Chrome crash "Aw, Snap!"

[jvailius]

Description

I have several chrome crashes with "Aw, Snap!" Error code: 4. Sorry for the unspecific title but I have not more insight what the issue might be.

Reproduce

It happens after opening 3 or more notebooks (containing some code but it just loads the kernel - no outputs present to render) or after repeatedly running several cells.

I have seen that memory usage for the chrome tab can very quickly grow close to 1GB. Maybe related. It may be a chrome issue when memory grows to fast in emscription - just guessing.

Context

• jupyterlite-core: 0.3.0
• jupyterlite-xeus: 0.1.8
• Operating System and version: Debian bookworm
• Browser and version: Chrome

I build jupyterlite with defaults except --XeusAddon.mount_jupyterlite_content=True and the notebook folder set to read only (chmod 0444)

My environment.yml:

name: xeus-python-kernel
channels:
  - https://repo.mamba.pm/emscripten-forge
  - conda-forge
dependencies:
  - xeus-python
  - pandas
  - matplotlib
  - requests
  - seaborn
  - xarray
  - jupyterlab_widgets
  - ipywidgets
  - pip
  - pip:
    - jlite/client  # a client site library

[jvailius]

Likely related to #339 (but in my case it is a xeus kernel)

Though I find it strange that it happened with a rather simple, single notebook where I executed cells repeatedly. I tested a few simple ipwidgets like Datepicker - just a couple of them. I played a bit more with it and I started with 10MB of memory for the tab and after running "Run all cells" it jumped to 900MB.

[jtpio]

Thanks @jvailius for the report 👍

Would it be possible to share the notebook that caused the issue?

[jvailius]

@jtpio Thanks for looking into it. Here it is: aw-snap.ipynb.tar.gz

After 3 or 4 "Run all cells" memory usage (reported in the chrome tab header) jumps to a value 10 and more times higher.

[jvailius]

Even if I just have one cell with content 1 and run it a few times ("Run all cells") memory usage will jump from a few MB to something close to 1 GB. If someone has an idea how to debug this issue, where to start then I could give it a try.

It will be hard to advertise this solution (although I'd like to!) in the company if it crashes my colleagues' browsers running even simple notebooks.