TR Forums

Arch (what I use) switched X's input driver from evdev to libinput a little while back. I noticed this first because my mouse, which I prefer to have no acceleration, suddenly had the most godawful acceleration I've ever had the misfortune to experience. There was a big (2x?) jump in sensitivity right around the speed that I most commonly move my mouse, meaning I had to make a point of going slower to get anything resembling predictability. I would have just gone back to xf86-input-evdev, but that's apparently going to be depreciated at some point, and I prefer not to accumulate more depreciated weirdness in my system than I have to, so I went looking for how to disable acceleration with xf86-input-libinput. It turns out to be packed so full of WTFs, the developers of Gnome could be proud.

First up, they pushed responsibility for config out to developers of desktop environments. Libinput has zero direct configurability - it needs to be handled by whatever's calling it. When that's X, things are basically alright and predictable, but if Wayland is taking over the world someday, that means we're dependent on DE developers to both take the time to handle this stuff and not screw it up like DE developers so often do.

Second, even in the best of cases, there isn't much config. Look at the bug report for permanent acceleration. Hutterer *really* didn't want to implement that feature, and (here too) looks like he still doesn't really believe in it or understand it. I use 400 cpi with no accel, and on a 1080p monitor, "excruciatingly slow" is only the right description if you mouse with your wrist only.

I get a terribly "we know what you want better than you know what you want" vibe from this guy anyway, but then he says "I'm also somewhat tempted to only enable the flat profile for devices we know have switchable DPI profiles, but I'm not sure that's worth the effort." Dude? I, the user, am going out of my way to set an option to make things behave in the only way I find sane, and you would prevent me from doing that because you don't believe I really know what I'm doing? The only words coming to mind here are not suitable for the non-R&P parts of this forum.

Anyway, I'm currently using xf86-input-libinput with the flat acceleration profile. It isn't quite right, and I can't figure out how. By moving the mouse fast one way and slowly the other repeatedly, I can't find any actual acceleration. I tried all three polling rates (125/500/1000 Hz) and verified they were actually working, but the wrongness apparently has nothing to do with that. It makes nice smooth loops in mtpaint - there was a touch of jitter, but it was associated with 1000 Hz polling (I'm now running 500). All I can figure out is that it used to be pretty much on the money at 400 cpi (within 10 points), but now it's ~375. However, the subjective feel is a touch faster if anything. I have no idea where to go with that one. Any thoughts?

Egads, yet another "it wasn't broke but we 'fixed' it anyway" moment.

Anyhow... if it is really as f**ked up as you say, we can hope that someone will fork it. I imagine someone will also provide a configuration wrapper of some sort.

Do developers actually use the software they develop? I sometimes wonder about that after I hear about the problems some users are having lately.

I was going to suggest a different distro, but it appears that Debian's switching to this for v9.

Crap like this is why "x is the year of Linux on the desktop" is a bad joke.

whm1974 wrote:

Do developers actually use the software they develop? I sometimes wonder about that after I hear about the problems some users are having lately.

Historically, a big part of the PROBLEM with Linux as a general desktop computing solution was that it was aimed TOO MUCH at developers, and not enough at non-developers!

That may even be part of the issue here -- someone who spends 99% of their time cranking out code is going to care less about mouse acceleration than someone for whom the mouse is the primary means of interacting with the system (e.g. graphic design, CAD...)

just brew it! wrote:

whm1974 wrote:

Do developers actually use the software they develop? I sometimes wonder about that after I hear about the problems some users are having lately.

Historically, a big part of the PROBLEM with Linux as a general desktop computing solution was that it was aimed TOO MUCH at developers, and not enough at non-developers!

That may even be part of the issue here -- someone who spends 99% of their time cranking out code is going to care less about mouse acceleration than someone for whom the mouse is the primary means of interacting with the system (e.g. graphic design, CAD...)

Yeah but surely even developers must do other things with their systems besides writing code, right? You know, like actually use the software, such as watch movies, play games, write documents, and etc.

Software often has an intended audience just like those movies and games do.  Two applications built for the same purpose, but targeted at a different base of users may end up being very different applications.

That, and "design by programmer" isn't just a figure of speech.  I often gripe and complain about design mockups I'm given to build, but I'd much rather have someone else build them for me, because there's a pretty good chance they'll do a better job of it than I would. I would imagine the same is true for system apps which don't have as much of a UI per say.

I'm thinking the remaining problem might be input lag. I know I have very little tolerance for that, and there seems to be a bit more of it than before. I don't have the stuff handy to test that objectively, though.