > 
> > Yeah, when inverting it, the capability should be assumed by a version
> > check (yuck) with old qemu.
> > 
> 
> Yeah, the qemu commit in question is v2.10.0-rc0~93^2~18 and before that it
> wasn't possible to compile out TCG. And what do you mean "too old to support
> TCG"? Isn't TCG how QEMU started (with adaptation to KVM happening later)?
> The oldest mention of TCG that I bothered to find is in
> v0.14.0-rc0-936-g303d4e865b which is way older than current minimal 1.5.0 so
> I think we are safe, aren't we? Or is there something that I am missing?

No, this is the artifact of you taking patches and modifying them and me
not checking the pushed code.

You correctly added that prior to 2.10 the capability is assumed. I was
refering to the need to have such a thing if you want to do a positive
capability to prevent regressing with old qemus.