Editor hangs forever in Linux upon opening modal dialogs, with a message about Vulkan swapchains being outdated.

[NMitchell2026]

Summary

UnrealEditor becomes unresponsive, though not completely crashed or locked as certain messages still appear periodically in the logs, when opening modal dialogs within the editor on Linux. Once this state is reached, the editor appears stuck in this state and the process must be forcibly killed.

The issue is easily reached by attempting to create a new project. Upon clicking the create button, an outline of a modal dialog appears, but the interior is never rendered. At the same time, the logs present this message:

[2026.06.22-03.35.31:688][978]LogVulkanRHI: AcquireNextImage() failed due to the outdated swapchain, not even attempting to present.

Further analysis of the issue shows, via Unreal Insights and manual debugging, that the engine is stuck within the FVulkanDynamicRHI::WaitAndProcessInterruptQueue method, apparently continually timing out on a call to VulkanRHI::vkWaitSemaphoresKHR. The Vulkan Semaphore apparently never reaches whatever value is being expected of it, as the editor waits here forever. This then blocks the main editor process from continuing with whatever it was doing (e.g. trying to render a modal dialog) and the whole process grinds to a halt.

System information:

Unreal Engine Version: 5.7.4
OS: Linux - Debian - Bullseye - Kernel 5.10.0-22-amd64
GPU: Nvidia GeForce RTX 3070
GPU Driver Version: 595.84

What Type of Bug are you experiencing?

Rendering (Graphics / Niagara)

Steps to Reproduce

1. Launch the editor
2. Open File → New Project
3. Configure Project settings (I used a C++ project, scalable, no VS-Code)
4. Click Create

Expected Result

Editor continues to run as expected, opening and displaying any necessary dialogs and responding to user input.

Observed Result

Editor freezes upon pressing Create, showing only the outlines of spawning in modal windows but never rendering them. User input is ignored.

Affects Versions

5.7

Platform(s)

Linux

Upload an image

[LoganTnk]

Similar hang /app frozen state when launching the app.
Cannot navigate around too much, select new project, a panel is created, but the app becomes frozen. Loaded w opengl or vulkan flags, ie ./UnrealEditor -opengl & also ./UnrealEditor -vulkan, app is frozen when attempting to navigate around or press any slate buttons. WHile the URL linked buttons open browser to url(s), anything else while on the apps freezing.

Although other versions, v5.7.4 launching w/o issue and not freeze ui.

My system is AMD 7900 X CPU, 7900 XT GPU, GPUs running mesa drivers, OS Linux Mint 22.3 Zena.

[gyunseo]

Unreal Editor 5.6.1 gets frozen when try to pick a color palette dialog in any editor such Material Editor, UMG.
This problem comes first since I upgrade Fedora44 and Nvidia 595.71.05.
UE5.6.1 have worked perfect with Nvidia 580.xx.
PC Environment : Fedora 44, Kernel 7.0.13-200.fc44.x86_64
Nvidia Driver: 595.71.05

[TheCupcakeisalie]

On a fresh install, I’m having the same issue. I had claude help me triage it but no fix.

I put claudes full diagnostic here:

System: RTX 3090, NVIDIA 595.71.05 (open), Ubuntu 24.04.4, kernel 6.17, GNOME on X11,
single 4K@60. Repro on both the UE 5.8.0 binary release and a 5.8.1 source build (branch 5.8).

Repro (~15s, 100% eventually): open any MaterialInstanceConstant, open the Color Picker,
close it. Crashes within a few frames. PIE is not required — a fresh editor is enough,
and it sometimes dies before I ever open the picker.

For me it’s a hard crash rather than a hang. Two faces of the same call:
SIGSEGV in libnvidia-glcore at FVulkanFenceManager::WaitForFence (VulkanSynchronization.cpp:127)
← FVulkanSwapChain::AcquireImageIndex (VulkanSwapChain.cpp:783)
← FVulkanViewport::Present ← EndDrawingViewport
…or VK_ERROR_DEVICE_LOST at VulkanSynchronization.cpp:136 (same function, the VERIFYVULKANRESULT).
No NVRM Xid in any crash window, DEVICE FAULT REPORT always empty, and the faulting reads
are garbage (0x2885f, 0x288f7, 0x100000001) — so the GPU isn’t faulting; it’s a userspace
segfault inside the driver.

Possibly useful: the “outdated swapchain” message looks protective, not causal. AcquireImageIndex
returns early on OUT_OF_DATE and never reaches the WaitForFence that crashes. In one run the log
flooded with “AcquireNextImage() failed due to the outdated swapchain” every ~0.5s and the editor
survived 25 picker open/close cycles. Runs without that sustained flood reach the fence wait and die.
Same binary, same config, opposite outcomes. When it doesn’t crash, the picker often opens
completely transparent — mapped, IsViewable, top of stacking order, rendering nothing.

Notably, 5.8.1 does NOT fix this, despite containing:
1e1137c3 “VulkanRHI: Add explicit sync for window destruction instead of 3 frame delay.
Add handling for Not_Ready when acquiring index.” (UE-380098, UE-382159)
1baafca3 “VulkanRHI: Fix SubmissionThread deadlock from recycled rendering done semaphore handle.”
(UE-385632)
Those add the “outdated swapchain” guard but the crash persists.

Ruled out here (all still crash):

• NVIDIA 590.48.01 → 595.71.05-open: no change
• r.Vulkan.EnablePipelineLRUCache=1: confirmed applied (LastSetBy: SystemSettingsIni), no effect
• r.Lumen.AsyncCompute=0: delays it only (doesn’t cover r.LumenScene.Lighting.AsyncCompute)
• r.RDG.AsyncCompute=0 (global async compute off): still crashes
• Dual mixed-refresh monitors vs single 4K@60: no difference
• GNOME fractional scaling (off), xrandr transform (identity)
• Sunshine / gnome-remote-desktop (GPU capture clients) stopped: no difference
• Blueprint-only demo template project, so not content-specific

[TheCupcakeisalie]

I was able to temporarily bypass this issue by rebuilding with

In Engine/Source/Runtime/VulkanRHI/Private/Linux/VulkanLinuxPlatform.h:

#define VULKAN_USE_IMAGE_ACQUIRE_FENCES 0

[gyunseo]

These days, I updated Fedora 44 system and I saw that’s fixed In kernel 7.1.4 and Nvidia 610.43.03 (RTX 3080).
But in my other Fedora 44 with RTX5090 I updated to kernel 7.1.5 today has still the same problem.