(I may be completely wrong in my analysis here, apologies in advance!)

I recently saw a regression in IPython 8 which I believe is due to stack_data. To illustrate the situation, I have adapted an example from ipython/ipython#6300. Save this to a file called cythontest.pyx:

cdef list foo = []

def f():
    raise ValueError('message')

and run this under IPython (pyximport is part of Cython):

import pyximport
pyximport.install(reload_support=True)
import cythontest
cythontest.f()

IPython 7 does show the following (great!) traceback:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-1-c52676a06430> in <module>
      4 import cythontest
      5 
----> 6 cythontest.f()

~/cythontest.pyx in cythontest.f()
      2 
      3 def f():
----> 4     raise ValueError('message')

ValueError: message

IPython 8, however, does not show the source code for cythontest.f:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Input In [3], in <module>
      3 pyximport.install(reload_support=True)
      4 import cythontest
----> 6 cythontest.f()

File ~/cythontest.pyx:4, in cythontest.f()

ValueError: message

I believe this happens because IPython 8 depends on this repository's package (stack_data), which tries to parse the pyx file and fails to produce an AST because Cython is a superset of Python. Given that we have line number information and the pyx file is valid source code (just not Python source code), I believe it would be better for the user if the relevant source fragment were printed even in cases where the AST is not available, similar to what IPython 7 does.