comparing ascii instead of bytes

pyproject.toml

@@ -10,7 +10,7 @@ build-backend = "setuptools.build_meta"
 write_to = "src/_pytest/_version.py"
 
 [tool.pytest.ini_options]
-minversion = "2.0"
+#minversion = "2.0"
 addopts = "-rfEX -p pytester --strict-markers"
 python_files = ["test_*.py", "*_test.py", "testing/python/*.py"]
 python_classes = ["Test", "Acceptance"]

src/_pytest/_io/saferepr.py

@@ -94,7 +94,7 @@ def safeformat(obj: object) -> str:
 DEFAULT_REPR_MAX_SIZE = 240
 
 
-def saferepr(obj: object, maxsize: Optional[int] = DEFAULT_REPR_MAX_SIZE) -> str:
+def saferepr(obj: object, maxsize: Optional[int] = DEFAULT_REPR_MAX_SIZE, use_ascii: bool = False) -> str:
     """Return a size-limited safe repr-string for the given object.
 
     Failing __repr__ functions of user instances will be represented
@@ -104,10 +104,14 @@ def saferepr(obj: object, maxsize: Optional[int] = DEFAULT_REPR_MAX_SIZE) -> str
     This function is a wrapper around the Repr/reprlib functionality of the
     stdlib.
     """
+
+    if use_ascii:
+        obj = ascii(obj)
+
     return SafeRepr(maxsize).repr(obj)
 
 
-def saferepr_unlimited(obj: object) -> str:
+def saferepr_unlimited(obj: object, use_ascii: bool = True) -> str:
     """Return an unlimited-size safe repr-string for the given object.
 
     As with saferepr, failing __repr__ functions of user instances
@@ -119,6 +123,8 @@ def saferepr_unlimited(obj: object) -> str:
     when maxsize=None, but that might affect some other code.
     """
     try:
+        if use_ascii:
+            obj = ascii(obj)
         return repr(obj)
     except Exception as exc:
         return _format_repr_exception(exc, obj)

src/_pytest/assertion/util.py

@@ -157,20 +157,24 @@ def has_default_eq(
     return True
 
 
-def assertrepr_compare(config, op: str, left: Any, right: Any) -> Optional[List[str]]:
+def assertrepr_compare(config, op: str, left: Any, right: Any, use_ascii: bool=False) -> Optional[List[str]]:
     """Return specialised explanations for some operators/operands."""
     verbose = config.getoption("verbose")
+
+    use_ascii = isinstance(left, str) and isinstance(right, str) and normalize("NFD", left) == normalize("NFD", right)
+
     if verbose > 1:
-        left_repr = saferepr_unlimited(left)
+        left_repr = saferepr_unlimited(left, use_ascii=use_ascii)
-        right_repr = saferepr_unlimited(right)
+        right_repr = saferepr_unlimited(right, use_ascii=use_ascii)
     else:
         # XXX: "15 chars indentation" is wrong
         #      ("E       AssertionError: assert "); should use term width.
         maxsize = (
             80 - 15 - len(op) - 2
         ) // 2  # 15 chars indentation, 1 space around op
-        left_repr = saferepr(left, maxsize=maxsize)
+
-        right_repr = saferepr(right, maxsize=maxsize)
+        left_repr = saferepr(left, maxsize=maxsize, use_ascii=use_ascii)
+        right_repr = saferepr(right, maxsize=maxsize, use_ascii=use_ascii)
 
     summary = f"{left_repr} {op} {right_repr}"
 
@@ -242,15 +246,6 @@ def _diff_text(left: str, right: str, verbose: int = 0) -> List[str]:
 
     explanation: List[str] = []
 
-    # if both are strings are truthy and normalize to the same string, but we're here,
-    # their byte representation is different. We convert to utf-8 and compare them as bytes.
-
-    if left and right and normalize("NFD", left) == normalize("NFD", right):
-        return [
-            "Strings are different but normalize to the same string. Comparing their utf-8 encoding.",
-            *_compare_eq_sequence(left.encode(), right.encode()),
-        ]
-
     if verbose < 1:
         i = 0  # just in case left or right has zero length
         for i in range(min(len(left), len(right))):

testing/test_assertion.py

@@ -781,11 +781,7 @@ class TestAssert_reprcompare:
         left = "hyv\xe4"
         right = "hyva\u0308"
         expl = callequal(left, right)
-        assert expl == [
+        assert expl == ['"\'hyv\\\\xe4\'" == "\'hyva\\\\u0308\'"', f'- {str(right)}', f'+ {str(left)}']
-            "'hyvä' == 'hyvä'",
-            "Strings are different but normalize to the same string. Comparing their utf-8 encoding.",
-            "At index 3 diff: b'\\xc3' != b'a'",
-        ]
 
 
 class TestAssert_reprcompare_dataclass: