Re: [PATCH 1/3] hw/intc/arm_gicv3: Move checking of redist-region-count to arm_gicv3_common_realize

[Peter Maydell]

On Thu, 30 Sept 2021 at 22:54, Philippe Mathieu-Daudé <f4bug@amsat.org> wrote:
> The pattern make me think gicv3_init_irqs_and_mmio() should be
> refactored as a ARMGICv3CommonClass::init_irqs_and_mmio handler,
> called in arm_gicv3_common_realize().

That won't work because the two subclasses want to call it
with different arguments.

More generally, I find QEMU's class method infrastructure
sufficiently clunky that it is better avoided except in
the case where it's actually needed, ie where you have one
callsite that might have to deal with an object whose
exact type may vary such that you don't know which of the
methods you want and you need the dynamic-dispatch.
In this case that doesn't apply: both of the callsites of
gicv3_init_irqs_and_mmio() know exactly what function they
want to be calling (there is only one implementation), and
moreover both callsites are part of the GICv3 implementation
themselves, so this isn't about presenting a nice (or "nice")
external interface to other parts of QEMU.

-- PMM