From c9dd9e96a1bc834fad23435242bc83b793d987ed Mon Sep 17 00:00:00 2001
From: Pavel Tyunin <pavel51tunin@gmail.com>
Date: Wed, 7 Oct 2020 17:31:42 +0300
Subject: [PATCH] Allow decoding even shorter strings in fallback encoding

Complete UTF-8 characters (except leading nulls) never appear in failed decoding attempts when the input is fed byte by byte.
---
 src/common/convauto.cpp       | 30 ++++++++++++++----------------
 tests/mbconv/convautotest.cpp |  2 +-
 2 files changed, 15 insertions(+), 17 deletions(-)

diff --git a/src/common/convauto.cpp b/src/common/convauto.cpp
index 8778295207..708096de5d 100644
--- a/src/common/convauto.cpp
+++ b/src/common/convauto.cpp
@@ -267,24 +267,21 @@ bool wxConvAuto::InitFromInput(const char *src, size_t len)
     return true;
 }
 
-// checks if the input can be the beginning of a valid UTF-8 string
-static bool wxIsUTF8Prefix(const char *src, size_t len)
+// checks if the input can be the beginning of a valid UTF-8 sequence
+static bool wxCanBeUTF8SequencePrefix(const char *src, size_t len)
 {
-    unsigned char l;
-    for ( size_t i = 0; i < len; ++i )
+    size_t i = 0;
+    unsigned char l = tableUtf8Lengths[(unsigned char)src[i]];
+    if ( !l )
+        return false; // invalid leading byte
+    while ( --l )
     {
-        l = tableUtf8Lengths[(unsigned char)src[i]];
-        if ( !l )
-            return false; // invalid leading byte
-        while ( --l )
-        {
-            if ( ++i == len )
-                return true; // truncated sequence
-            if ( (src[i] & 0xC0) != 0x80 )
-                return false; // invalid continuation byte
-        }
+        if ( ++i == len )
+            return true; // truncated sequence
+        if ( (src[i] & 0xC0) != 0x80 )
+            return false; // invalid continuation byte
     }
-    return true;
+    return false; // complete sequence
 }
 
 size_t
@@ -339,7 +336,8 @@ wxConvAuto::ToWChar(wchar_t *dst, size_t dstLen,
         size_t nNull = 0;
         if ( srcLen != wxNO_LEN && srcLen >= 2 && !src[0] )
             nNull = ( src[1]? 1 : 2 );
-        if ( srcLen < nNull + m_conv->GetMaxCharLen() && wxIsUTF8Prefix(src, srcLen) )
+        if ( srcLen < nNull + m_conv->GetMaxCharLen() &&
+             wxCanBeUTF8SequencePrefix(src + nNull, srcLen - nNull) )
             return wxCONV_FAILED;
 
         // if the conversion failed but we didn't really detect anything and
diff --git a/tests/mbconv/convautotest.cpp b/tests/mbconv/convautotest.cpp
index 27839f93a5..12e19c21ed 100644
--- a/tests/mbconv/convautotest.cpp
+++ b/tests/mbconv/convautotest.cpp
@@ -226,7 +226,7 @@ void ConvAutoTestCase::FallbackMultibyte()
 
 void ConvAutoTestCase::FallbackShort()
 {
-    TestFirstChar("\x61\x61\x61\xc4", 'a', 4,
+    TestFirstChar("\x61\xc4", 'a', 2,
                   ConvState(wxBOM_None, wxFONTENCODING_ISO8859_5, true),
                   wxFONTENCODING_ISO8859_5);
 }