Validate ASCII host values

Host values must match the unreserved, sub-delims or pct-encoded grammar
rules. The IDNA encoder already checks for this, but for ASCII values we
skip IDNA encoding.

If the invalid character is `@`, or `:` followed by `@` further in the
host value, tack on advice about using `authority` instead of `host`.

CHANGES/880.bugfix.rst

@@ -0,0 +1,3 @@
+Started rejecting ASCII hostnames with invalid characters. For host strings that
+look like authority strings, the exception message includes advice on what to do
+instead -- by :user:`mjpieters`.

CHANGES/954.bugfix.rst

@@ -0,0 +1,3 @@
+Started rejecting ASCII hostnames with invalid characters. For host strings that
+look like authority strings, the exception message includes advice on what to do
+instead -- by :user:`mjpieters`.

docs/spelling_wordlist.txt

@@ -18,6 +18,7 @@ downstreams
 facto
 glibc
 google
+hostnames
 mailto
 manylinux
 multi

tests/test_url_build.py

@@ -84,6 +84,25 @@ def test_build_with_authority_and_host():
         URL.build(authority="host.com", host="example.com")
 
 
+@pytest.mark.parametrize(
+    ("host", "is_authority"),
+    [
+        ("user:pass@host.com", True),
+        ("user@host.com", True),
+        ("host:com", False),
+        ("not_percent_encoded%Zf", False),
+        ("still_not_percent_encoded%fZ", False),
+        *(("other_gen_delim_" + c, False) for c in "/?#[]"),
+    ],
+)
+def test_build_with_invalid_host(host: str, is_authority: bool):
+    match = r"Host '[^']+' cannot contain '[^']+' \(at position \d+\)"
+    if is_authority:
+        match += ", if .* use 'authority' instead of 'host'"
+    with pytest.raises(ValueError, match=f"{match}$"):
+        URL.build(host=host)
+
+
 def test_build_with_authority():
     url = URL.build(scheme="http", authority="ваня:bar@host.com:8000", path="path")
     assert str(url) == "http://%D0%B2%D0%B0%D0%BD%D1%8F:bar@host.com:8000/path"

tests/test_url_update_netloc.py

@@ -166,6 +166,26 @@ def test_with_host_non_ascii():
     assert url2.authority == "историк.рф:123"
 
 
+@pytest.mark.parametrize(
+    ("host", "is_authority"),
+    [
+        ("user:pass@host.com", True),
+        ("user@host.com", True),
+        ("host:com", False),
+        ("not_percent_encoded%Zf", False),
+        ("still_not_percent_encoded%fZ", False),
+        *(("other_gen_delim_" + c, False) for c in "/?#[]"),
+    ],
+)
+def test_with_invalid_host(host: str, is_authority: bool):
+    url = URL("http://example.com:123")
+    match = r"Host '[^']+' cannot contain '[^']+' \(at position \d+\)"
+    if is_authority:
+        match += ", if .* use 'authority' instead of 'host'"
+    with pytest.raises(ValueError, match=f"{match}$"):
+        url.with_host(host=host)
+
+
 def test_with_host_percent_encoded():
     url = URL("http://%25cf%2580%cf%80:%25cf%2580%cf%80@example.com:123")
     url2 = url.with_host("%cf%80.org")

yarl/_url.py

@@ -1,5 +1,6 @@
 import functools
 import math
+import re
 import warnings
 from collections.abc import Mapping, Sequence
 from contextlib import suppress
@@ -15,6 +16,22 @@
 
 sentinel = object()
 
+# reg-name: unreserved / pct-encoded / sub-delims
+# this pattern matches anything that is *not* in those classes. and is only used
+# on lower-cased ASCII values.
+_not_reg_name = re.compile(
+    r"""
+        # any character not in the unreserved or sub-delims sets, plus %
+        # (validated with the additional check for pct-encoded sequences below)
+        [^a-z0-9\-._~!$&'()*+,;=%]
+    |
+        # % only allowed if it is part of a pct-encoded
+        # sequence of 2 hex digits.
+        %(?![0-9a-f]{2})
+    """,
+    re.VERBOSE,
+)
+
 
 def rewrite_module(obj: object) -> object:
     obj.__module__ = "yarl"
@@ -765,11 +782,27 @@ def _encode_host(cls, host, human=False):
             ip = ip_address(ip)
         except ValueError:
             host = host.lower()
-            # IDNA encoding is slow,
-            # skip it for ASCII-only strings
+            # skip it for humanized and ASCII-only strings
             # Don't move the check into _idna_encode() helper
             # to reduce the cache size
-            if human or host.isascii():
+            if human:
+                return host
+            if host.isascii():
+                # Check for invalid characters explicitly; _idna_encode() does this
+                # for non-ascii host names.
+                invalid = _not_reg_name.search(host)
+                if invalid is not None:
+                    value, pos, extra = invalid.group(), invalid.start(), ""
+                    if value == "@" or (value == ":" and "@" in host[pos:]):
+                        # this looks like an authority string
+                        extra = (
+                            ", if the value includes a username or password, "
+                            "use 'authority' instead of 'host'"
+                        )
+                    raise ValueError(
+                        f"Host {host!r} cannot contain {value!r} (at position "
+                        f"{pos}){extra}"
+                    ) from None
                 return host
             host = _idna_encode(host)
         else: