[Regression] Scaling artifacts on modals

[tekstryder]

Hyprland version

Hyprland 0.53.0 built from branch main at commit c8b5023 dirty (opengl: allow texture filter to be changed ( 13078)).
Date: Tue Jan 27 16:21:53 2026
Tag: v0.53.0-84-gc8b5023bb, commits: 6840

Libraries:
Hyprgraphics: built against 0.5.0, system has 0.5.0
Hyprutils: built against 0.11.0, system has 0.11.0
Hyprcursor: built against 0.1.13, system has 0.1.13
Hyprlang: built against 0.6.8, system has 0.6.8
Aquamarine: built against 0.10.0, system has 0.10.0

Version ABI string: c8b5023bb0610be4d1d2987ef05168834b2661ba_aq_0.10_hu_0.11_hg_0.5_hc_0.1_hlg_0.6
flags set:
no xwayland

Regression

Yes.

Bisected to 6fce2d7 as first bad commit.

Describe the bug

Since modal props were added in #12024 I've been using window rules...

windowrule = border_size 0, match:class org.gnome.FileRoller, match:title ^$
windowrule = border_size 0, match:class org.gnome.Nautilus, match:title ^$

...to prevent wonky squared borders on rounded GTK4 modal dialogs.

Since 6fce2d7 there now appears various pixels surrounding the dialogs, which change when dialogs are moved.

Reproduction steps

1. View a GTK4 modal dialog using fractional scaling on git after 6fce2d7

2. Notice border pixel corruption, move the window around and see it change

monitor = DP-1, 3840x2160, 0x0, 1.333333
monitor = DVI-D-1, 2560x1440, 2880x-425, 1.066667, transform, 3

Does not occur without scaling factor.

[tekstryder]

As mentioned in #12878 (reply in thread), my current workaround for this new regression is:

Set debug:disable_scale_checks and set scaling factors to 1.35 and 1.05 respectively.

monitor = DP-1, 3840x2160, 0x0, 1.35
monitor = DVI-D-1, 2560x1440, 2845x-455, 1.05, transform, 3

These happen to be the same scaling factors I use on Sway without issue as well.

Works around this problem, #11099, and #12878.

Also provides a near-perfect PPI match between my two displays, so I'll prolly stay with these fairly ideal scaling factors.

[vaxerski]

can't repro

1.25 scale, disabled shadows as well.

1.6

1.07

[tekstryder]

1.33x in the OP?

[tekstryder]

Oh curious.

I can't repro on latest git either. Hmm.

[tekstryder]

Without building against older hyprwire tag, I can't do a full bisect to find the fixing commit but narrowed to 3 commits...

Still bad as of ec120d5

All good as of db6114c

So likely #13124 fixed this.

[tekstryder]

Un-marking as answered, this issue is back.

Previously bisected to 6fce2d7

See also, likely dupe/related #13612

[tekstryder]

Marking this as answered by the fact that such scaling issues are inherent in Wayland compositors at this time.

See also:

• Fractional Scaling Artifacts | Background Bleeding #12878

Perhaps we can look forward to future work with the xx-fractional-scale-v2 proposal for an alternate way to handle fractional scaling:

• https://gitlab.freedesktop.org/wayland/wayland-protocols/-/merge_requests/499

This ultimately came down to finding the sweet spot of:

1. scaling factors that are easy on the eyes
2. scaling factors that result in a good PPI match between displays of different dimensions/resolutions
3. finding scaling factors that satisfy 1 & 2, but do not cause this pixel corruption issue.