Extension host terminated uexpectedly when using ipywidgets.Fileupload

[beralves]

Environment data

• VS Code version: 1.52.1
• Jupyter Extension version (available under the Extensions sidebar): v2020.12.414227025
• Python Extension version (available under the Extensions sidebar): v2021.1.502429796
• OS (Windows | Mac | Linux distro) and version: Windows 10
• Python and/or Anaconda version: 3.8
• Type of virtual environment used (N/A | venv | virtualenv | conda | ...): venv
• Jupyter server running: Local

Expected behaviour

ipywidgets.Fileupload should work on vscode while running the notebook. Fileupload works if I launch "jupyter notebook" in the command and run my notebook there. It would be helpful to run it inside vscode, to avoid context switching and use debug features.

Actual behaviour

ipywidgets.Fileupload crashes extension host on vscode while running the notebook. I get the message "Extension host terminated uexpectedly."

Steps to reproduce:

1. Create a notebook and include an ipywidgets.Fileupload. Run the cell, it displays fine. Click "upload", select the file you would like to upload. When you click "OK" to close the file browser dialog you get the error message.

Logs

Output for Jupyter in the Output panel (View→Output, change the drop-down the upper-right of the Output panel to Jupyter)

XXX

[rchiodo]

Thanks for the bug. We haven't tried using FileUpload before, but we'll see if we can repro with it.

[rchiodo]

Yep crashes the host for me too. Just do this:

widgets.FileUpload(accept='', multiple = False)

Then click on the Upload button and pick a text file. Extension host will die

[rchiodo]

Looks to me like a message buffer being to big. Callstack for crash is here:

postToSocket (d:\Source\vscode-jupyter\src\client\datascience\raw-kernel\rawSocket.ts:230)
(anonymous function) (d:\Source\vscode-jupyter\src\client\datascience\raw-kernel\rawSocket.ts:222)
VM_Unknown (undefined:97)
[ Promise.then ]
sendMessage (d:\Source\vscode-jupyter\src\client\datascience\raw-kernel\rawSocket.ts:221)
sendToRealKernel (d:\Source\vscode-jupyter\src\client\datascience\raw-kernel\rawSocket.ts:104)
sendPendingMessages (d:\Source\vscode-jupyter\src\client\datascience\ipywidgets\ipyWidgetMessageDispatcher.ts:354)
sendRawPayloadToKernelSocket (d:\Source\vscode-jupyter\src\client\datascience\ipywidgets\ipyWidgetMessageDispatcher.ts:142)
receiveMessage (d:\Source\vscode-jupyter\src\client\datascience\ipywidgets\ipyWidgetMessageDispatcher.ts:109)
receiveMessage (d:\Source\vscode-jupyter\src\client\datascience\ipywidgets\ipyWidgetMessageDispatcherFactory.ts:42)
onMessage (d:\Source\vscode-jupyter\src\client\datascience\ipywidgets\commonMessageCoordinator.ts:106)
(anonymous function) (d:\Source\vscode-jupyter\src\client\datascience\ipywidgets\notebookIPyWidgetCoordinator.ts:88)
fire (event.ts:623)
onDidReceiveMessage (extHostNotebook.ts:44)
$onDidReceiveMessage (extHostNotebook.ts:612)
_doInvokeHandler (rpcProtocol.ts:409)
_invokeHandler (rpcProtocol.ts:394)
_receiveRequest (rpcProtocol.ts:310)
_receiveOneMessage (rpcProtocol.ts:246)
(anonymous function) (rpcProtocol.ts:105)

[rchiodo]

The ZMQ library is crashing trying to send the data to kernel.

[beralves]

Hi, any plans for this issue? I'm happy to test if needed.

[rchiodo]

There are no plans to fix this at the moment. We generally wait for more feedback before prioritizing.

A workaround might be to start your own jupyter server and set it as the 'remote' URI to connect to. That should use a different communication mechanism to send the data to the kernel then. I haven't tried it though.