From 96397778144722f7983774e9fc88521f4d36d3c7 Mon Sep 17 00:00:00 2001
From: Přemysl Eric Janouch <p@janouch.name>
Date: Sat, 24 Oct 2020 19:06:41 +0200
Subject: 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.
---
 liberty.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

(limited to 'liberty.c')

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)
-- 
cgit v1.2.3-70-g09d2