From 426d83a86afa628ad25dd3c0ba79ac407696f7c5 Mon Sep 17 00:00:00 2001
From: Pavel Charvat <pchar@ucw.cz>
Date: Fri, 29 Dec 2017 14:45:26 +0100
Subject: [PATCH] UTF-8: Slightly changed sanitization.

The disassembled code from my compiler seems to be OK now.
---
 ucw/unicode.h | 27 ++++++++++++++-------------
 1 file changed, 14 insertions(+), 13 deletions(-)

diff --git a/ucw/unicode.h b/ucw/unicode.h
index 8966167c..d40e8105 100644
--- a/ucw/unicode.h
+++ b/ucw/unicode.h
@@ -282,29 +282,26 @@ static inline void *utf16_be_get(const void *p, uint *uu)
 }
 
 // Internal helper for functions below
-static inline uint unicode_sanitize_char_internal(uint u, uint limit, uint allowed_ctl, uint repl)
+static inline bool unicode_check_char_internal(uint u, uint limit, uint allowed_ctl)
 {
   if (u - 0x20 >= 0x60)			// Otherwise printable ASCII 0x20..0x7f (optimized for this case)
     {
       if (u < 0x20)
 	{
-	  if (!(1U << u) & allowed_ctl)
-	    goto bad;			// Latin-1 control chars except the allowed ones
+	  if (!((1U << u) & allowed_ctl))
+	    return false;		// Latin-1 control chars except the allowed ones
 	}
       else if (u >= 0xd800)
 	{
 	  if (u < 0xf900)		// 0xd800..0xf8ff Surrogate pair
-	    goto bad;
-	  if (u >= limit)
-	    goto bad;			// Outside Unicode range
+	    return false;
+	  if (u > limit)
+	    return false;		// Outside allowed range
 	}
       else if (u < 0xa0)
-	{				// 0x80..0x9f Latin-1 control chars
-bad:
-	  u = repl;
-	}
+	return false;			// 0x80..0x9f Latin-1 control chars
     }
-  return u;
+  return true;
 }
 
 /**
@@ -314,7 +311,9 @@ bad:
  **/
 static inline uint unicode_sanitize_char(uint u)
 {
-  return unicode_sanitize_char_internal(u, 0x10000, 1U<<'\t', UNI_REPLACEMENT);
+  if (!unicode_check_char_internal(u, 0xffff, 1U << '\t'))
+    u = UNI_REPLACEMENT;
+  return u;
 }
 
 /**
@@ -324,7 +323,9 @@ static inline uint unicode_sanitize_char(uint u)
  **/
 static inline uint unicode_full_sanitize_char(uint u)
 {
-  return unicode_sanitize_char_internal(u, 0x110000, 1U<<'\t', UNI_REPLACEMENT);
+  if (!unicode_check_char_internal(u, 0x10ffff, 1U << '\t'))
+    u = UNI_REPLACEMENT;
+  return u;
 }
 
 /* unicode-utf8.c */
-- 
2.39.2