From 75d063e363f1f6c74c80246ffed71f180b8c10d5 Mon Sep 17 00:00:00 2001 From: Přemysl Janouch Date: Thu, 10 Dec 2015 19:34:10 +0100 Subject: Fix a nasty bug in utf8_next() Uppercase ASCII was read incorrectly. --- liberty.c | 20 +++++++++----------- 1 file changed, 9 insertions(+), 11 deletions(-) diff --git a/liberty.c b/liberty.c index 290d4ef..4373e91 100644 --- a/liberty.c +++ b/liberty.c @@ -2242,14 +2242,11 @@ utf8_next (const char *s, size_t len, int32_t *codepoint) if (!len) return NULL; - // In the middle of a character -> error - const uint8_t *p = (const unsigned char *) s; - if ((*p & 0xC0) == 0x80) - return NULL; + // Find out how long the sequence is (0 for ASCII) + unsigned mask = 0x80; + unsigned sequence_len = 0; - // Find out how long the sequence is - unsigned mask = 0xC0; - unsigned tail_len = 0; + const uint8_t *p = (const uint8_t *) s; while ((*p & mask) == mask) { // Invalid start of sequence @@ -2257,15 +2254,16 @@ utf8_next (const char *s, size_t len, int32_t *codepoint) return NULL; mask |= mask >> 1; - tail_len++; + sequence_len++; } - // Check the rest of the sequence - if (tail_len > --len) + // In the middle of a character or the input is too short + if (sequence_len == 1 || sequence_len > len) return NULL; + // Check the rest of the sequence uint32_t cp = *p++ & ~mask; - while (tail_len--) + while (sequence_len && --sequence_len) { if ((*p & 0xC0) != 0x80) return NULL; -- cgit v1.2.3-70-g09d2