summaryrefslogtreecommitdiffstats
path: root/third_party/rust/tokio-util/tests/codecs.rs
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-19 01:47:29 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-19 01:47:29 +0000
commit0ebf5bdf043a27fd3dfb7f92e0cb63d88954c44d (patch)
treea31f07c9bcca9d56ce61e9a1ffd30ef350d513aa /third_party/rust/tokio-util/tests/codecs.rs
parentInitial commit. (diff)
downloadfirefox-esr-0ebf5bdf043a27fd3dfb7f92e0cb63d88954c44d.tar.xz
firefox-esr-0ebf5bdf043a27fd3dfb7f92e0cb63d88954c44d.zip
Adding upstream version 115.8.0esr.upstream/115.8.0esr
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'third_party/rust/tokio-util/tests/codecs.rs')
-rw-r--r--third_party/rust/tokio-util/tests/codecs.rs464
1 files changed, 464 insertions, 0 deletions
diff --git a/third_party/rust/tokio-util/tests/codecs.rs b/third_party/rust/tokio-util/tests/codecs.rs
new file mode 100644
index 0000000000..f9a780140a
--- /dev/null
+++ b/third_party/rust/tokio-util/tests/codecs.rs
@@ -0,0 +1,464 @@
+#![warn(rust_2018_idioms)]
+
+use tokio_util::codec::{AnyDelimiterCodec, BytesCodec, Decoder, Encoder, LinesCodec};
+
+use bytes::{BufMut, Bytes, BytesMut};
+
+#[test]
+fn bytes_decoder() {
+ let mut codec = BytesCodec::new();
+ let buf = &mut BytesMut::new();
+ buf.put_slice(b"abc");
+ assert_eq!("abc", codec.decode(buf).unwrap().unwrap());
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"a");
+ assert_eq!("a", codec.decode(buf).unwrap().unwrap());
+}
+
+#[test]
+fn bytes_encoder() {
+ let mut codec = BytesCodec::new();
+
+ // Default capacity of BytesMut
+ #[cfg(target_pointer_width = "64")]
+ const INLINE_CAP: usize = 4 * 8 - 1;
+ #[cfg(target_pointer_width = "32")]
+ const INLINE_CAP: usize = 4 * 4 - 1;
+
+ let mut buf = BytesMut::new();
+ codec
+ .encode(Bytes::from_static(&[0; INLINE_CAP + 1]), &mut buf)
+ .unwrap();
+
+ // Default capacity of Framed Read
+ const INITIAL_CAPACITY: usize = 8 * 1024;
+
+ let mut buf = BytesMut::with_capacity(INITIAL_CAPACITY);
+ codec
+ .encode(Bytes::from_static(&[0; INITIAL_CAPACITY + 1]), &mut buf)
+ .unwrap();
+ codec
+ .encode(BytesMut::from(&b"hello"[..]), &mut buf)
+ .unwrap();
+}
+
+#[test]
+fn lines_decoder() {
+ let mut codec = LinesCodec::new();
+ let buf = &mut BytesMut::new();
+ buf.reserve(200);
+ buf.put_slice(b"line 1\nline 2\r\nline 3\n\r\n\r");
+ assert_eq!("line 1", codec.decode(buf).unwrap().unwrap());
+ assert_eq!("line 2", codec.decode(buf).unwrap().unwrap());
+ assert_eq!("line 3", codec.decode(buf).unwrap().unwrap());
+ assert_eq!("", codec.decode(buf).unwrap().unwrap());
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!(None, codec.decode_eof(buf).unwrap());
+ buf.put_slice(b"k");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!("\rk", codec.decode_eof(buf).unwrap().unwrap());
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!(None, codec.decode_eof(buf).unwrap());
+}
+
+#[test]
+fn lines_decoder_max_length() {
+ const MAX_LENGTH: usize = 6;
+
+ let mut codec = LinesCodec::new_with_max_length(MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"line 1 is too long\nline 2\nline 3\r\nline 4\n\r\n\r");
+
+ assert!(codec.decode(buf).is_err());
+
+ let line = codec.decode(buf).unwrap().unwrap();
+ assert!(
+ line.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ line,
+ MAX_LENGTH
+ );
+ assert_eq!("line 2", line);
+
+ assert!(codec.decode(buf).is_err());
+
+ let line = codec.decode(buf).unwrap().unwrap();
+ assert!(
+ line.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ line,
+ MAX_LENGTH
+ );
+ assert_eq!("line 4", line);
+
+ let line = codec.decode(buf).unwrap().unwrap();
+ assert!(
+ line.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ line,
+ MAX_LENGTH
+ );
+ assert_eq!("", line);
+
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!(None, codec.decode_eof(buf).unwrap());
+ buf.put_slice(b"k");
+ assert_eq!(None, codec.decode(buf).unwrap());
+
+ let line = codec.decode_eof(buf).unwrap().unwrap();
+ assert!(
+ line.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ line,
+ MAX_LENGTH
+ );
+ assert_eq!("\rk", line);
+
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!(None, codec.decode_eof(buf).unwrap());
+
+ // Line that's one character too long. This could cause an out of bounds
+ // error if we peek at the next characters using slice indexing.
+ buf.put_slice(b"aaabbbc");
+ assert!(codec.decode(buf).is_err());
+}
+
+#[test]
+fn lines_decoder_max_length_underrun() {
+ const MAX_LENGTH: usize = 6;
+
+ let mut codec = LinesCodec::new_with_max_length(MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"line ");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"too l");
+ assert!(codec.decode(buf).is_err());
+ buf.put_slice(b"ong\n");
+ assert_eq!(None, codec.decode(buf).unwrap());
+
+ buf.put_slice(b"line 2");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"\n");
+ assert_eq!("line 2", codec.decode(buf).unwrap().unwrap());
+}
+
+#[test]
+fn lines_decoder_max_length_bursts() {
+ const MAX_LENGTH: usize = 10;
+
+ let mut codec = LinesCodec::new_with_max_length(MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"line ");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"too l");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"ong\n");
+ assert!(codec.decode(buf).is_err());
+}
+
+#[test]
+fn lines_decoder_max_length_big_burst() {
+ const MAX_LENGTH: usize = 10;
+
+ let mut codec = LinesCodec::new_with_max_length(MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"line ");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"too long!\n");
+ assert!(codec.decode(buf).is_err());
+}
+
+#[test]
+fn lines_decoder_max_length_newline_between_decodes() {
+ const MAX_LENGTH: usize = 5;
+
+ let mut codec = LinesCodec::new_with_max_length(MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"hello");
+ assert_eq!(None, codec.decode(buf).unwrap());
+
+ buf.put_slice(b"\nworld");
+ assert_eq!("hello", codec.decode(buf).unwrap().unwrap());
+}
+
+// Regression test for [infinite loop bug](https://github.com/tokio-rs/tokio/issues/1483)
+#[test]
+fn lines_decoder_discard_repeat() {
+ const MAX_LENGTH: usize = 1;
+
+ let mut codec = LinesCodec::new_with_max_length(MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"aa");
+ assert!(codec.decode(buf).is_err());
+ buf.put_slice(b"a");
+ assert_eq!(None, codec.decode(buf).unwrap());
+}
+
+// Regression test for [subsequent calls to LinesCodec decode does not return the desired results bug](https://github.com/tokio-rs/tokio/issues/3555)
+#[test]
+fn lines_decoder_max_length_underrun_twice() {
+ const MAX_LENGTH: usize = 11;
+
+ let mut codec = LinesCodec::new_with_max_length(MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"line ");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"too very l");
+ assert!(codec.decode(buf).is_err());
+ buf.put_slice(b"aaaaaaaaaaaaaaaaaaaaaaa");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"ong\nshort\n");
+ assert_eq!("short", codec.decode(buf).unwrap().unwrap());
+}
+
+#[test]
+fn lines_encoder() {
+ let mut codec = LinesCodec::new();
+ let mut buf = BytesMut::new();
+
+ codec.encode("line 1", &mut buf).unwrap();
+ assert_eq!("line 1\n", buf);
+
+ codec.encode("line 2", &mut buf).unwrap();
+ assert_eq!("line 1\nline 2\n", buf);
+}
+
+#[test]
+fn any_delimiters_decoder_any_character() {
+ let mut codec = AnyDelimiterCodec::new(b",;\n\r".to_vec(), b",".to_vec());
+ let buf = &mut BytesMut::new();
+ buf.reserve(200);
+ buf.put_slice(b"chunk 1,chunk 2;chunk 3\n\r");
+ assert_eq!("chunk 1", codec.decode(buf).unwrap().unwrap());
+ assert_eq!("chunk 2", codec.decode(buf).unwrap().unwrap());
+ assert_eq!("chunk 3", codec.decode(buf).unwrap().unwrap());
+ assert_eq!("", codec.decode(buf).unwrap().unwrap());
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!(None, codec.decode_eof(buf).unwrap());
+ buf.put_slice(b"k");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!("k", codec.decode_eof(buf).unwrap().unwrap());
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!(None, codec.decode_eof(buf).unwrap());
+}
+
+#[test]
+fn any_delimiters_decoder_max_length() {
+ const MAX_LENGTH: usize = 7;
+
+ let mut codec =
+ AnyDelimiterCodec::new_with_max_length(b",;\n\r".to_vec(), b",".to_vec(), MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"chunk 1 is too long\nchunk 2\nchunk 3\r\nchunk 4\n\r\n");
+
+ assert!(codec.decode(buf).is_err());
+
+ let chunk = codec.decode(buf).unwrap().unwrap();
+ assert!(
+ chunk.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ chunk,
+ MAX_LENGTH
+ );
+ assert_eq!("chunk 2", chunk);
+
+ let chunk = codec.decode(buf).unwrap().unwrap();
+ assert!(
+ chunk.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ chunk,
+ MAX_LENGTH
+ );
+ assert_eq!("chunk 3", chunk);
+
+ // \r\n cause empty chunk
+ let chunk = codec.decode(buf).unwrap().unwrap();
+ assert!(
+ chunk.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ chunk,
+ MAX_LENGTH
+ );
+ assert_eq!("", chunk);
+
+ let chunk = codec.decode(buf).unwrap().unwrap();
+ assert!(
+ chunk.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ chunk,
+ MAX_LENGTH
+ );
+ assert_eq!("chunk 4", chunk);
+
+ let chunk = codec.decode(buf).unwrap().unwrap();
+ assert!(
+ chunk.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ chunk,
+ MAX_LENGTH
+ );
+ assert_eq!("", chunk);
+
+ let chunk = codec.decode(buf).unwrap().unwrap();
+ assert!(
+ chunk.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ chunk,
+ MAX_LENGTH
+ );
+ assert_eq!("", chunk);
+
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!(None, codec.decode_eof(buf).unwrap());
+ buf.put_slice(b"k");
+ assert_eq!(None, codec.decode(buf).unwrap());
+
+ let chunk = codec.decode_eof(buf).unwrap().unwrap();
+ assert!(
+ chunk.len() <= MAX_LENGTH,
+ "{:?}.len() <= {:?}",
+ chunk,
+ MAX_LENGTH
+ );
+ assert_eq!("k", chunk);
+
+ assert_eq!(None, codec.decode(buf).unwrap());
+ assert_eq!(None, codec.decode_eof(buf).unwrap());
+
+ // Delimiter that's one character too long. This could cause an out of bounds
+ // error if we peek at the next characters using slice indexing.
+ buf.put_slice(b"aaabbbcc");
+ assert!(codec.decode(buf).is_err());
+}
+
+#[test]
+fn any_delimiter_decoder_max_length_underrun() {
+ const MAX_LENGTH: usize = 7;
+
+ let mut codec =
+ AnyDelimiterCodec::new_with_max_length(b",;\n\r".to_vec(), b",".to_vec(), MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"chunk ");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"too l");
+ assert!(codec.decode(buf).is_err());
+ buf.put_slice(b"ong\n");
+ assert_eq!(None, codec.decode(buf).unwrap());
+
+ buf.put_slice(b"chunk 2");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b",");
+ assert_eq!("chunk 2", codec.decode(buf).unwrap().unwrap());
+}
+
+#[test]
+fn any_delimiter_decoder_max_length_underrun_twice() {
+ const MAX_LENGTH: usize = 11;
+
+ let mut codec =
+ AnyDelimiterCodec::new_with_max_length(b",;\n\r".to_vec(), b",".to_vec(), MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"chunk ");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"too very l");
+ assert!(codec.decode(buf).is_err());
+ buf.put_slice(b"aaaaaaaaaaaaaaaaaaaaaaa");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"ong\nshort\n");
+ assert_eq!("short", codec.decode(buf).unwrap().unwrap());
+}
+#[test]
+fn any_delimiter_decoder_max_length_bursts() {
+ const MAX_LENGTH: usize = 11;
+
+ let mut codec =
+ AnyDelimiterCodec::new_with_max_length(b",;\n\r".to_vec(), b",".to_vec(), MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"chunk ");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"too l");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"ong\n");
+ assert!(codec.decode(buf).is_err());
+}
+
+#[test]
+fn any_delimiter_decoder_max_length_big_burst() {
+ const MAX_LENGTH: usize = 11;
+
+ let mut codec =
+ AnyDelimiterCodec::new_with_max_length(b",;\n\r".to_vec(), b",".to_vec(), MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"chunk ");
+ assert_eq!(None, codec.decode(buf).unwrap());
+ buf.put_slice(b"too long!\n");
+ assert!(codec.decode(buf).is_err());
+}
+
+#[test]
+fn any_delimiter_decoder_max_length_delimiter_between_decodes() {
+ const MAX_LENGTH: usize = 5;
+
+ let mut codec =
+ AnyDelimiterCodec::new_with_max_length(b",;\n\r".to_vec(), b",".to_vec(), MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"hello");
+ assert_eq!(None, codec.decode(buf).unwrap());
+
+ buf.put_slice(b",world");
+ assert_eq!("hello", codec.decode(buf).unwrap().unwrap());
+}
+
+#[test]
+fn any_delimiter_decoder_discard_repeat() {
+ const MAX_LENGTH: usize = 1;
+
+ let mut codec =
+ AnyDelimiterCodec::new_with_max_length(b",;\n\r".to_vec(), b",".to_vec(), MAX_LENGTH);
+ let buf = &mut BytesMut::new();
+
+ buf.reserve(200);
+ buf.put_slice(b"aa");
+ assert!(codec.decode(buf).is_err());
+ buf.put_slice(b"a");
+ assert_eq!(None, codec.decode(buf).unwrap());
+}
+
+#[test]
+fn any_delimiter_encoder() {
+ let mut codec = AnyDelimiterCodec::new(b",".to_vec(), b";--;".to_vec());
+ let mut buf = BytesMut::new();
+
+ codec.encode("chunk 1", &mut buf).unwrap();
+ assert_eq!("chunk 1;--;", buf);
+
+ codec.encode("chunk 2", &mut buf).unwrap();
+ assert_eq!("chunk 1;--;chunk 2;--;", buf);
+}