Next up is the big one: usb_audio. It is well past the 1,000 line cap, so per REVIEW.md §2 I have opened an issue with a splitting plan rather than just dropping a huge PR on you:
https://github.com/RenkuOS/Source/issues/34
Short version: the module goes from 5,184 lines to 9,702, about +4,700/−180 across the same 12 files, no new ones. That is 36 commits of work, and it covers four of my UPSTREAM findings plus UAC2 support. I do not think it splits usefully. The natural seams leave you with middle stages that compile and boot but do not work as a driver, so nobody would have run them, and bisecting them would tell you nothing. What I have real hardware evidence for is the whole module, on the DN-HC4500, DN-HD2500, RMX2 and DDJ-SR over UHCI, EHCI and xHCI.
So I am asking for one PR with oversize-approved. If you would rather have it split two ways I will do that instead, it is in the issue. One practical note: that label does not exist in the repo yet, only the stock GitHub ones, and lint.yml looks for it by name, so someone with write access will need to create it or CI stays red no matter what we agree.
midi_server is the other oversize one later on. I will raise that separately with an RFC first, but it may be worth deciding both at once.
No rush from my end, I know everyone has weekday jobs and houses to move into. In the meantime there is still small stuff I can send that stays under the cap.
Also, @rainygirl , I went through your VAIO P PRs and left comments on #26, #27, #29 and #33 since they land in USB code I have been living in. Nothing that changes the approach, mostly edge cases in the give-up paths.
