Better - more complete - fix for [d433c0e0ad]: TCL_UTF_MAX == 4 problems. It allows emoji to be produced by the system encoding, even for other values of TCL_UTF_MAX.

Also added test-cases for this.

2 files changed, 41 insertions, 12 deletions

diff --git a/generic/tclEncoding.c b/generic/tclEncoding.c
index f159b32..00b97f5 100644
--- a/generic/tclEncoding.c
+++ b/generic/tclEncoding.c
@@ -2155,7 +2155,7 @@ BinaryProc(
 /*
  *-------------------------------------------------------------------------
  *
- * UtfExtToUtfIntProc --
+ * UtfIntToUtfExtProc --
  *
  *	Convert from UTF-8 to UTF-8. While converting null-bytes from the
  *	Tcl's internal representation (0xc0, 0x80) to the official
@@ -2296,7 +2296,7 @@ UtfToUtfProc(
 				 * output buffer. */
     int pureNullMode)		/* Convert embedded nulls from internal
 				 * representation to real null-bytes or vice
-				 * versa. */
+				 * versa. Also combine or separate surrogate pairs */
 {
     const char *srcStart, *srcEnd, *srcClose;
     const char *dstStart, *dstEnd;
@@ -2312,14 +2312,14 @@ UtfToUtfProc(
     srcEnd = src + srcLen;
     srcClose = srcEnd;
     if ((flags & TCL_ENCODING_END) == 0) {
-	srcClose -= TCL_UTF_MAX;
+	srcClose -= 6;
     }
     if (flags & TCL_ENCODING_CHAR_LIMIT) {
 	charLimit = *dstCharsPtr;
     }
 
     dstStart = dst;
-    dstEnd = dst + dstLen - TCL_UTF_MAX;
+    dstEnd = dst + dstLen - ((pureNullMode == 1) ? 4 : TCL_UTF_MAX);
 
     for (numChars = 0; src < srcEnd && numChars <= charLimit; numChars++) {
 	if ((src > srcClose) && (!Tcl_UtfCharComplete(src, srcEnd - src))) {
@@ -2361,15 +2361,28 @@ UtfToUtfProc(
 	    src += 1;
 	    dst += Tcl_UniCharToUtf(*chPtr, dst);
 	} else {
-	    int len = TclUtfToUniChar(src, chPtr);
-	    src += len;
-	    dst += Tcl_UniCharToUtf(*chPtr, dst);
-#if TCL_UTF_MAX == 4
-	    if ((*chPtr >= 0xD800) && (len < 3)) {
-		src += Tcl_UtfToUniChar(src, chPtr);
-		dst += Tcl_UniCharToUtf(*chPtr, dst);
+	    src += TclUtfToUniChar(src, chPtr);
+	    if ((*chPtr & 0xFC00) == 0xD800) {
+		/* A high surrogate character is detected, handle especially */
+		Tcl_UniChar low = *chPtr;
+		size_t len = Tcl_UtfToUniChar(src, &low);
+		if ((low & 0xFC00) != 0xDC00) {
+			*dst++ = (char) (((*chPtr >> 12) | 0xE0) & 0xEF);
+			*dst++ = (char) (((*chPtr >> 6) | 0x80) & 0xBF);
+			*dst++ = (char) ((*chPtr | 0x80) & 0xBF);
+			continue;
+		} else if (pureNullMode == 1) {
+		    int full = (((*chPtr & 0x3FF) << 10) | (low & 0x3FF)) + 0x10000;
+		    *dst++ = (char) (((full >> 18) | 0xF0) & 0xF7);
+		    *dst++ = (char) (((full >> 12) | 0x80) & 0xBF);
+		    *dst++ = (char) (((full >> 6) | 0x80) & 0xBF);
+		    *dst++ = (char) ((full | 0x80) & 0xBF);
+			*chPtr = 0;
+		    src += len;
+		    continue;
+		}
 	    }
-#endif
+	    dst += Tcl_UniCharToUtf(*chPtr, dst);
 	}
     }
 
diff --git a/tests/encoding.test b/tests/encoding.test
index ed0e6a4..cf27190 100644
--- a/tests/encoding.test
+++ b/tests/encoding.test
@@ -328,6 +328,22 @@ test encoding-15.3 {UtfToUtfProc null character input} {
     binary scan [encoding convertto identity $y] H* z
     list [string bytelength $x] [string bytelength $y] $z
 } {1 2 c080}
+test encoding-15.4 {UtfToUtfProc emoji character input} {
+    set x \xED\xA0\xBD\xED\xB8\x82
+    set y [encoding convertfrom utf-8 \xED\xA0\xBD\xED\xB8\x82]
+    list [string length $x] [string length $y] $y
+} "6 2 \uD83D\uDE02"
+test encoding-15.5 {UtfToUtfProc emoji character input} {
+    set x \xF0\x9F\x98\x82
+    set y [encoding convertfrom utf-8 \xF0\x9F\x98\x82]
+    list [string length $x] [string length $y] $y
+} "4 2 \uD83D\uDE02"
+test encoding-15.6 {UtfToUtfProc emoji character output} {
+    set x \uD83D\uDE02
+    set y [encoding convertto utf-8 \uD83D\uDE02]
+    binary scan $y H* z
+    list [string length $x] [string length $y] $z
+} {2 4 f09f9882}
 
 test encoding-16.1 {UnicodeToUtfProc} {
     set val [encoding convertfrom unicode NN]