Should initialization errors be exposed to the user?

[krassowski]

Currently, if user configures InteractiveShellApp.exec_lines and it fails, there is no way for them to see the error in notebook.

For example set c.InteractiveShellApp.exec_lines=["raise ValueError()"].

This is visible in IPython/jupyter-server because it gets written to the terminal, but it does not show up in the notebook. There is even explicit code making sure that it won't be attached to the output of the first cell once it gets run:

ipykernel/ipykernel/kernelapp.py

Lines 730 to 737 in 84211d7

	if self.shell:
	self.init_gui_pylab()
	self.init_extensions()
	self.init_code()
	# flush stdout/stderr, so that anything written to these streams during
	# initialization do not get associated with the first execution request
	sys.stdout.flush()
	sys.stderr.flush()

With code to the same effect in IPython itself: https://github.com/ipython/ipython/blob/cc614bdc1d3f61303e265c26d593d384e4a90d1e/IPython/core/shellapp.py#L341-L360

I wonder if we should emit a special message for any output so that it would be shown in the Notebook console? The same way as unhandled ipywidget messages land in the console?

I mean this log console: