Fix the protect-pip-on-windows logic

Author: pradyunsg

src/pip/_internal/commands/install.py

@@ -344,9 +344,9 @@ def run(self, options: Values, args: List[str]) -> int:
             except KeyError:
                 modifying_pip = False
             else:
-                # If we're not replacing an already installed pip,
-                # we're not modifying it.
-                modifying_pip = pip_req.satisfied_by is None
+                # If we're replacing an already installed pip, we're modifying it.
+                modifying_pip = pip_req.satisfied_by is not None
+
             protect_pip_from_modification_on_windows(modifying_pip=modifying_pip)
 
             check_binary_allowed = get_check_binary_allowed(finder.format_control)

tests/functional/test_install_upgrade.py

@@ -1,5 +1,6 @@
 import itertools
 import os
+import sys
 import textwrap
 
 import pytest
@@ -396,3 +397,11 @@ def test_install_find_existing_package_canonicalize(script, req1, req2):
     )
     satisfied_message = f"Requirement already satisfied: {req2}"
     assert satisfied_message in result.stdout, str(result)
+
+
+@pytest.mark.skipif(sys.platform != "win32", reason="Windows-only test")
+def test_modifying_pip_presents_error(script):
+    result = script.pip("install", "pip", "--force-reinstall", expect_error=True)
+
+    assert "python.exe" in result.stderr, str(result)
+    assert " -m " in result.stderr, str(result)