Static resources fail to load in Notebook renderer with 401 error

[DonJayamanne]

Originally filed here microsoft/vscode-jupyter#12843

Problem:

• Notebook kernel pre-load scripts are now defined statically via the package.json contribution notebookPreload
• However VS Code throws a 401 error when loading resources from within the Jupyter extension,

If however we have a dummy script file (in the Jupyter extension directory) that is defined as a rendererScripts in the createNotebookController VS Code API , then all resources from Jupyter extension directory are loaded

I.e. using the above API in some way registers the Juptyer extension directory as a supported root for scripts, and using the static contribution ntoebookPreload does not.

[DonJayamanne]

@mjbvz I have a work around in the Jupyter extension right now, i've added a dummy placeholder script that does nothing,

[mjbvz]

@DonJayamanne Can you share steps to reproduce this with the fix in place? Also, does this happen on normal desktop too, or only sshe

Unclear why this is not working as we should add preloads to the local resource roots:

vscode/src/vs/workbench/contrib/notebook/browser/view/renderers/backLayerWebView.ts

Line 1009 in e9229a1

Array.from(this.notebookService.getStaticPreloads(this.notebookViewType), x => dirname(x.entrypoint)),

[DonJayamanne]

@mjbvz
Here are the steps

• Install Python
• Install VS Code insiders and latest pre-release versions of Jupyter and Python
• NOTE: Install the pre-release version of Jupyter extension that is 2 days old
• Add the following setting into vscode

	"jupyter.widgetScriptSources": [
	],

• Create a python env and install the following dependencies
  You can use the command python -m venv .venv to create a .venv directory with Pytohn in a particular workspace folder (optionally you can use Python extension to do this for you)

python -m pip install ipykernel ipywidgets ipympl matplotlib

• Run the following code in a notebook pointing to the above venv (i.e. select the above venv as a kernel)

%matplotlib widget
import matplotlib.pyplot as plt
import numpy as np

X = np.linspace(0, 2*np.pi)
Y = np.sin(X)

fig, ax = plt.subplots()
ax.plot(X, Y)

• You might get prompted to enable CDN download, hit cancel for that
• The widget will not load, network tab in vscode will display a 401 error

Note: This is now working in the latest pre-release version of Jupyter extension

[mjbvz]

I believe the broken version is be v2023.2.1000451018 but I'm having trouble reproducing this

@DonJayamanne The output either never renders anything (but there is no network error) or renders as an image:

Any other steps?

[DonJayamanne]

Not sure which pre-release to use anymore, as the pre-release has changed since

• Please use the vsixc from this build (check the artifacts for the vsix) Remove dummy script vscode-jupyter#13066
  • here is the VSIX
• Install Python
• Add the following setting into vscode (reload VSCode)

	"jupyter.widgetScriptSources": [
	],

• Create a python env and install the following dependencies
  You can use the command python -m venv .venv to create a .venv directory with Pytohn in a particular workspace folder (optionally you can use Python extension to do this for you)

python -m pip install ipykernel ipywidgets ipympl matplotlib

• Run the following code in a notebook pointing to the above venv (i.e. select the above venv as a kernel)

%matplotlib widget
import matplotlib.pyplot as plt
import numpy as np

X = np.linspace(0, 2*np.pi)
Y = np.sin(X)

fig, ax = plt.subplots()
ax.plot(X, Y)

• You might get prompted to enable CDN download, hit cancel for that
• The widget will not load, network tab in vscode will display a 401 error

[mjbvz]

I'm still unable to reproduce this issue

@DonJayamanne Are you running the jupyter out of sources by chance or have you patched VS Code/jupyter in some other way? The path to the resource we a trying to load looks weird: ~/Desktop/development/vcs/vscode-jupyter/..... This looks like an extension development path

In normal installs, we should load resources from a path that looks more like:

~/.vscode-insiders/extensions/ms-toolsai.jupyter-renderers-1.0.15/..

If you are able, you can also try setting a breakpoint in _getResourceRootsCache and sharing the result of that function compared to the path of the js file we try loading in the webview

[vscodenpa]

This issue has been closed automatically because it needs more information and has not had recent activity. See also our issue reporting guidelines.

Happy Coding!

[DonJayamanne]

information was provided

[mjbvz]

@DonJayamanne #182664 lets you specific an additional set of localResourceRoots that should be allowed when loading the preload. Can you please try adding these for the python static preload contribution? Let me know if fixes the problem!

[DonJayamanne]

Thanks @mjbvz , seems to work, microsoft/vscode-jupyter#13531

[DonJayamanne]

Changes were made upstream in VS Code, marking as verified.