Fix validation of overlong UTF-8

It was too strict and Egyptian dicks didn't want to pass,
so we'll do it half-arsedly for a subset.

2 files changed, 6 insertions, 8 deletions

diff --git a/liberty.c b/liberty.c
index 1695fd3..9705081 100644
--- a/liberty.c
+++ b/liberty.c
@@ -2748,16 +2748,12 @@ utf8_decode (const char **s, size_t len)
 	}
 
 	// In the middle of a character
-	if (sequence_len == 1)
+	// or an overlong sequence (subset, possibly MUTF-8, not supported)
+	if (sequence_len == 1 || *p == 0xC0 || *p == 0xC1)
 		return -1;
 
 	// Check the rest of the sequence
 	uint32_t cp = *p++ & ~mask;
-
-	// Overlong sequence (possibly MUTF-8, not supported)
-	if (!cp && sequence_len)
-		return -1;
-
 	while (sequence_len && --sequence_len)
 	{
 		if (p == end)
diff --git a/tests/liberty.c b/tests/liberty.c
index 3f0bd3f..dc445d8 100644
--- a/tests/liberty.c
+++ b/tests/liberty.c
@@ -331,10 +331,12 @@ test_utf8 (void)
 	soft_assert (utf8_decode (&partial, 1) == -2);
 	soft_assert (utf8_decode (&empty,   0) == -1);
 
-	const char valid[] = "2H₂ + O₂ ⇌ 2H₂O, R = 4.7 kΩ, ⌀ 200 mm";
+	const char valid_1[] = "2H₂ + O₂ ⇌ 2H₂O, R = 4.7 kΩ, ⌀ 200 mm";
+	const char valid_2[] = "\xf0\x93\x82\xb9";
 	const char invalid_1[] = "\xf0\x90\x28\xbc";
 	const char invalid_2[] = "\xc0\x80";
-	soft_assert ( utf8_validate (valid,     sizeof valid));
+	soft_assert ( utf8_validate (valid_1,   sizeof valid_1));
+	soft_assert ( utf8_validate (valid_2,   sizeof valid_2));
 	soft_assert (!utf8_validate (invalid_1, sizeof invalid_1));
 	soft_assert (!utf8_validate (invalid_2, sizeof invalid_2));