10:44naemi: pq: Does the cursor change need to be managed directly from said application? There is an application running as part of the script
10:56pq: naemi, yes. And the cursor image will only apply while the pointer is on the app window.
10:56naemi: hrm ok
10:56naemi: thanks for the info
10:56pq: naemi, are you sure cursor image is a good design, and not something like e.g. notifications?
10:57naemi: it'd be the best imo
10:57naemi: guess I'll use notifications instead
10:57pq: the user can always move the pointer away from the window, and then your cursor image is removed.
10:58pq: so it won't be a global'ish indicator
10:58pq: I understood you wanted someething gloabl'ish when you talked about forcing the cursor image.
11:00pq: It would be strange for an application to not control its own cursor image.
11:00naemi: yeah that was the idea
11:01naemi: (it's a password manager auto-typer, I need an indication that it's writing)
11:02naemi: I don't mind if it's just for a specific window if I can set it with high-level tools (not code)
11:03pq: no, not with tools. It needs to come through the same Wayland display connection as the app is using.
11:04pq: I'd expect the typing to be an even bigger problem.
11:10pq: Wayland is fundamentally not scriptable. You need a compositor designed for scripting things, that's not provided by Wayland.
11:13kennylevinsen: naemi: the problem would be that the specific window is that it would only be the password manager's own windows, not the window of the "target" so to speak
11:13kennylevinsen: ... well that became word salad, "the problem is that the specific window would only be..."
11:13naemi: I understood :)
11:14naemi: bit of a shame, this worked fine in x11 and I got really used to it
11:14pq: naemi, how do you do the "typing in" part?
11:14naemi: dotool
11:14pq: what does that do?
11:14kennylevinsen: it's a uinput thing :/
11:15naemi: yup
11:15naemi: https://sr.ht/~geb/dotool/
11:15pq: aha, that won't work
11:15pq: you don't know if the input device created by uinput will be used by the compositor (maybe it goes to a wrong physical seat), and you don't know what keymap you should use to translate characters into key codes.
11:16pq: you can get lucky though, but it's far from a good idea
11:16pq: you also cannot control which Wayland window will receive the typing
11:17pq: I guess making the password manager an input method could work? Or was there something more specific?
11:18pq: aside from direct compositor integration
11:19pq: copy & paste would work, if you allow the user to click on the password manager window to copy, and then explicitly paste into the target app
11:20pq: wasn't there some kind of password manager d-bus framework even?
11:20naemi: I can't use the clipboard for this, as one of the reasons I use dotool/uinput is because I need to type into software that has no clipboard too (e.g. spice remote)
11:21pq: aha, that complicates things
11:22naemi: I'll just send a short notification or play a sound whenever it's done typing, doesn't seem like this is solveable with cursor
11:22pq: right
11:24kennylevinsen: note: spice has clipboard support, but it requires the agent. Also no guarantee if those clients are well-behaved - some try to copy in the background for example.
11:25pq: I guess you also limit your passwords into 7-bit ascii and use only qwerty layout to avoid most of the keymap problems :-)
11:27kennylevinsen: 7 bits for characters?! anything past 6 bits is just throwing resources away!
11:28naemi: I can't depend on an agent, and my workflow has worked for more than a decade without issue, it's just the cursor thing I missed from x11
11:29kennylevinsen: you do you, just saying that clipboard can be fixed for spice for most use-cases :)
11:29kennylevinsen: emphasis on most, rather than all
11:29naemi: for normal operation usecases yes, but there are a lot of times where an agent can't be run or simply doesn't exist
11:30naemi: like when installing a new VM, or connecting to some arcane system
11:30naemi: and that is a large portion of the VMs I connect to
11:30kennylevinsen: well, if that OS is either windows, linux or freebsd it should be perfectly viable - need for password managed credentials usually come after installing some packages anyway
11:31kennylevinsen: but if it's plan9, netbsd or some arcane windows 3.11 setup for weird outdated control software, then maybe not
11:31pq: It's as if "paste as keyboard events" would be a useful feature in a VM viewer GUI.
11:31kennylevinsen: but possible is one thing, practical another
11:31naemi: connecting to stuff (proxy, network, repo) during installation that require passowords can't be solved with an agent
11:32naemi: it's just one usecase anyways, there are other situations where the clipboard is not a solution
11:32kennylevinsen: naemi: if network access to install qemu-agent requires a password, then yeah there's a gap for that password
11:33kennylevinsen: pq: hmm, not the worst of ideas
11:33pq: "paste as keyboard events" could be even a compositor feature...
11:33naemi: there are systems I have to use daily which explicitly prohibits clipboard anyways, so I can't use it as a one-stop solution
11:34kennylevinsen: Hah I was just about to say that :)
11:34kennylevinsen: naemi: heh, nothing like IT policies that actively enforce *bad security practices* like disallowing clipboards :)
11:35pq: Wayland clipboard is not like X11 free-for-all-clipboard
11:35kennylevinsen: but tbf, danish government login tries to disallow clipboards too, password managers just usually hack around it...
11:42naemi: in my experience nothing will prevent uinput from typing for me, so that's what I use
11:43pq: just never switch to dvorak
13:59krathul[m]: How does libinput convert the relative deltas in mouse movement from evdev to relative movement on the screen? Is there any transformation matrix?
14:00pq: krathul[m], maybe start from https://wayland.freedesktop.org/libinput/doc/latest/configuration.html#pointer-acceleration
14:03pq: krathul[m], libinput is configured by each compositor at runtime, and has no configuration files of its own. So the conversion varies over user to user and mouse to mouse.
14:06pq: krathul[m], libinput also has APIs for accelerated and non-accelerated motion. Callers choose which one they need.
15:00zetaE: greets
15:00zetaE: one major complain about from x11 users is lack of xhost
15:00zetaE: what's wayland contingency plan in this regards?
15:02daniels: for local use, bind-mount the UNIX socket; for remote use, use waypipe or RDP or any other remote solution that fits your needs
15:11zetaE: daniels: I know
15:11zetaE: hasn't stop me from trying
15:54Momentum: anyone knows where i can find documentation for what the `data` field is used for on various `wlr_*` structures?
15:54bl4ckb0ne: thats a question for #wlroots on libera
15:54Momentum: oups
15:54Momentum: i thought i'm on #wlroots
15:54Momentum: my bad