Merge #126

126: Fix num parsing for invalid multi-byte sign chars r=cuviper a=HeroicKatora

Ensure that splitting the potential sign character from the remainder
respects UTF8 boundaries. This lets invalid characters fail correctly
with an error, instead of panicking.

Closes: #125 

Co-authored-by: Andreas Molzer <[email protected]>

src/lib.rs

@@ -216,7 +216,12 @@ macro_rules! float_trait_impl {
                 }
 
                 fn slice_shift_char(src: &str) -> Option<(char, &str)> {
-                    src.chars().nth(0).map(|ch| (ch, &src[1..]))
+                    let mut chars = src.chars();
+                    if let Some(ch) = chars.next() {
+                        Some((ch, chars.as_str()))
+                    } else {
+                        None
+                    }
                 }
 
                 let (is_positive, src) =  match slice_shift_char(src) {
@@ -395,6 +400,15 @@ fn from_str_radix_unwrap() {
     assert_eq!(f, 0.0);
 }
 
+#[test]
+fn from_str_radix_multi_byte_fail() {
+    // Ensure parsing doesn't panic, even on invalid sign characters
+    assert!(f32::from_str_radix("™0.2", 10).is_err());
+
+    // Even when parsing the exponent sign
+    assert!(f32::from_str_radix("0.2E™1", 10).is_err());
+}
+
 #[test]
 fn wrapping_is_num() {
     fn require_num<T: Num>(_: &T) {}