denyhosts/clamav/libclamav_rust/.cargo/vendor/proc-macro2/tests/test.rs

646 lines
19 KiB
Rust
Raw Normal View History

2023-01-14 18:28:39 +08:00
#![allow(clippy::assertions_on_result_states, clippy::non_ascii_literal)]
use proc_macro2::{Ident, Literal, Punct, Spacing, Span, TokenStream, TokenTree};
use std::panic;
use std::str::{self, FromStr};
#[test]
fn idents() {
assert_eq!(
Ident::new("String", Span::call_site()).to_string(),
"String"
);
assert_eq!(Ident::new("fn", Span::call_site()).to_string(), "fn");
assert_eq!(Ident::new("_", Span::call_site()).to_string(), "_");
}
#[test]
fn raw_idents() {
assert_eq!(
Ident::new_raw("String", Span::call_site()).to_string(),
"r#String"
);
assert_eq!(Ident::new_raw("fn", Span::call_site()).to_string(), "r#fn");
}
#[test]
#[should_panic(expected = "`r#_` cannot be a raw identifier")]
fn ident_raw_underscore() {
Ident::new_raw("_", Span::call_site());
}
#[test]
#[should_panic(expected = "`r#super` cannot be a raw identifier")]
fn ident_raw_reserved() {
Ident::new_raw("super", Span::call_site());
}
#[test]
#[should_panic(expected = "Ident is not allowed to be empty; use Option<Ident>")]
fn ident_empty() {
Ident::new("", Span::call_site());
}
#[test]
#[should_panic(expected = "Ident cannot be a number; use Literal instead")]
fn ident_number() {
Ident::new("255", Span::call_site());
}
#[test]
#[should_panic(expected = "\"a#\" is not a valid Ident")]
fn ident_invalid() {
Ident::new("a#", Span::call_site());
}
#[test]
#[should_panic(expected = "not a valid Ident")]
fn raw_ident_empty() {
Ident::new("r#", Span::call_site());
}
#[test]
#[should_panic(expected = "not a valid Ident")]
fn raw_ident_number() {
Ident::new("r#255", Span::call_site());
}
#[test]
#[should_panic(expected = "\"r#a#\" is not a valid Ident")]
fn raw_ident_invalid() {
Ident::new("r#a#", Span::call_site());
}
#[test]
#[should_panic(expected = "not a valid Ident")]
fn lifetime_empty() {
Ident::new("'", Span::call_site());
}
#[test]
#[should_panic(expected = "not a valid Ident")]
fn lifetime_number() {
Ident::new("'255", Span::call_site());
}
#[test]
fn lifetime_invalid() {
let result = panic::catch_unwind(|| Ident::new("'a#", Span::call_site()));
match result {
Err(box_any) => {
let message = box_any.downcast_ref::<String>().unwrap();
let expected1 = r#""\'a#" is not a valid Ident"#; // 1.31.0 .. 1.53.0
let expected2 = r#""'a#" is not a valid Ident"#; // 1.53.0 ..
assert!(
message == expected1 || message == expected2,
"panic message does not match expected string\n\
\x20 panic message: `{:?}`\n\
\x20expected message: `{:?}`",
message,
expected2,
);
}
Ok(_) => panic!("test did not panic as expected"),
}
}
#[test]
fn literal_string() {
assert_eq!(Literal::string("foo").to_string(), "\"foo\"");
assert_eq!(Literal::string("\"").to_string(), "\"\\\"\"");
assert_eq!(Literal::string("didn't").to_string(), "\"didn't\"");
}
#[test]
fn literal_raw_string() {
"r\"\r\n\"".parse::<TokenStream>().unwrap();
}
#[test]
fn literal_byte_string() {
assert_eq!(Literal::byte_string(b"").to_string(), "b\"\"");
assert_eq!(
Literal::byte_string(b"\0\t\n\r\"\\2\x10").to_string(),
"b\"\\0\\t\\n\\r\\\"\\\\2\\x10\"",
);
}
#[test]
fn literal_character() {
assert_eq!(Literal::character('x').to_string(), "'x'");
assert_eq!(Literal::character('\'').to_string(), "'\\''");
assert_eq!(Literal::character('"').to_string(), "'\"'");
}
#[test]
fn literal_integer() {
assert_eq!(Literal::u8_suffixed(10).to_string(), "10u8");
assert_eq!(Literal::u16_suffixed(10).to_string(), "10u16");
assert_eq!(Literal::u32_suffixed(10).to_string(), "10u32");
assert_eq!(Literal::u64_suffixed(10).to_string(), "10u64");
assert_eq!(Literal::u128_suffixed(10).to_string(), "10u128");
assert_eq!(Literal::usize_suffixed(10).to_string(), "10usize");
assert_eq!(Literal::i8_suffixed(10).to_string(), "10i8");
assert_eq!(Literal::i16_suffixed(10).to_string(), "10i16");
assert_eq!(Literal::i32_suffixed(10).to_string(), "10i32");
assert_eq!(Literal::i64_suffixed(10).to_string(), "10i64");
assert_eq!(Literal::i128_suffixed(10).to_string(), "10i128");
assert_eq!(Literal::isize_suffixed(10).to_string(), "10isize");
assert_eq!(Literal::u8_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::u16_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::u32_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::u64_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::u128_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::usize_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::i8_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::i16_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::i32_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::i64_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::i128_unsuffixed(10).to_string(), "10");
assert_eq!(Literal::isize_unsuffixed(10).to_string(), "10");
}
#[test]
fn literal_float() {
assert_eq!(Literal::f32_suffixed(10.0).to_string(), "10f32");
assert_eq!(Literal::f64_suffixed(10.0).to_string(), "10f64");
assert_eq!(Literal::f32_unsuffixed(10.0).to_string(), "10.0");
assert_eq!(Literal::f64_unsuffixed(10.0).to_string(), "10.0");
}
#[test]
fn literal_suffix() {
fn token_count(p: &str) -> usize {
p.parse::<TokenStream>().unwrap().into_iter().count()
}
assert_eq!(token_count("999u256"), 1);
assert_eq!(token_count("999r#u256"), 3);
assert_eq!(token_count("1."), 1);
assert_eq!(token_count("1.f32"), 3);
assert_eq!(token_count("1.0_0"), 1);
assert_eq!(token_count("1._0"), 3);
assert_eq!(token_count("1._m"), 3);
assert_eq!(token_count("\"\"s"), 1);
assert_eq!(token_count("r\"\"r"), 1);
assert_eq!(token_count("b\"\"b"), 1);
assert_eq!(token_count("br\"\"br"), 1);
assert_eq!(token_count("r#\"\"#r"), 1);
assert_eq!(token_count("'c'c"), 1);
assert_eq!(token_count("b'b'b"), 1);
assert_eq!(token_count("0E"), 1);
assert_eq!(token_count("0o0A"), 1);
assert_eq!(token_count("0E--0"), 4);
assert_eq!(token_count("0.0ECMA"), 1);
}
#[test]
fn literal_iter_negative() {
let negative_literal = Literal::i32_suffixed(-3);
let tokens = TokenStream::from(TokenTree::Literal(negative_literal));
let mut iter = tokens.into_iter();
match iter.next().unwrap() {
TokenTree::Punct(punct) => {
assert_eq!(punct.as_char(), '-');
assert_eq!(punct.spacing(), Spacing::Alone);
}
unexpected => panic!("unexpected token {:?}", unexpected),
}
match iter.next().unwrap() {
TokenTree::Literal(literal) => {
assert_eq!(literal.to_string(), "3i32");
}
unexpected => panic!("unexpected token {:?}", unexpected),
}
assert!(iter.next().is_none());
}
#[test]
fn literal_parse() {
assert!("1".parse::<Literal>().is_ok());
assert!("-1".parse::<Literal>().is_ok());
assert!("-1u12".parse::<Literal>().is_ok());
assert!("1.0".parse::<Literal>().is_ok());
assert!("-1.0".parse::<Literal>().is_ok());
assert!("-1.0f12".parse::<Literal>().is_ok());
assert!("'a'".parse::<Literal>().is_ok());
assert!("\"\n\"".parse::<Literal>().is_ok());
assert!("0 1".parse::<Literal>().is_err());
assert!(" 0".parse::<Literal>().is_err());
assert!("0 ".parse::<Literal>().is_err());
assert!("/* comment */0".parse::<Literal>().is_err());
assert!("0/* comment */".parse::<Literal>().is_err());
assert!("0// comment".parse::<Literal>().is_err());
assert!("- 1".parse::<Literal>().is_err());
assert!("- 1.0".parse::<Literal>().is_err());
assert!("-\"\"".parse::<Literal>().is_err());
}
#[test]
fn roundtrip() {
fn roundtrip(p: &str) {
println!("parse: {}", p);
let s = p.parse::<TokenStream>().unwrap().to_string();
println!("first: {}", s);
let s2 = s.parse::<TokenStream>().unwrap().to_string();
assert_eq!(s, s2);
}
roundtrip("a");
roundtrip("<<");
roundtrip("<<=");
roundtrip(
"
1
1.0
1f32
2f64
1usize
4isize
4e10
1_000
1_0i32
8u8
9
0
0xffffffffffffffffffffffffffffffff
1x
1u80
1f320
",
);
roundtrip("'a");
roundtrip("'_");
roundtrip("'static");
roundtrip("'\\u{10__FFFF}'");
roundtrip("\"\\u{10_F0FF__}foo\\u{1_0_0_0__}\"");
}
#[test]
fn fail() {
fn fail(p: &str) {
if let Ok(s) = p.parse::<TokenStream>() {
panic!("should have failed to parse: {}\n{:#?}", p, s);
}
}
fail("' static");
fail("r#1");
fail("r#_");
fail("\"\\u{0000000}\""); // overlong unicode escape (rust allows at most 6 hex digits)
fail("\"\\u{999999}\""); // outside of valid range of char
fail("\"\\u{_0}\""); // leading underscore
fail("\"\\u{}\""); // empty
fail("b\"\r\""); // bare carriage return in byte string
fail("r\"\r\""); // bare carriage return in raw string
fail("\"\\\r \""); // backslash carriage return
fail("'aa'aa");
fail("br##\"\"#");
fail("\"\\\n\u{85}\r\"");
}
#[cfg(span_locations)]
#[test]
fn span_test() {
check_spans(
"\
/// This is a document comment
testing 123
{
testing 234
}",
&[
(1, 0, 1, 30), // #
(1, 0, 1, 30), // [ ... ]
(1, 0, 1, 30), // doc
(1, 0, 1, 30), // =
(1, 0, 1, 30), // "This is..."
(2, 0, 2, 7), // testing
(2, 8, 2, 11), // 123
(3, 0, 5, 1), // { ... }
(4, 2, 4, 9), // testing
(4, 10, 4, 13), // 234
],
);
}
#[cfg(procmacro2_semver_exempt)]
#[cfg(not(nightly))]
#[test]
fn default_span() {
let start = Span::call_site().start();
assert_eq!(start.line, 1);
assert_eq!(start.column, 0);
let end = Span::call_site().end();
assert_eq!(end.line, 1);
assert_eq!(end.column, 0);
let source_file = Span::call_site().source_file();
assert_eq!(source_file.path().to_string_lossy(), "<unspecified>");
assert!(!source_file.is_real());
}
#[cfg(procmacro2_semver_exempt)]
#[test]
fn span_join() {
let source1 = "aaa\nbbb"
.parse::<TokenStream>()
.unwrap()
.into_iter()
.collect::<Vec<_>>();
let source2 = "ccc\nddd"
.parse::<TokenStream>()
.unwrap()
.into_iter()
.collect::<Vec<_>>();
assert!(source1[0].span().source_file() != source2[0].span().source_file());
assert_eq!(
source1[0].span().source_file(),
source1[1].span().source_file()
);
let joined1 = source1[0].span().join(source1[1].span());
let joined2 = source1[0].span().join(source2[0].span());
assert!(joined1.is_some());
assert!(joined2.is_none());
let start = joined1.unwrap().start();
let end = joined1.unwrap().end();
assert_eq!(start.line, 1);
assert_eq!(start.column, 0);
assert_eq!(end.line, 2);
assert_eq!(end.column, 3);
assert_eq!(
joined1.unwrap().source_file(),
source1[0].span().source_file()
);
}
#[test]
fn no_panic() {
let s = str::from_utf8(b"b\'\xc2\x86 \x00\x00\x00^\"").unwrap();
assert!(s.parse::<TokenStream>().is_err());
}
#[test]
fn punct_before_comment() {
let mut tts = TokenStream::from_str("~// comment").unwrap().into_iter();
match tts.next().unwrap() {
TokenTree::Punct(tt) => {
assert_eq!(tt.as_char(), '~');
assert_eq!(tt.spacing(), Spacing::Alone);
}
wrong => panic!("wrong token {:?}", wrong),
}
}
#[test]
fn joint_last_token() {
// This test verifies that we match the behavior of libproc_macro *not* in
// the range nightly-2020-09-06 through nightly-2020-09-10, in which this
// behavior was temporarily broken.
// See https://github.com/rust-lang/rust/issues/76399
let joint_punct = Punct::new(':', Spacing::Joint);
let stream = TokenStream::from(TokenTree::Punct(joint_punct));
let punct = match stream.into_iter().next().unwrap() {
TokenTree::Punct(punct) => punct,
_ => unreachable!(),
};
assert_eq!(punct.spacing(), Spacing::Joint);
}
#[test]
fn raw_identifier() {
let mut tts = TokenStream::from_str("r#dyn").unwrap().into_iter();
match tts.next().unwrap() {
TokenTree::Ident(raw) => assert_eq!("r#dyn", raw.to_string()),
wrong => panic!("wrong token {:?}", wrong),
}
assert!(tts.next().is_none());
}
#[test]
fn test_debug_ident() {
let ident = Ident::new("proc_macro", Span::call_site());
#[cfg(not(span_locations))]
let expected = "Ident(proc_macro)";
#[cfg(span_locations)]
let expected = "Ident { sym: proc_macro }";
assert_eq!(expected, format!("{:?}", ident));
}
#[test]
fn test_debug_tokenstream() {
let tts = TokenStream::from_str("[a + 1]").unwrap();
#[cfg(not(span_locations))]
let expected = "\
TokenStream [
Group {
delimiter: Bracket,
stream: TokenStream [
Ident {
sym: a,
},
Punct {
char: '+',
spacing: Alone,
},
Literal {
lit: 1,
},
],
},
]\
";
#[cfg(not(span_locations))]
let expected_before_trailing_commas = "\
TokenStream [
Group {
delimiter: Bracket,
stream: TokenStream [
Ident {
sym: a
},
Punct {
char: '+',
spacing: Alone
},
Literal {
lit: 1
}
]
}
]\
";
#[cfg(span_locations)]
let expected = "\
TokenStream [
Group {
delimiter: Bracket,
stream: TokenStream [
Ident {
sym: a,
span: bytes(2..3),
},
Punct {
char: '+',
spacing: Alone,
span: bytes(4..5),
},
Literal {
lit: 1,
span: bytes(6..7),
},
],
span: bytes(1..8),
},
]\
";
#[cfg(span_locations)]
let expected_before_trailing_commas = "\
TokenStream [
Group {
delimiter: Bracket,
stream: TokenStream [
Ident {
sym: a,
span: bytes(2..3)
},
Punct {
char: '+',
spacing: Alone,
span: bytes(4..5)
},
Literal {
lit: 1,
span: bytes(6..7)
}
],
span: bytes(1..8)
}
]\
";
let actual = format!("{:#?}", tts);
if actual.ends_with(",\n]") {
assert_eq!(expected, actual);
} else {
assert_eq!(expected_before_trailing_commas, actual);
}
}
#[test]
fn default_tokenstream_is_empty() {
let default_token_stream = <TokenStream as Default>::default();
assert!(default_token_stream.is_empty());
}
#[test]
fn tokenstream_size_hint() {
let tokens = "a b (c d) e".parse::<TokenStream>().unwrap();
assert_eq!(tokens.into_iter().size_hint(), (4, Some(4)));
}
#[test]
fn tuple_indexing() {
// This behavior may change depending on https://github.com/rust-lang/rust/pull/71322
let mut tokens = "tuple.0.0".parse::<TokenStream>().unwrap().into_iter();
assert_eq!("tuple", tokens.next().unwrap().to_string());
assert_eq!(".", tokens.next().unwrap().to_string());
assert_eq!("0.0", tokens.next().unwrap().to_string());
assert!(tokens.next().is_none());
}
#[cfg(span_locations)]
#[test]
fn non_ascii_tokens() {
check_spans("// abc", &[]);
check_spans("// ábc", &[]);
check_spans("// abc x", &[]);
check_spans("// ábc x", &[]);
check_spans("/* abc */ x", &[(1, 10, 1, 11)]);
check_spans("/* ábc */ x", &[(1, 10, 1, 11)]);
check_spans("/* ab\nc */ x", &[(2, 5, 2, 6)]);
check_spans("/* áb\nc */ x", &[(2, 5, 2, 6)]);
check_spans("/*** abc */ x", &[(1, 12, 1, 13)]);
check_spans("/*** ábc */ x", &[(1, 12, 1, 13)]);
check_spans(r#""abc""#, &[(1, 0, 1, 5)]);
check_spans(r#""ábc""#, &[(1, 0, 1, 5)]);
check_spans(r###"r#"abc"#"###, &[(1, 0, 1, 8)]);
check_spans(r###"r#"ábc"#"###, &[(1, 0, 1, 8)]);
check_spans("r#\"a\nc\"#", &[(1, 0, 2, 3)]);
check_spans("r#\"á\nc\"#", &[(1, 0, 2, 3)]);
check_spans("'a'", &[(1, 0, 1, 3)]);
check_spans("'á'", &[(1, 0, 1, 3)]);
check_spans("//! abc", &[(1, 0, 1, 7), (1, 0, 1, 7), (1, 0, 1, 7)]);
check_spans("//! ábc", &[(1, 0, 1, 7), (1, 0, 1, 7), (1, 0, 1, 7)]);
check_spans("//! abc\n", &[(1, 0, 1, 7), (1, 0, 1, 7), (1, 0, 1, 7)]);
check_spans("//! ábc\n", &[(1, 0, 1, 7), (1, 0, 1, 7), (1, 0, 1, 7)]);
check_spans("/*! abc */", &[(1, 0, 1, 10), (1, 0, 1, 10), (1, 0, 1, 10)]);
check_spans("/*! ábc */", &[(1, 0, 1, 10), (1, 0, 1, 10), (1, 0, 1, 10)]);
check_spans("/*! a\nc */", &[(1, 0, 2, 4), (1, 0, 2, 4), (1, 0, 2, 4)]);
check_spans("/*! á\nc */", &[(1, 0, 2, 4), (1, 0, 2, 4), (1, 0, 2, 4)]);
check_spans("abc", &[(1, 0, 1, 3)]);
check_spans("ábc", &[(1, 0, 1, 3)]);
check_spans("ábć", &[(1, 0, 1, 3)]);
check_spans("abc// foo", &[(1, 0, 1, 3)]);
check_spans("ábc// foo", &[(1, 0, 1, 3)]);
check_spans("ábć// foo", &[(1, 0, 1, 3)]);
check_spans("b\"a\\\n c\"", &[(1, 0, 2, 3)]);
check_spans("b\"a\\\n\u{00a0}c\"", &[(1, 0, 2, 3)]);
}
#[cfg(span_locations)]
fn check_spans(p: &str, mut lines: &[(usize, usize, usize, usize)]) {
let ts = p.parse::<TokenStream>().unwrap();
check_spans_internal(ts, &mut lines);
assert!(lines.is_empty(), "leftover ranges: {:?}", lines);
}
#[cfg(span_locations)]
fn check_spans_internal(ts: TokenStream, lines: &mut &[(usize, usize, usize, usize)]) {
for i in ts {
if let Some((&(sline, scol, eline, ecol), rest)) = lines.split_first() {
*lines = rest;
let start = i.span().start();
assert_eq!(start.line, sline, "sline did not match for {}", i);
assert_eq!(start.column, scol, "scol did not match for {}", i);
let end = i.span().end();
assert_eq!(end.line, eline, "eline did not match for {}", i);
assert_eq!(end.column, ecol, "ecol did not match for {}", i);
if let TokenTree::Group(g) = i {
check_spans_internal(g.stream().clone(), lines);
}
}
}
}
#[test]
fn byte_order_mark() {
let string = "\u{feff}foo";
let tokens = string.parse::<TokenStream>().unwrap();
match tokens.into_iter().next().unwrap() {
TokenTree::Ident(ident) => assert_eq!(ident, "foo"),
_ => unreachable!(),
}
let string = "foo\u{feff}";
string.parse::<TokenStream>().unwrap_err();
}