The documentation at https://docs.python.org/3/reference/simple_stmts.html says that:

"Names listed in a global statement must not be used in the same code block textually preceding that global statement"

But then later:

"CPython implementation detail: The current implementation does not enforce the two restrictions,
but programs should not abuse this freedom, as future implementations may enforce them..."

Code like this

def f():
    x = 1
    global x

gives SyntaxWarning for several releases, maybe it is time to make it a SyntaxError?

I like fixing this. I suppose it'll be fixed for nonlocal too?

Yes, nonlocal will be fixed too. This will be a relatively small patch, but I am a bit afraid there could be merge conflicts (the code affected could overlap with changes for PEP 526).

Easiest way would be to simply make this a single patch with PEP 526 implementation (this will actually even slightly simplify the implementation).

Alternatively we could fix this right after the implementation patch is applied. What do you prefer?

I strongly prefer to have it as a separate patch (either before or
after) to avoid the appearance of piggy-backing this in with a much
larger change.

OK, then I think after will be safer. Let us came back to this right after PEP 526.

I am attaching the patch.

Yury, I also added you here, just in case you have time for this.

Btw, while working on this I have found that the second restriction:

"""
Names listed in a global statement must not be ... in a for loop control target, class definition, function definition, or import statement.
"""

is not enforced at all, it does not give even a SyntaxWarning.
Is it something worth fixing?

Added few tests to the patch

I'll commit this.

New changeset 804b71d43c85 by Guido van Rossum in branch 'default':
Issue #27999: Make "global after use" a SyntaxError, and ditto for nonlocal.
https://hg.python.org/cpython/rev/804b71d43c85

Thanks again Ivan!