00:00 <apritzel> have you ever checked what it would take to set SYS_SOC to something else than sunxi?

00:00 <smaeul> it wouldn't be too hard to fix up at runtime: find r_intc's phandle, then delete all interrupt-parent properties referencing it (except where the interrupts number matches the interrupts number in r_intc)

00:01 <smaeul> the question is deciding when to apply the fixup

00:02 <smaeul> I tried changing it once, and ran into a giant mess -- that's why I did https://lore.kernel.org/u-boot/20211104035516.28268-5-samuel@sholland.org/

00:02 <smaeul> I couldn't add a "wdt-reboot" node in sunxi-u-boot.dtsi, because the watchdog node labels were different across the SoCs

00:04 <apritzel> yeah, the runtime approach has conceptional issues: when you load grub, you don't know which kernel it will be

00:05 <apritzel> was the r_intc driver change ever backported to stable kernels?

00:08 <smaeul> it doesn't look like it, and I didn't expect it to be

00:08 <apritzel> for SYS_SOC: maybe adding some extra variable for the DT override, defaulting to SYS_SOC?

00:14 <apritzel> another approach I was thinking about is to add something like a user selected overlay, to switch between "bleeding edge" and "compatible" DTs

00:35 <apritzel> smaeul: with just a super trivial fix to drivers/net/sunxi_emac.c I get past buildman