Incorrect line number of wrong symbol name inside f-strings

This is not a problem in PyFlakes but in the formatted literals implementation in the core of CPython itself. The AST child Name node for the formatting variables reports 1 as both line and column as the nodes are generated by PyParser_ASTFromString "automagically" in fstring_find_expr (Python/pythoinrun.c and Python/ast.c respectively).

I don't see how this should/could be fixed by PyFlakes

Sadly this is true, the Name elements of an fstring carry an useless lineno, as the following script demonstrates:

from ast import FormattedValue, Str, parse

code = """\
a = 1
b = f"a={a} c={c}"
"""

parsed = parse(code)
first, second = parsed.body
bvalue = second.value

print(type(bvalue))

for n in bvalue.values:
    print(type(n))
    if isinstance(n, Str):
        print("%d:%d %r" % (n.lineno, n.col_offset, n.s))
    else:
        assert isinstance(n, FormattedValue)
        print("%d:%d" % (n.lineno, n.col_offset))
        v = n.value
        print(type(v))
        print(" %d:%d %r" % (v.lineno, v.col_offset, v.id))

that emits:

<class '_ast.JoinedStr'>
<class '_ast.Str'>
2:4 'a='
<class '_ast.FormattedValue'>
2:4
<class '_ast.Name'>
  1:1 'a'
<class '_ast.Str'>
2:4 ' c='
<class '_ast.FormattedValue'>
2:4
<class '_ast.Name'>
  1:1 'c'

However, couldn't PyFlakes recognize those Name elements and go up to the enclosing FormattedValue and use its position as the error location?

FYI, this is solved in Python 3.6.3, yay!