Can I finally start using Wayland in 2026?
128 points by stapelberg
128 points by stapelberg
@stapelberg Your font rendering issues stem from a different set of fonts being used on Wayland. Eyeballing it, on X11 they used to be set to Cantarell, on Wayland it's Adwaita Sans I believe.
I think it could be caused by GTK sourcing its config files from a different place under X11, but I'm not sure. I've always used GNOME Tweaks to change GTK font settings, which is the same as editing dconf.
Thanks for the tip! https://mas.to/@whynothugo@fosstodon.org/115836011190535672 also wrote me a few minutes ago:
@zekjur Re font rendering: under Wayland, gtk3 ignores .config/gtk-3.0/settings.ini, and uses dconf exclusively. Is it possible that you have some setting there which Firefox is using in X11 resulting in the discrepancies?
And indeed, I do have such a file:
cat ~/.config/gtk-3.0/settings.ini
[Settings]
gtk-toolbar-style=GTK_TOOLBAR_ICONS
gtk-font-name=Cantarell 11
Will play around with this when I get a chance.
Update: Yep, that was it. Now updated in the article: https://michael.stapelberg.ch/posts/2026-01-04-wayland-sway-in-2026/#font-rendering
I've tried Wayland over the holidays and I switched back to X11 yesterday. I've experienced the double shortcuts thing, the laggyness and some issues that are related to electron. I also had high hopes it would 'just work'.
The double shortcuts thing drove me mad, I didn't know it was a Wayland thing untill I read the post, and cleaned my keyboard and did a reset, but those didn't work.
I've switched (sort of) a while ago because I saw the writing on the wall and figured it's better to deal with the quickly-fixed early adopter bugs than the long tail of all-but-abandoned X11 branches. It's still a mixed bag of weird for me.
On Plasma 6 the color picker is either working or not working depending on the phase of the moon, I guess, and remote desktop has never given me anything other than a black window. I was at least looking forward to a clean slate dropping some of the old cruft (like playing whack-a-mole with the umpteen places where font DPI and anti-aliasing are being set to guess why text is blurry) but I think at this point that's not happening, either. Hence "sort of" -- the Wayland session is my default but I find myself logging out and going back to the X11 session once every couple of days.
I tried Sway for a while -- I'm not a fan of dynamic tiling WMs but i3's model is bearable, and I don't like how sluggish Plasma is (on either Wayland or X11). I ran into the double shortcuts thing right away but I was long past the point where nothing surprised me anymore.
Not gonna lie, to me "can I finally start using Wayland in 2026" is slowly morphing into "am I going to finally pay for Windows in 2027". Like, I agree with pretty much every reason why X11 should be replaced, but if this is what it's being replaced with, I'd rather fast forward through the next ten years or so of FOSS desktops collapsing under their own weight and just get that stupid Surface already.
Thanks for sharing! It’s good to know others are encountering this issue, too. That makes it more likely to be reproducible for th developers who can fix it :)
Nice review of the problems that arise when using Wayland although some seem "self-inflicted" by using uncommon hardware like a nice 8k monitor :P
In any case: I don't think I have noticed most of these issues in my standard Fedora 43 KDE Plasma setup, except for the laggy Emacs+pgtk issue... and I'm glad to see it wasn't me imagining things. Emacs feels OK on my laptop screen, but when I use my external 4K display and try to run Emacs in it full-screen, the lag is unbearable. I even resort to running Emacs on a full-screen Konsole terminal because of this, which is sad... (although being used to running Emacs on the terminal means I don't care much for "X11 forwarding", which is another limitation highlighted in the article about the transition to Wayland).
although some seem "self-inflicted" by using uncommon hardware like a nice 8k monitor :P
I've been using sway/wayland for more than 5 years now and I would quit using it and go back to i3 if I could self inflict that 8K monitor on myself, assuming I couldn't make it work with sway/wayland.
I don't know why Emacs+pgtk is slow, but it is. All my displays are low to moderate resolution, so I never really see lag as such, it does feel like input latency is too high. But the thing that gets me is that if an Emacs+pgtk window is open on any virtual desktop (Sway, Niri, Gnome), one CPU is always showing high usage, from Emacs, even if Emacs is completely idle. I think this might have something to do with Emacs and GTK both having their own event loops, but I don't really know enough to say for sure. These days, I run Emacs under XWayland mainly.
I also had the same thought about most of the problems being self-inflicted - but what a problem to self-inflict!
Fractional scaling needs to be seen as an accessibility issue and prioritized accordingly.
I'm visually impaired so I really rely on 150% scaling on my normal sized monitor. Pop!_OS with Cosmic is supposed to be past alpha now, but I still have to jump through tediously researched hoops for each Electron app, one by one.
Gnome has an experimental setting for fractional scaling that ive been using over a year now with no issue. It applies globally to all apps. Theres an 'xwayland-native-scaling' feature now too for x11 clients. Recently they made it quite a bit sharper too.
No Xwayland scaling: programs started via Xwayland are blurry (by default) or double-scaled (when setting Xft.dpi: 288).
Yeah, this is really annoying. I use my https://github.com/talex5/wayland-proxy-virtwl proxy to undo the Wayland scaling (running with --x-unscale=2) as a work-around. However, it does also require patching wlroots so that it doesn't reject windows where the size isn't a multiple of 2 (wlroots is following the Wayland spec, but the spec is silly).
I switched very recently as well for no particular reason and my experience was also similarly unfruitful.
--release mode which seems to execute after "p" is released but before "Super" and "shift" are released sometimes. I did not face this with i3.Probably a few other things I'm forgetting right now.
Synergy/barrier
I had a good experience with rkvm[1] recently (except for it's documentation maybe). Don't remember if Synergy/barrier has additional features over input device sharing, but if you need only that I think it's a nice solution. Afaik it creates it's own uinput device and doesn't care about what display server is running.
like so many of these projects rkvm seems to be abandoned too.
It just uses kernel interfaces though, and hopefully they are considered fully userspace in the sense of not breaking userspace, so it will probably keep working for years without changes anyway.
I moved to i3 in preparation for an eventual (forced) Wayland move. Looks like it's still a ways off being ready.
Very much depends on your hardware and your desired workflow. On the flip side, I’ve been using Wayland exclusively and extensively since 2018, and have literally never looked back. There were growing pains and quirks early on, but from 2020-21 ish onwards I have been able to adamantly hold that having to use X would be an active downgrade. Doubly so starting in 2025 when I moved to Niri - my only complaint is the lack of per-keyboard layout configuration.
This is even with some weird hardware: my monitor is a Dell U4025QW, a 5k2k ultra wide running at 120Hz (VRR down to 48 as needed while gaming). Not as many pixels as the 8k the author has to push, but still weird and high refresh rate.
I will second this. The bad reports of Wayland and Sway that I've heard over the years scared me away from Wayland for a while, but I've been using Wayland for a few months now, and other than the afore-mentioned screen-sharing issues I haven't had any real problems. The only real way to know if it'll work for you is to give it a shot, I think.
Which GPU and driver are you driving your 5k2k display with?
Both the on-board iGPU in my AMD 7900 CPU as well as my dGPU (AMD 7900GRE) both push it just fine using the stock in-kernel driver bits + any modern Mesa, over DisplayPort at least. This includes doing weird stuff like plugging my displays into the motherboard (so using iGPU most of the time for productivity tasks, getting nice low power draw) but having the dGPU headlesslly render the frames of intensive tasks (eg. games).
AFAIK HDMI will have issues due to the lack of any open-source HDMI 2.1 implementations, because HDMI is a complicated, proprietary, behind-closed-doors-licensing-agreements, beast.
It seems like most of the problems reported in this article stem from either nvidia or Chrome. I've been happy on Wayland for like a decade at this point.
I’ve been a happy Wayland user for five years now. Though I have a very different setup of programs I use and needs. I use Firefox and a terminal, and anything else I can take or leave. I don’t have a need to screen share or use any X only programs.
I do know that Chrome and electron based apps continue to have problems, but I rarely use them. If only there were one or more very wealthy companies who could back their development!
I switched around 3-4 years ago after long being vaguely interested in the new stack, but finding it too experimental yet.
After some initial hurdles, mostly related to screensharing, I've been happily and painlessly using it for years now. Tearing being a thing of the past is downright magical and it's great not to need to bother with the various config files. Hell, I recently (and regrettably) switched back to an NVIDIA card a few weeks ago and I was very happy to see how It Just Works (TM) after the rocky history the two had.
The future of Linux desktop is brighter than ever.
mostly related to screensharing
I still don’t have this figured out despite follow the portal wiki, Arch’s wiki, & literally copying a friend’s NixOS config for the screen share stuff verbatim—as well as trying Hyprland & its portal as downgrade to sway.
I threw in the towel & have used OBS + v4l2loopback if required to screen share (luckily terminal sharing is often good enough for 80% of my use cases).
OK, so when you start swaylock via SSH for testing, remember to always unlock instead of just cancelling swaylock with Ctrl+C. And hope it never crashes.
I think that a crashed locker leaving the system in a locked state is the correct behavior, or else you get fun bugs where you can unlock a machine by crashing the screensaver. JWZ has a post about this.
X11 servers are not fully aware of locker roles. If one both introduces a protocol to make the compositor aware of locking and writes all the protocols as local-first, the compositor being able to respawn crashed locker makes sense.
The Chrome GPU issue might be related to this, though there doesn't seem to be much of a concrete resolution yet. (Once for a similar issue I saw someone remark that the Vulkan backend worked, so that might be worth a try? I don't think that was specifically on Nvidia though...)
I shouldn’t see a chooser when clicking on Chrome’s “Window” tab. I should see previews of all windows. I should be able to select the window in Chrome, not with a separate chooser.
The separate chooser likely won't be going away; the entire purpose is so that the application won't see everything you have running just for you to share some limited subset of it. That being said, ofc there shouldn't be two dialogs anyway...but an issue has been open in Chromium for that for years. Not sure if/when that would be resolved.
I've been using Wayland since 2022, trying to keep xlsclients output empty. In 2026, it's nearly possible, I think even proton goes through wayland these days.
Cool times. I've been looking forward to this moment since 2000 at least.
Aside from the different text rendering, the other major issue for me is input latency: Emacs-pgtk feels significantly slower (less responsive) than Emacs.
This is especially true with fractional scaling and a relatively high DPI (200). Right now I have xwayland-satellite just for Emacs, because even Wine works with Wayland now (good enough for games)
Just to offer a counterpoint: switching from i3/X11 to wayland/sway fixed every single issues I had with my setup, sometimes not even knowing it was an X issue.
The old proprietary Antidote software was painfully slow and nearly unusable (but I must use it for professional reason). Under Xwayland, it works perfectly. It’s proprietary. It’s old. It has never been ported to Wayland and still works better!
But I completely agree that YMMV depending on your hardware and setup.
OK, so when you start swaylock via SSH for testing, remember to always unlock instead of just cancelling swaylock with Ctrl+C. And hope it never crashes.
If swaylock / the session lock program dies, you can start a new instance of it which would grab the lock and then normally unlock or SIGUSR1 it to release the lock. Not sure if this is part of the protocol but Niri definitely does that.
So from my perspective, switching from this existing, flawlessly working stack (for me) to Sway only brings downsides.
Yup.
It pains me to see how much time and effort has been, is, and will continue to be wasted on this Wayland nonsense. We have thousands of people feeling forced to break their workflows and do hundreds of hours of unnecessary work.... because an elite few dictated a change from on high.
Really pains me coming from so-called "free-as-in-freedom" software.
because an elite few dictated a change from on high. Really pains me coming from so-called "free-as-in-freedom" software.
I feel like this misunderstands what software freedom entitles you to. Any "dictum from on high" only has weight because of it's convinced some people who are spending their time working on free software to direct their efforts in a certain way. You have the freedom to entirely ignore their efforts, make changes to X11, and continue to maintain it (and some groups have sprung up do to this!)
The fact that software is "free software" doesn't automatically compel developers to write/maintain software to your specifications, but neither does it compel you to use those changes.
It's a "do-ocracy". Those who do the hard work get to make decisions. It has always been this way. If you don't participate in the making of the graphics stack, you don't get to make decisions about the graphics stack.
You can't demand that people who work on graphics drivers, desktop environments, GUI toolkits and applications, and give that software to you for free, do tonnes of duplicate effort to support both X11 and Wayland just to keep you personally happy. They don't work for you. Stop acting entitled.
If you don't participate in the making of the graphics stack, you don't get to make decisions about the graphics stack.
I do participate in the making of the graphics stack, including contributions to the X server, interclient communications standards, and maintaining a variety of applications, including a gui toolkit. I guess my ballot must have gotten lost in the mail.
You can't demand that people who work on graphics drivers, desktop environments, GUI toolkits and applications, and give that software to you for free, do tonnes of duplicate effort to support both X11 and Wayland just to keep you personally happy.
You have this completely backward. I haven't demanded any duplicate effort. If anything, my demand would simply be that they stop pushing their broken nonsense unnecessarily, putting more work on all the downstream maintainers like me, like the OP, like god knows how many other programmers and users who had to adapt to their breakage.
Meanwhile, my programs written for Microsoft Windows ages ago continue to run just fine to this day...
Wayland is not going away. By demanding that desktop environments and GUI toolkits keep supporting X11, you are demanding duplicate effort.
You don't maintain GNOME and KDE and Qt and GTK and all the graphics drivers. Your individual contribution is meaningless here. The consensus among most of the people working on this stuff is to focus on Wayland, and your disagreement won't change that.
Wayland is not going away.
I think it might, eventually.
Curious why you say this/what you think the path to that is?
While I probably wouldn't mind the outcome, I'm not sure I can see a likely scenario where that would happen.
By demanding that desktop environments and GUI toolkits keep supporting X11, you are demanding duplicate effort.
I'm not demanding this. I'm just pointing out that a lot of effort has going toward Wayland support from a lot of different people, most of whom I guess you dismiss as "meaningless".
But keep arguing with that strawman, I'm out of this thread.
Wait. If your initial post didn't mean "the Linux graphics/GUI stack should keep supporting X11" (whether in addition to or instead of Wayland), then ... what did it mean? What do you want?
Nice motte and bailey you got there.
I'm just pointing out that a lot of effort has going toward Wayland support from a lot of different people,
No, you talked about a supposed "elite" forcing everyone to do this. Literally the only thing that has happened is that some people have decided to spend their effort on Wayland.
Who exactly is pushing what on you and how? As you pointed out, you could just use Windows if you wanted to, and your programs would run just fine. Nobody is forcing you to do anything.
Those who do the hard work get to make decisions.
Exactly. And Linux/BSD distributions decides whose hard work actually gets shipped, and when. AlmaLinux decided that X11 is supported to 2032. So there really is no rush to migrate to Wayland.
I feel like a boiled frog, having to abandon long-used software because things change on me for no good reason. Worth noting is that the protean (lovecraftian??) insanity of constantly shifting methods of human-computer-interaction is not confined to the FOSS community -- things are in flux constantly in Windows and iOS, too. The new Glass UI on my family's iPhones keeps tripping us up, despite having used iPhones exclusively for over a decade now. Having to help my mom with usability issues on her Windows laptop makes me want to go live in a cave in the mountains.
Our industry seems to have lost respect for the actual human beings on the receiving end of our UX experiments. We are living in an Eternal September of HCI. I even wrote a post about how I moved away from thunderbird after over a decade of constant use, but realistically, the majority of computer users are forced into interacting with whatever has been poured into the trough.
not being able to share a specific windows is deal breaker for me. good to hear it's getting fixed soon.
Is it getting fixed across all Wayland compositors, or in each one as it gets around to implementing a new protocol?
Serious question, not meant to be snarky. As an external observer, one of the things which scares me the most about Wayland is the extent to which it appears it will absolutely annihilate the window manager ecosystem and drive folks to GNOME, KDE and sway, none of which I want to run.
I noticed this issue as well, I had a ton of issues with both sway and hyprland that I just don't have on i3, including horrible performance in the yuzu emulators forks (eden and citron).
(Note I have an AMD gpu, RX580. Bought with the hopes that sway would be better for nothing)
The performance issue is not present in KDE plasma wayland but I do not want to use plasma, I want a tiling WM.
I really hope phoenix takes off and implements the wayland compat layer the dev has talked about.
I love these kinds of articles! The only way Wayland is going to get polished is by documenting issues. Especially when it comes to smaller projects with less resources than KDE or Gnome. When I moved to Wayland I also switched to Plasma so my experience is much more refined than the alternative wms are having.
I also wonder why the x crowd isn't gravitating more towards wayback (a wayland compositor who's entire job is being a thin xserver). You get all the benefits of Wayland development with all the familiar comforts of x. Seems like a win win to me unless you just flat refuse to use Wayland things because reasons.
I think XRandR with Wayback is still work in progress? Not to mention dynamic reconfiguration via XInput2. Right now it looks like Wayback is busy covering the basics like multi-monitor, and they know what basics are working on next — so I won't be able to report anything useful.
benefits of Wayland development
Development is means to ends, not a good on its own. It is arguably a benefit when getting fresh enough hardwarefor issues with Xorg, but, as I like my current setup just fine, not before. Even if there are benefits applicable to my setup and mu goals, there are not enough to lose dynamic-input-reconfiguration features.
Well yes, but if the majority of linux graphics efforts are going into wayland (which seems to be the main complaint from X enthusiasts) there's a maintenance benefit to using wayland with a compat layer vs maintaining a parallel stack. I know wayback is new, but it seems like a good place for X users to concentrate their efforts so all the neat mini wms can keep working on modern systems.
if the majority of linux graphics efforts are going into wayland
The majority of linux graphics efforts go into the kernel or mesa. The input device stuff (another huge part of the work) is kernel + libinput. This stuff is shared between all the graphics implementations, wayland and X. (The Wayland devs cite this as being the reason Wayland became possible in the first place; otherwise they'd never have even been able to get started. Before 2008, most graphics drivers were actually part of X, but this got separated out.)
And so so so so so so so so so so so important to realize that graphics are a tiny slice of a gui. The user interface stuff and end user applications are way more important. Wayland is still playing catchup to X in this area, even after 17 years.
There is absolutely no benefit in moving before Wayback has reasonable feature coverage, and there are drawbacks to moving: the more explicit clinging to Xorg there is, the slower browsers will be at dropping usable X11 screenshare (and in general the better network-effect-based software usability will be).
Ah I meant in terms of developing a modern X. I'm not sure of the implementation complexity of something like XLibre, but maybe that project is moving faster than wayback. From what I remember xorg is on life support now because most of the core devs would rather work on wayland.
The best plausible outcome is probably a small team that is more competent (and probably more change-averse) than XLibre taking over doing the bare minimum to keep Xorg generally working after the current upstream abandons it; that being close to the current upstream's mode of operation already will probably help. Although maybe Phoenix will catch a Zig enthusiasm wave?
As long as AlmaLinux or somebody publishes the needed patches, Xorg under minimal-change policy is quite likely to keep having benefits over Wayback.
The nice thing about sway in this space is that it supports loading ICC profiles unlike many wlroots-based WMs.
I have been using KDE 6 on Debian 13 for at least the last 6 months and only recently noticed it was running on Wayland.
(Coming from someone who, in prior decades, would have spent a week customizing every package and pixel on their desktop.)
I wanted to start using Cosmic on its beta release, but the input settings menu was only a placeholder without custom keyboard layout implemented...
A here I thought that when I saw stapelberg on the xdg-desktop-portal-wlr issue tracker, it's because he switched to sway at some point (and even runs the main branch directly :0). Turns out I was quite off :P