Matthew GarrettEvery single ACPI vs Device Tree argument needs to start with the observation that I can boot a modern Linux kernel on an arbitrary x86 board from 1998 and it will probably suspend and resume correctly, and I can't do that with an arbitrary Arm board from 2026
CounterPillow@mjg59 if you're going to subtweet my LWN comment then at least you should've read it. But feel free to buy a Qualcomm SoC, it should just work after all since it's ACPI, right?
@CounterPillow Not meaningfully
@mjg59 Ah, so they're just doing ACPI wrong! No true Scotsman would fail to boot.
Thank you for your kind directions from your armchair, here's your Reddit-V Gold. Now I can finally stop writing drivers, since ACPI means I will no longer need those after all.
@CounterPillow as someone who's spent a shitload of time working on ACPI issues and who's deployed DT-based devices (including on x86!) and who's worked on hardware support for both consumer and enterprise distros, I will politely suggest that I have broader expertise in this domain than you
@mjg59 probably, but I'm still not dumbfuck enough to think forcing me to work around pre-flashed vendor ACPI firmware is going to mean you get to meaningfully use any OS on your PoopPi 4 without any upstreaming of drivers needed.
Though I guess I'll already be forced to talk to "AI agents" so what difference does it make, shovel the turds right in my mouth.
@mjg59 whole industry is a fucking joke and people who get tech opinions from YouTube like the ACPI pushers and "RISC-V will make my CPU open source!!!" troglodytes are at the centre of it.
@CounterPillow yeah I'm definitely getting my opinions from YouTube rather than having been employed to work on mainline ACPI support. Maybe the problem here is you?
@mjg59 I'm not talking about you specifically but the commenter I was replying to. And nobody seems to address the thing about drivers, which apparently magically appear.
@CounterPillow yes because the point is that the non-embedded ecosystem isn't a shit show of NIH and basic platform functionality can be assumed, and what weirdness there is can be described by ACPI. Removing ACPI from x86 and whatever the Arm server spec is called this week would require more drivers and more platform knowledge and the world would be a worse place. Obviously sticking ACPI into a space that doesn't have those characteristics won't work.
@mjg59 Okay phew so we are in agreement then, good talk, sorry for the tone.