Fix handling invalid utf8 characters in http paths (#492)

* Possible fix for handling invalid http paths

* Try to use utf8_lossy_decoding and add dev dependencies

* Attempt to fix the tests

* Review tests

* Apply utf8 lossy patch for path also to RSGI

---------

Co-authored-by: Giovanni Barillari <giovanni.barillari@sentry.io>

src/asgi/utils.rs

@@ -17,7 +17,7 @@ macro_rules! scope_native_parts {
             .uri
             .path_and_query()
             .map_or_else(|| ("", ""), |pq| (pq.path(), pq.query().unwrap_or("")));
-        let $path = percent_encoding::percent_decode_str(path_raw).decode_utf8().unwrap();
+        let $path = percent_encoding::percent_decode_str(path_raw).decode_utf8_lossy();
         let $version = match $req.version {
             hyper::Version::HTTP_10 => "1",
             hyper::Version::HTTP_11 => "1.1",

src/rsgi/types.rs

@@ -184,8 +184,8 @@ macro_rules! rsgi_scope_cls {
             }
 
             #[getter(path)]
-            fn get_path(&self) -> Result<Cow<str>> {
-                Ok(percent_decode_str(self.uri.path()).decode_utf8()?)
+            fn get_path(&self) -> Cow<str> {
+                percent_decode_str(self.uri.path()).decode_utf8_lossy()
             }
 
             #[getter(query_string)]

tests/apps/asgi.py

@@ -174,5 +174,5 @@ def app(scope, receive, send):
         '/err_app': err_app,
         '/err_proto': err_proto,
         '/timeout_n': timeout_n,
-        '/timeout_w': timeout_w,
-    }[scope['path']](scope, receive, send)
+        '/timeout_w': timeout_w
+    }.get(scope['path'], info)(scope, receive, send)

tests/test_asgi.py

@@ -27,6 +27,18 @@ async def test_scope(asgi_server, threading_mode):
     assert data['state']['global'] == 'test'
 
 
+@pytest.mark.asyncio
+@pytest.mark.skipif(bool(os.getenv('PGO_RUN')), reason='PGO build')
+@pytest.mark.parametrize('threading_mode', ['runtime', 'workers'])
+async def test_scope_path_utf8_lossy(asgi_server, threading_mode):
+    async with asgi_server(threading_mode) as port:
+        res = httpx.get(f'http://localhost:{port}/%c0')
+
+    assert res.status_code == 200
+    data = res.json()
+    assert data['path'] == '/\ufffd'
+
+
 @pytest.mark.asyncio
 @pytest.mark.parametrize('threading_mode', ['runtime', 'workers'])
 async def test_body(asgi_server, threading_mode):