Peter Krempa <pkrempa@xxxxxxxxxx> writes: > This is a little off-topic: > > So I wanted to make libvirt use the new parameter to stay ahead > deprecation. I've applied this patch to qemu, dumped capabilities and > pretty much expected a bunch of test cases in libvirt fail as they'd be > using a deprecated field as libvirt is supposed to validate everything. > > And the test suite passed unexpectedly. I've dug further and noticed > that for some reason libvirt doesn't still use JSON parameters for > -chardev (which is the pre-requisite for validation). > > I've also noticed that at some point I attempted to convert it over > witnessed by having an (unused) capability named QEMU_CAPS_CHARDEV_JSON > that I've introduced. > > My questions are: > 1) Does '-chardev' accept JSON identical to 'chardev-add' QMP command? Sadly, no. How badly do you want it? > If yes: If we implemented it: > 2) Since when can that be used? (What can I use as a witness) I figure we'd provide a witness the same way we did when we added JSON support to -device: add a feature @json-cli to chardev-add. > 3) Are there any gotchas? Not aware of any. Can't be 100% sure until we try. > I wonder this as I'd love to finish that out, but I really don't fancy > digging into qemu to find a gotcha 3/4 of the way there. Understandable :) > Anyways, as I've already stated, this patch is okay for libvirt, but I > didn't review the implementation, thus, on behalf of libvirt: > > ACKed-by: Peter Krempa <pkrempa@xxxxxxxxxx> Thanks!
