Conda package defines a Windows‑only style kernel on Linux

[gisfromscratch]

Describe the bug
Installing jupyterlab from the esri conda channel on Linux creates a broken Jupyter kernel whose kernel.json incorrectly references a Windows executable (python.exe) and uses a Windows-style backslash path.

This causes JupyterLab to fail to start any kernel, producing a “No such file or directory … python.exe” error.

The issue appears to be caused by distributing a Windows‑specific kernel spec even when installed on Linux.

To Reproduce
Steps to reproduce the behavior:

conda create -n fail
conda activate fail
conda install -c esri jupyterlab
jupyter kernelspec list

Inspect the kernel:

cat $CONDA_PREFIX/share/jupyter/kernels/python3/kernel.json

{
 "argv": [
  ".../envs/fail\\python.exe",
  "-m",
  "ipykernel_launcher",
  "-f",
  "{connection_file}"
 ],
 "display_name": "Python 3 (ipykernel)",
 "language": "python"
}

Expected behavior
The jupyter kernel should just work on Linux.

"argv": [
  ".../envs/fail/bin/python",
  "-m",
  "ipykernel_launcher",
  "-f",
  "{connection_file}"
]

Platform (please complete the following information):

• OS: Linux (Debian/Ubuntu-based; same behavior under WSL2 and native Linux)
• Browser not relevant
• Python API Version 2.4.3

[achapkowski]

@gisfromscratch thank you for reporting this, we are going to look at it.

[rwmajor2]

@gisfromscratch Thank you for your feedback. We were able to reproduce the behavior. We have determined the root cause of this and will plan for a fix at our next release.

The root cause is actually in our noarch build of ipykernel.

As a workaround, can you try the following? This will update the ipykernel package to a defaults channel version.

conda create -n success -c esri jupyterlab python=3.13
conda activate success
conda install main::ipykernel
jupyter kernelspec list

cat /data01/miniconda3/envs/success/share/jupyter/kernels/python3/kernel.json

{
 "argv": [
  "/data01/miniconda3/envs/success/bin/python",
  "-Xfrozen_modules=off",
  "-m",
  "ipykernel_launcher",
  "-f",
  "{connection_file}"
 ],
 "display_name": "Python 3 (ipykernel)",
 "language": "python",
 "metadata": {
  "debugger": true
 }
}