[Bug]: Google Meet tab crashes (SIGILL on renderer Media thread) seconds after stopping a screen share #2720

Closed
opened 2026-10-15 17:53:47 +00:00 by adrianfagerland · 1 comment
adrianfagerland commented 2026-10-15 17:53:47 +00:00 (Migrated from github.com)

Version

0.18.1.1 (Chromium 154.0.8037.57), official x86_64 AppImage (run through NixOS appimageTools)

Have you tested that this is not an upstream issue or an issue with your configuration?

  • I have tried reproducing this issue in Chrome and it could not be reproduced there
  • I have tried reproducing this issue in ungoogled-chromium and it could not be reproduced there
  • I have tried reproducing this issue in Helium with a new and empty profile using --user-data-dir command line argument and it could not be reproduced there

Description

The Google Meet tab's renderer crashes (SIGILL on its Media thread) about 5 seconds after stopping a screen share on Wayland, which drops me from the call.

How to Reproduce?

  1. Join a Google Meet call in Helium on Wayland (Hyprland + xdg-desktop-portal-hyprland).
  2. Present the entire screen through the portal picker (it was shared for ~35 minutes when this happened).
  3. Click "Stop presenting" and go back to the Meet tab.
  4. About 5 seconds later the Meet tab's renderer crashes.

It isn't deterministic: stopping a screen share the day before, with the same setup, did not crash.

Actual behavior

Timeline from the system journal (local time):

  • 15:14:13 Helium releases its "Native desktop capture" screensaver inhibit (the share has stopped).
  • 15:14:18 The kernel logs traps: Media[…] trap invalid opcode ip:57fe95cf2bc3 sp:73fdb17fd050 error:0 in helium[73fabc3,57fe919e1000+d561000]. The --type=renderer process for https://meet.google.com dies with SIGILL (ILL_ILLOPN), i.e. an intentional crash at offset 0x73fabc3 in the helium binary, on the renderer thread named Media.
  • 15:14:19 The GPU process also writes a minidump (its gpu-url-chunk is a blob:https://meet.google.com/… URL, and the strings gpu.mojom.GpuChannel / null field 1 appear in it). The GPU process keeps running, so this looks like a side effect of the renderer dying.
  • 15:14:19 Meet's Blink wake lock is released, and the call is gone.

Expected behavior

Stopping the screen share returns to the normal Meet view and the call continues.

Additional context

  • OS: NixOS 26.11, kernel 7.2.8 (CachyOS), x86_64
  • Compositor: Hyprland 0.56.2, xdg-desktop-portal-hyprland 1.4.1 with screencopy { force_shm = true; max_fps = 30 } (DMA-BUF negotiation for whole-monitor capture fails on this setup, so capture is SHM-only), PipeWire 1.6.8
  • GPU: Intel Iris Xe (ADL GT2, 0x46a6), Mesa 26.2.3, ANGLE on OpenGL ES
  • Flags: --ozone-platform=wayland --enable-features=UsePortalFileDialog --disable-features=VaapiVideoDecoder,VaapiVideoEncoder,VaapiVideoDecodeLinuxGL (video decode/encode is in software)
  • The system was not low on memory (about half of RAM was available), and nothing killed the process externally.
  • Minidumps for the renderer and the GPU process are saved locally but haven't been uploaded yet. I can send them through helium://crashes if that helps.
  • I've since updated to 0.19.3.1 (Chromium 155.0.8059.79) and will comment here if it happens again.
### Version 0.18.1.1 (Chromium 154.0.8037.57), official x86_64 AppImage (run through NixOS `appimageTools`) ### Have you tested that this is not an upstream issue or an issue with your configuration? - [ ] I have tried reproducing this issue in Chrome and it could not be reproduced there - [ ] I have tried reproducing this issue in ungoogled-chromium and it could not be reproduced there - [ ] I have tried reproducing this issue in Helium with a new and empty profile using `--user-data-dir` command line argument and it could not be reproduced there ### Description The Google Meet tab's renderer crashes (SIGILL on its `Media` thread) about 5 seconds after stopping a screen share on Wayland, which drops me from the call. ### How to Reproduce? 1. Join a Google Meet call in Helium on Wayland (Hyprland + xdg-desktop-portal-hyprland). 2. Present the entire screen through the portal picker (it was shared for ~35 minutes when this happened). 3. Click "Stop presenting" and go back to the Meet tab. 4. About 5 seconds later the Meet tab's renderer crashes. It isn't deterministic: stopping a screen share the day before, with the same setup, did not crash. ### Actual behavior Timeline from the system journal (local time): - `15:14:13` Helium releases its "Native desktop capture" screensaver inhibit (the share has stopped). - `15:14:18` The kernel logs `traps: Media[…] trap invalid opcode ip:57fe95cf2bc3 sp:73fdb17fd050 error:0 in helium[73fabc3,57fe919e1000+d561000]`. The `--type=renderer` process for `https://meet.google.com` dies with SIGILL (`ILL_ILLOPN`), i.e. an intentional crash at offset `0x73fabc3` in the `helium` binary, on the renderer thread named `Media`. - `15:14:19` The GPU process also writes a minidump (its `gpu-url-chunk` is a `blob:https://meet.google.com/…` URL, and the strings `gpu.mojom.GpuChannel` / `null field 1` appear in it). The GPU process keeps running, so this looks like a side effect of the renderer dying. - `15:14:19` Meet's Blink wake lock is released, and the call is gone. ### Expected behavior Stopping the screen share returns to the normal Meet view and the call continues. ### Additional context - OS: NixOS 26.11, kernel 7.2.8 (CachyOS), x86_64 - Compositor: Hyprland 0.56.2, xdg-desktop-portal-hyprland 1.4.1 with `screencopy { force_shm = true; max_fps = 30 }` (DMA-BUF negotiation for whole-monitor capture fails on this setup, so capture is SHM-only), PipeWire 1.6.8 - GPU: Intel Iris Xe (ADL GT2, `0x46a6`), Mesa 26.2.3, ANGLE on OpenGL ES - Flags: `--ozone-platform=wayland --enable-features=UsePortalFileDialog --disable-features=VaapiVideoDecoder,VaapiVideoEncoder,VaapiVideoDecodeLinuxGL` (video decode/encode is in software) - The system was not low on memory (about half of RAM was available), and nothing killed the process externally. - Minidumps for the renderer and the GPU process are saved locally but haven't been uploaded yet. I can send them through `helium://crashes` if that helps. - I've since updated to 0.19.3.1 (Chromium 155.0.8059.79) and will comment here if it happens again.
wukko commented 2026-10-15 18:00:52 +00:00 (Migrated from github.com)

could you upload the crash and reply with the uploaded crash id?

could you upload the crash and reply with the uploaded crash id?
Sign in to join this conversation.
No milestone
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
imputnet/helium#2720
No description provided.