#asahi-dev · 2023-03-19 14:16
marcankettenis: it is not, this is why they have separate tables per model too, which is yuck#asahi-dev · 2023-03-19 11:33
marcanthe keyboard thing did but the m1n1 bump is backwards-compatible (and I don't know if you even use the /chosen kblang node yet?)#asahi-dev · 2023-03-19 11:32
marcanpurely a linux thing#asahi-dev · 2023-03-19 11:32
marcannope#asahi-dev · 2023-03-19 11:31
marcanguess that needs some cleanup#asahi-dev · 2023-03-19 11:31
marcan(in a lot of codepaths)#asahi-dev · 2023-03-19 11:30
marcanjannau: heh, why are we always calling fdt_get_alias then?#asahi-dev · 2023-03-19 11:01
marcansounds like an issue with initial sync between the compositor and DCP? but only when buffers first get created (wouldn't be surprised if sleep uses an extra dummy buffer or something)#asahi-dev · 2023-03-19 11:01
marcanyeah, I get it fairly consistently on first sleep too, right when the screen goes black (but never subsequent sleeps)#asahi-dev · 2023-03-19 10:58
marcanyes#asahi-dev · 2023-03-19 10:56
marcanbut never randomly in use or anything#asahi-dev · 2023-03-19 10:55
marcanit's only on startup right? and I also noticed it on first sleep/resume#asahi-dev · 2023-03-19 10:55
marcanI did notice the "pink squares" issue on kwin, but I guess we'll call that a known regression until either lina tracks it down or someone finds a worse repro#asahi-dev · 2023-03-19 10:54
marcanif nobody has any objections let's push this out tomorrow or something (cc lina, you get to write the blog post)#asahi-dev · 2023-03-19 10:54
marcaneverything is on asahi-dev now#asahi-dev · 2023-03-19 10:51
marcanjannau: seems to work for me#asahi-dev · 2023-03-19 10:41
marcanit almost smells similar to the CS42L84 race I fixed that was breaking everything#asahi-dev · 2023-03-19 10:41
marcanI think some altdistro folks have reported problems with that for a while now (i2c lockups), but I have no idea how to repro it#asahi-dev · 2023-03-19 10:40
marcanlet me test that#asahi-dev · 2023-03-19 10:31
marcanjannau, sven, kettenis, povik: any release blockers at this point?#asahi-dev · 2023-03-19 10:28
marcanpushed asahi-6.2-11 which should fix the M2 keyboard country code silliness (alongside new m1n1) and pulls in some fixes from lina#asahi-dev · 2023-03-18 02:42
marcancool, I'll go with that#asahi-dev · 2023-03-18 02:38
marcanand leave it to userspace to figure out what to do with all that, or we can make the driver just mirror them into /sys#asahi-dev · 2023-03-18 02:38
marcanboth in the keyboard node (which does exist on the MTP machines, and we can add something for the SPI machines)#asahi-dev · 2023-03-18 02:37
marcan(the former is based on the latter according to a lookup table, but I'd rather have that in the firmware and just provide both in the DT)#asahi-dev · 2023-03-18 02:37
marcanrobher: how about `apple,hid-country-code` for the HID country bit and `apple,kblayout-id` for the actual specific Apple layout ID?#asahi-dev · 2023-03-18 02:35
marcanjannau: I think that's around the limit, a bit more and you start getting panics on boot policy mutations (due to timeouts, I think)#asahi-dev · 2023-03-17 08:44
marcan_jannau_: we do have a DT keyboard node on MTP machines#asahi-dev · 2023-03-17 08:31
marcanand we still want something for the HID country-code, though that is mostly an issue for MTP (since it's not a full HID device and we have to construct the country code), so if you're happy with it I can just use apple,hid-country-code#asahi-dev · 2023-03-17 08:30
marcanrobher: Where do we put that identifier then? /chosen as I do now?#asahi-dev · 2023-03-15 05:47
marcankblang it is#asahi-dev · 2023-03-15 05:46
marcanyeah I don't think it's in there, meh#asahi-dev · 2023-03-15 05:27
marcanand there's more than one table, of course#asahi-dev · 2023-03-15 05:27
marcanit's actually just in AppleTopCase.kext/Contents/PlugIns/AppleTopCaseDriverV2.kext/Contents/Info.plist#asahi-dev · 2023-03-15 05:27
marcanso we get to do the same#asahi-dev · 2023-03-15 05:27
marcannevermind, macOS just has a table of the kblang -> everything else#asahi-dev · 2023-03-15 04:55
marcanjannau: can you run experiments/mtp.py and paste the output? I think I might have found the real keyboard type.#asahi-dev · 2023-03-15 04:41
marcanokay, that's one set of "what if we need more data" questions answered#asahi-dev · 2023-03-15 04:40
marcanno other info#asahi-dev · 2023-03-15 04:40
marcanso this is really just the language code#asahi-dev · 2023-03-15 04:40
marcanand the first byte is the version#asahi-dev · 2023-03-15 04:40
marcanokay, that eliminates a lot of mystery.#asahi-dev · 2023-03-15 04:39
marcanjannau: oh lol, kblang-calibration is just 2 bytes and then a CRC32 of those 2 bytes.#asahi-dev · 2023-03-15 04:32
marcanright now we have that in /chosen which is probably also suboptimal (I imagine it should go in the HID device node)#asahi-dev · 2023-03-15 04:31
marcanalternatively we could just punt on standardizing any of this and keep doing what we do now, which is just shove the Apple-specific enum code somewhere and have userspace handle mapping that, but that's... maybe not ideal.#asahi-dev · 2023-03-15 04:30
marcanif we go with outright putting xkb type/model/options into the DT, that would nicely solve the problem but creates an interesting spec dependency with the xkb stuff (and that isn't as strictly managed as DT bindings are re backwards compat etc)#asahi-dev · 2023-03-15 04:29
marcanhow should this be expressed in the device tree? if we go with a country code type thing, that's lossy and userspace then still needs at least per-vendor tables (and Apple sometimes has more than one layout for a given country and not-technically-country-specific layouts, which is an issue).#asahi-dev · 2023-03-15 04:28
marcanoutput: userspace eventually needs to know what xkb layout to use (and the equivalent best-fit console keymap, though console keymaps are pretty hit and miss anyway)#asahi-dev · 2023-03-15 04:28
marcaninput: our bootloader knows the layout from system firmware, as a proprietary enum#asahi-dev · 2023-03-15 04:27
marcan- so then we need something to specify the *actual* layout, in detail, in the device tree