Fix incorrect linenos on fstring tokens with escaped newlines

I don't think this can affect Black itself much (maybe for formatting
ranges), but I ran into this with https://github.com/JelleZijlstra/lib2toast

CHANGES.md

@@ -39,6 +39,8 @@
 
 - Fix bug with Black incorrectly parsing empty lines with a backslash (#4343)
 
+- Fix incorrect line numbers in the tokenizer for certain tokens within f-strings (#4423)
+
 ### Performance
 
 <!-- Changes that improve Black's performance. -->

src/blib2to3/pgen2/tokenize.py

@@ -638,7 +638,7 @@ def generate_tokens(
                 else:
                     if is_fstring_start(token):
                         fstring_start, token = _split_fstring_start_and_middle(token)
-                        fstring_start_epos = (lnum, spos[1] + len(fstring_start))
+                        fstring_start_epos = (spos[0], spos[1] + len(fstring_start))
                         yield (
                             FSTRING_START,
                             fstring_start,

tests/test_tokenize.py

@@ -0,0 +1,117 @@
+"""Tests for the blib2to3 tokenizer."""
+
+import io
+import black
+import sys
+import textwrap
+from blib2to3.pgen2 import tokenize, token
+from dataclasses import dataclass
+
+
+@dataclass
+class Token:
+    type: str
+    string: str
+    start: tokenize.Coord
+    end: tokenize.Coord
+
+
+def get_tokens(text: str) -> list[Token]:
+    """Return the tokens produced by the tokenizer."""
+    readline = io.StringIO(text).readline
+    tokens: list[Token] = []
+
+    def tokeneater(
+        type: int, string: str, start: tokenize.Coord, end: tokenize.Coord, line: str
+    ) -> None:
+        tokens.append(Token(token.tok_name[type], string, start, end))
+
+    tokenize.tokenize(readline, tokeneater)
+    return tokens
+
+
+def assert_tokenizes(text: str, tokens: list[Token]) -> None:
+    """Assert that the tokenizer produces the expected tokens."""
+    actual_tokens = get_tokens(text)
+    assert actual_tokens == tokens
+
+
+def test_simple() -> None:
+    assert_tokenizes(
+        "1",
+        [Token("NUMBER", "1", (1, 0), (1, 1)), Token("ENDMARKER", "", (2, 0), (2, 0))],
+    )
+    assert_tokenizes(
+        "'a'",
+        [
+            Token("STRING", "'a'", (1, 0), (1, 3)),
+            Token("ENDMARKER", "", (2, 0), (2, 0)),
+        ],
+    )
+    assert_tokenizes(
+        "a",
+        [Token("NAME", "a", (1, 0), (1, 1)), Token("ENDMARKER", "", (2, 0), (2, 0))],
+    )
+
+
+def test_fstring() -> None:
+    assert_tokenizes(
+        'f"x"',
+        [
+            Token("FSTRING_START", 'f"', (1, 0), (1, 2)),
+            Token("FSTRING_MIDDLE", 'x', (1, 2), (1, 3)),
+            Token("FSTRING_END", '"', (1, 3), (1, 4)),
+            Token("ENDMARKER", "", (2, 0), (2, 0)),
+        ],
+    )
+    assert_tokenizes(
+        'f"{x}"',
+        [
+            Token("FSTRING_START", 'f"', (1, 0), (1, 2)),
+            Token("FSTRING_MIDDLE", '', (1, 2), (1, 2)),
+            Token("LBRACE", '{', (1, 2), (1, 3)),
+            Token("NAME", "x", (1, 3), (1, 4)),
+            Token("RBRACE", '}', (1, 4), (1, 5)),
+            Token("FSTRING_MIDDLE", '', (1, 5), (1, 5)),
+            Token("FSTRING_END", '"', (1, 5), (1, 6)),
+            Token("ENDMARKER", "", (2, 0), (2, 0)),
+        ],
+    )
+    assert_tokenizes(
+        'f"{x:y}"\n',
+        [
+            Token(type="FSTRING_START", string='f"', start=(1, 0), end=(1, 2)),
+            Token(type="FSTRING_MIDDLE", string="", start=(1, 2), end=(1, 2)),
+            Token(type="LBRACE", string="{", start=(1, 2), end=(1, 3)),
+            Token(type="NAME", string="x", start=(1, 3), end=(1, 4)),
+            Token(type="OP", string=":", start=(1, 4), end=(1, 5)),
+            Token(type="FSTRING_MIDDLE", string="y", start=(1, 5), end=(1, 6)),
+            Token(type="RBRACE", string="}", start=(1, 6), end=(1, 7)),
+            Token(type="FSTRING_MIDDLE", string="", start=(1, 7), end=(1, 7)),
+            Token(type="FSTRING_END", string='"', start=(1, 7), end=(1, 8)),
+            Token(type="NEWLINE", string="\n", start=(1, 8), end=(1, 9)),
+            Token(type="ENDMARKER", string="", start=(2, 0), end=(2, 0)),
+        ],
+    )
+    assert_tokenizes(
+        'f"x\\\n{a}"\n',
+        [
+            Token(type="FSTRING_START", string='f"', start=(1, 0), end=(1, 2)),
+            Token(type="FSTRING_MIDDLE", string="x\\\n", start=(1, 2), end=(2, 0)),
+            Token(type="LBRACE", string="{", start=(2, 0), end=(2, 1)),
+            Token(type="NAME", string="a", start=(2, 1), end=(2, 2)),
+            Token(type="RBRACE", string="}", start=(2, 2), end=(2, 3)),
+            Token(type="FSTRING_MIDDLE", string="", start=(2, 3), end=(2, 3)),
+            Token(type="FSTRING_END", string='"', start=(2, 3), end=(2, 4)),
+            Token(type="NEWLINE", string="\n", start=(2, 4), end=(2, 5)),
+            Token(type="ENDMARKER", string="", start=(3, 0), end=(3, 0)),
+        ],
+    )
+
+
+if __name__ == "__main__":
+    code = sys.stdin.read()
+    tokens = get_tokens(code)
+    text = f"assert_tokenizes({code!r}, {tokens!r})"
+    text = black.format_str(text, mode=black.FileMode())
+    print(textwrap.indent(text, "    "))