09:14 <vipulogmx> Any workaround to get river in Debian bullseye or i should lose hope for debian as of now and get arch to try river?

09:29 <kraem> hey i'm having a problem with river / libinput-gestures. for my laptop internal touchpad i can define libinput-gestures swipes and they get executed as expected. whenever i have an external touchpad plugged in, the swipe events from the external touchpad seems to not be propagated to libinput-gestures. is there any option i must set on the input with riverctl? i've read the riverctl man pages but

09:29 <kraem> can't find anything obvious

11:17 <ifreund> vipulogmx: If you can get wlroots 0.15.1 on debian bullseye you could then use a static zig tarball from https://ziglang.org/download/ to build river

11:17 <ifreund> I don't know how feasible getting wlroots 0.15.1 is though, I see that it's packaged for debian experimental

11:21 <ifreund> kraem: I'm not sure what's going on there. Have you seen that hardware work with some other wayland compositor?

12:18 <ahmadraniri1994[> <vipulogmx> "Any workaround to get river in..." <- Just switch to debian sid or unstable, I'n using Debian unstable and everything is fine.

12:19 <ahmadraniri1994[> * Just switch to debian sid or unstable, I'm using Debian unstable and everything is fine.

16:40 <novakane> ifreund: I was messing around with a rgba converter function and I'm curious if you would like something like this in river https://zigbin.io/a08d0f

17:00 <ifreund> novakane: I mean yeah the bit-twiddling version almost certainly generates better code but I find the river version easier to maintain and it's far from being performance critical

17:01 <ifreund> we could probably get the best of both worlds though...

17:04 <novakane> sure, it's not really gonna improve perfomance, I kinda like it though, and since I was improving mine I though about sharing it

17:06 <novakane> what would you change for the best of both worlds?

17:15 <ifreund> novakane: check this out https://0x0.st/oPUi.txt

17:21 <ifreund> This would be a bit stricter: var input = try fmt.parseUnsigned(u32, string[2..], 16);

17:32 <novakane> ifreund: oh yeah, that's nice too!

17:32 <novakane> does 0xff.0 is a floats?

17:34 <ifreund> yeah, same as 255.0

17:35 <ifreund> novakane: check it out, llvm auto vectorizes my version: https://godbolt.org/z/h8Wfr8hKM

17:37 <ifreund> the interesting part to compare is .LBB1_18 for your bit shift based approach with LBB0_18 for the bitcast to array approach

17:37 <ifreund> I'd guess the reason it's easier for llvm to autovectorize my code is because I only have a single branch and it's before the mathy bit

17:39 <ifreund> well, a single branch after the std.fmt.parseUnsigned() call I mean

17:40 <ifreund> this was a fun little diversion :)

17:41 <novakane> oh, that's interesting

17:41 <novakane> haha yeah I've been messing with this for too much time, just because it was kinda fun :P

18:10 <kraem> ifreund: not under wayland, but under x with libinput-gestures though. so there's something misconfigured i think.

18:12 <ifreund> kraem: does the external touchpad show up under riverctl list-inputs?

18:12 <kraem> yeah it's everything but gestures are working as expected

18:13 <kraem> s/it's//

18:14 <kraem> i'm running ydotoold as a service and it's somehow registering itself under list-inputs

18:19 <ifreund> kraem: the way ydotool works is that it intercepts input events from the kernel before they reach river

18:19 <ifreund> it exposes virtual input devices for river (or any other wayland compositor) to use

18:21 <kraem> oh alright, makes sense yeah

18:22 <kraem> i guess i need a way to debug exactly what the trackpad is sending

18:22 <ifreund> `libinput debug-events` exists

18:23 <ifreund> I don't really have any experience interperting the output of that though

18:25 <kraem> yeah just found it by searching :) and it's indeed sending swipe events

18:29 <ifreund> I guess the next thing to do would be to debug what libinput-gestures is recieving

18:48 <kraem> yup, so libinput debug-events are getting all the events. but libinput-gestures isn't recognizing anything, not even pointer_motion

18:53 <ifreund> kraem: that's strange, as libinput-gestures seems to be implemented by parsing the output of libinput debug-events

18:55 <ifreund> kraem: perhaps it's a permissions issuee? Are you runnning libinput-gestures as a user in the input group or othewise with permission to use libinput debug-events?

18:56 <ifreund> aside: parsing the output of libinput debug-events seems like an insane way to implement this kind of thing :/

19:04 <novakane> huh that's sounds crazy

19:11 <kraem> ifreund: yeah i've learned that it does just listen to debug-events while reading the readme + issues just now :)

19:12 <kraem> don't think it's a permission issue as i'm debugging it by running it as my user who's in the input group and it's finding the builtin trackpad just fine

19:14 <kraem> what i've found though is that it seems the trackpad registers two devices under /dev/input (event3 and event4). `libinput debug-events` is reporting all the actions from event4, but `libinput-gestures -r` is just reporting the device event3

19:17 <kraem> https://paste.rs/EwU

19:43 <kraem> thanks for rubber-ducking. it was of course an option to set it to listen to specific devices / all devices :)

21:16 <NickH> vipulogmx: I have river packaged for Debian bullseye

21:17 <NickH> Currently the binary packages are just on a private server, but I have uploaded src packages for zig and river to mentors.debian.net

21:19 <NickH> There are a few additonal packages that need to be backported but it's very easy.

21:19 <NickH> I'd be happy to make my packages public but don't really want to host them on my own server.

21:20 <NickH> s/packages/binary packages/

21:25 <NickH> See https://0x0.st/oPGq.txt

21:31 <NickH> Zig package https://0x0.st/oPGc.txt

21:33 <NickH> Maybe I could set up a VPS somewhere...

23:33 <ifreund> kraem: no problem, glad you got it worked out!