@lysnikolaou Do you remember why this was expressions and not star_expressions? I cannot see any edge case that makes this change incorrect but now I am curious.

I think that this was for consistency with 3.8. While 3.8 and 3.9 currently fail at parser level for eval('*a'), this patch fails at bytecode compiler level.

We could make that fail at the parser level if you want but not sure if is worth it because we would need to factor that into some rule and I don't know if we can reuse the rest. What's your opinion?

Wouldn't leaving it as is work?

The reason I'm pushing for the error to be in parser level is due to the quality of the error message. Here's an example:

Without this patch:

➜  cpython git:(bpo-40631) ✗ cat a.py       
a = [1, 2, 3]
eval('''


*a''')
➜  cpython git:(bpo-40631) ✗ python a.py
Traceback (most recent call last):
  File "/home/lysnikolaou/repos/cpython/a.py", line 2, in <module>
    eval('''
  File "<string>", line 4
    *a
    ^
SyntaxError: invalid syntax

With this patch:

➜  cpython git:(bpo-40631) ✗ cat a.py
a = [1, 2, 3]
eval('''


*a''')
➜  cpython git:(bpo-40631) ✗ ./python a.py
Traceback (most recent call last):
  File "/home/lysnikolaou/repos/cpython/a.py", line 2, in <module>
    eval('''
  File "<string>", line 4
SyntaxError: can't use starred expression here

The first one makes a much better job of making clear what the error is I feel. But if it's too much effort and leaving it as is doesn't work, I'm okay with it.

I very much agree. I have closed the PR based on this discussion and Guido's request. Thanks for the examples and the comment.

This patch makes (*x), y valid, but with multiple layers of parentheses (((*x))), y is still considered invalid.