06:48 orowith2os[m]: question: how bad of an idea would it be to set a super specific acceleration curve for libinput
06:48 orowith2os[m]: like, I'm talking a step of 0.1, and probably close to 50 or 100 points
06:49 orowith2os[m]: if not even more specific
06:55 vyivel: should be fine
06:56 vyivel: (i think)
06:57 orowith2os[m]: vyivel: time to screw around and find out!
06:58 orowith2os[m]: (I would use more, uh, descriptive language, but trying to be family friendly. I used enough of my sailor mouth at work today)
06:58 vyivel: are you trying to make an actually curvy curve
07:00 orowith2os[m]: perhaps
07:01 orowith2os[m]: oh god. imagine a sine wave for the curve.
07:01 vyivel: peak ux
07:02 orowith2os[m]: I need an app to generate and visualize acceleration curves for this....
07:02 orowith2os[m]: good first time for Python I guess.
07:03 orowith2os[m]: I'll play around with GTK and see if I can make something that plots it out
07:04 orowith2os[m]: erhm, later.
07:04 orowith2os[m]: something something gnuplot
07:06 orowith2os[m]: I think my biggest concern right now is making sure I set the libinput config on x11 correctly...
07:07 orowith2os[m]: `change_property (settings, device, "libinput Accel Custom Motion Step", XA_STRING, 8, step_as_str->str, step_as_str->len);`
07:07 orowith2os[m]: this scares me
08:34 orowith2os[m]: vyivel: so, as it turns out. We can have a max of 64 points on the curve
08:34 orowith2os[m]: https://gitlab.freedesktop.org/libinput/libinput/-/blob/main/src/libinput-private.h#L267
08:34 orowith2os[m]: damn you, yinon
08:35 orowith2os[m]: also curious to know if npoints_min could be changed to 1, and it just level it out (a flat acceleration profile) if it ever hits that condition
08:36 vyivel: the code in filter-custom.c assumes there are at least 2 points, the special case isn't worth it
08:41 orowith2os[m]: fair enough
08:41 orowith2os[m]: I did file an issue for a higher NPOINTS_MAX though: https://gitlab.freedesktop.org/libinput/libinput/-/issues/1094
08:43 orowith2os[m]: with a low step and high amount of points, you could hit the limits
09:15 whot: orowith2os[m]: beyond "nicer visualisation" I doubt you'd notice a difference with >64 points
09:16 orowith2os[m]: yeah
09:16 whot: this very much sounds like the ikea effect :)
09:17 orowith2os[m]: feel free to close as not planned, we can always wait and see if someone ever hits those limits and would give a real world case to justify raising the limit
09:17 orowith2os[m]: I also hope my Mutter patches caught your eye? ;)
09:17 orowith2os[m]: you might want to scrutinize the X11 implementation over there
09:17 whot: orowith2os[m]: they did, I had the tab open all day! and probably will for another few days :)
09:18 orowith2os[m]: hehe
09:18 orowith2os[m]: I think there's a good chance we could eek it into 48, considering there isn't any UI changes involved
09:18 orowith2os[m]: or backport it to other versions
09:18 whot: what's the deadline for that?
09:19 orowith2os[m]: I think, uh, Soon
09:19 whot: like days or a week?
09:19 orowith2os[m]: beginning of March, looks like
09:19 whot: curse the short february
09:19 orowith2os[m]: March 19th
09:19 whot: thats the release date or the deadline?
09:20 orowith2os[m]: release date for the final stable release
09:20 whot: ah, so a tad sooner then
09:20 orowith2os[m]: yeahhhh
09:20 orowith2os[m]: it's not a worry for me. distributions or users will patch it in if they want to
09:20 orowith2os[m]: so long as gnome doesn't NACK the idea of backporting the patches to 47 or earlier
09:22 orowith2os[m]: I also hope seeing the process of me figuring out how to write glib-flavored C and slowly get to a more fully working implementation was fun
09:22 orowith2os[m]: whew boy, 60 more files to build in webkitgtk (well, 120, since it's doing two different ABIs) until I can test the X11 stuff...
09:23 orowith2os[m]: nix is weird.
09:24 orowith2os[m]: just waiting on a solution for https://github.com/NixOS/nixpkgs/issues/384705 so I can test it more, and faster
09:31 MrCooper: mutter 48.rc is cut at the end of this week, after that it's bug fixes only
09:32 orowith2os[m]: MrCooper: so if I want it in then, gotta go fast, gotcha
09:32 orowith2os[m]: I'd have to request a string break, right?
09:32 orowith2os[m]: string freeze break*
09:40 davidre:has bad experience wit rushing features
09:57 orowith2os[m]: whot: so, uh.... in an X11 session right now
09:57 whot: my condolences
09:57 orowith2os[m]: libinput accels no worky
09:58 orowith2os[m]: xinput list-props confirms it
09:58 orowith2os[m]: I probably messed up with the mutter stuff
09:58 orowith2os[m]: if you have any pointers for X11
09:58 whot: does the property change if you set it via mutter?
09:58 orowith2os[m]: that's what I'm doing right now
09:59 orowith2os[m]: makes me think I'm using change_property wrong
09:59 orowith2os[m]: ah, journalctl my beloved
09:59 orowith2os[m]: XIChangeProperty failed on device 11 property "libinput Accel Profile Enabled" with X error 2
10:00 orowith2os[m]: omg I can't even switch workspaces using the touchpad gestures with X11. I'm so glad I never touch it
10:01 orowith2os[m]: wait. I forgot a break statement.
10:01 orowith2os[m]: I'm an idiot
10:01 orowith2os[m]: let me try that
10:02 orowith2os[m]: it might've been that it was falling through and attempting to set custom profiles or something
10:02 orowith2os[m]: all without the custom profile actually being enabled
10:13 orowith2os[m]: whot: would you be available in the gnome-shell matrix to hash it out a bit more?
10:14 whot: looking
12:12 wlb: weston Issue #997 opened by Link Mauve (linkmauve) Weston main fails to build due to LED_COMPOSE https://gitlab.freedesktop.org/wayland/weston/-/issues/997
12:16 mdb977: Hi, I'm on BeagleBoneBlack with Weston 13.0.3. I have a problem with my physical USB keyboard. If I connect the keyboard before starting weston-editor, everything will work fine. If I connect the keyboard after starting weston-editor, it does not work until I restart the editor. What's the best way to debug this? The log does not show anything obvious to me.
13:31 pq: mdb977, a wild guess is that weston-editor (window.c more like) just doesn't understand the keyboard capability appearing later.
13:35 pq: could be the Weston implementation of text-input-unstable-v1 not handling it
13:37 pq: mdb977, is weston-editor really the program you need?
13:46 mdb977: pq, I do not need weston-editor, but my Qt6 app behaves the same
13:48 pq: mdb977, ok, might be better to concentrate on the case with the Qt6 app.
13:49 pq: weston-editor might not raise as much interest
13:49 pq: personally I have no idea
14:34 bl4ckb0ne: should we clean LLM answers? https://gitlab.freedesktop.org/wayland/wayland/-/merge_requests/455#note_2794549
14:35 kennylevinsen: what in the actual
14:36 kchibisov: I'd rather ban the person all together.
14:37 kchibisov: We had the same thing happened the other day to us, where someone wanted to promote their AI by replying to issue report.
14:37 kennylevinsen:would like to change to a different timeline, he doesn't like this one anymore
14:39 karolherbst: bl4ckb0ne: I think you linked the wrong reply, didn't you?
14:39 karolherbst: probably wanted note_2794341
14:39 bl4ckb0ne: woops, its indeed the wrong reply
14:39 bl4ckb0ne: https://gitlab.freedesktop.org/wayland/wayland/-/merge_requests/455#note_2794341
14:39 pq: yeah, I was scratching my head here
14:40 bl4ckb0ne: iirc that user made 2 LLM issues on wlroots and a bunch of spam comments on other MRs
14:40 karolherbst: I think the person in question uses LLM for translation purposes, but might overdo it a little... might be best for you to have a chat with them to check what's up there
14:41 bl4ckb0ne: yup, it doesnt look like a regular spambot
14:41 karolherbst: but in the end if you want to forbid LLMs for technical contributions that might be reasonable if it's only wasting maintainers time, but it's a tough call to make if the person in question doesn't understand english
14:41 karolherbst: or not well enough
14:41 daniels: I'm handling it
14:42 kchibisov: The point is that they are used in a way to confuse people in the current state of things.
14:42 bl4ckb0ne: thanks
14:42 daniels: I saw that on the weekend but forgot it
14:42 karolherbst: thanks
14:42 daniels: I think that person is genuinely enthusiastic, but as karolherbst is getting at, is quite possibly using LLMs to better express themselves in a non-native language
14:42 bl4ckb0ne: i have the same feel
14:43 kennylevinsen: this isn't my experience of how LLM's handle translation requests, so it seems more like they told it to try to make a review or create constructive commentary for them
14:44 daniels: oh yeah, for sure they've also asked the LLM to make a compelling point for them
14:44 karolherbst: well maybe "translation" is the wrong term here, they might just try to overcome language barriers
14:45 karolherbst: anyway, I did check out that account a while ago and at least from a quick look it seems plausible that they might not be able to communicate properly in English.
14:46 kennylevinsen: but then again, who can? :)
14:47 karolherbst: not the point I was trying to make
14:51 kennylevinsen: nah, but as someone using technical assistance to navigate other (non-english) languages, I don't think it's an excuse. You will know that you have just generated an essay on a subject where your only actual opinion is "can I haz plz".
14:52 kennylevinsen: but it's fine to reach out and give them a heads up instead of insta-banning, as they might not fully have realized just how annoying and spammy such comments are
21:11 orowith2os[m]: Uhm..... could libinput expose LIBINPUT_ACCEL_NPOINTS_{MIN, MAX} publicly?
21:11 orowith2os[m]: MAX, at least
21:11 Guest9726: What is the context for this?
21:12 orowith2os[m]: I filed https://gitlab.freedesktop.org/libinput/libinput/-/issues/1094 but it was closed by whot for being a niche use case that nobody's hit yet (no complaints there), but I also want to keep in mind if that does change, having mutter Just Work to support the larger NPOINTS_MAX value
21:14 orowith2os[m]: whot: your choice here. I'll just change my mutter patches to hardcode 64 for now