If TCL_UTF_MAX>=4, make Tcl_ParseBackslash combine two surrogates so they appear as one 4-byte UTF-8 byte sequence from the start. Add test-case for this.

3 files changed, 12 insertions, 2 deletions

diff --git a/generic/tclParse.c b/generic/tclParse.c
index 1532c05..4f30f8b 100644
--- a/generic/tclParse.c
+++ b/generic/tclParse.c
@@ -920,6 +920,16 @@ TclParseBackslash(
 	     * No hexadigits -> This is just "u".
 	     */
 	    result = 'u';
+#if TCL_UTF_MAX > 3
+	} else if (((result & 0xDC00) == 0xD800) && (count == 6) && (p[5] == '\\') && (p[6] == 'u') && (numBytes >= 10)) {
+	    /* If high surrogate is immediately followed by a low surrogate escape, combine them. */
+	    int low;
+	    int count2 = TclParseHex(p+7, 4, &low);
+	    if ((low & 0xDC00) == 0xDC00) {
+		result = ((result & 0x3FF)<<10 | (low & 0x3FF)) + 0x10000;
+		count += count2 + 2;
+	    }
+#endif
 	}
 	break;
     case 'U':
diff --git a/generic/tclUtf.c b/generic/tclUtf.c
index 9c2ef03..ce80bd0 100644
--- a/generic/tclUtf.c
+++ b/generic/tclUtf.c
@@ -846,7 +846,7 @@ Tcl_UtfBackslash(
 	 * We ate a whole line. Pay the price of a strlen()
 	 */
 
-	result = TclParseBackslash(src, (int)strlen(src), &numRead, dst);
+	result = TclParseBackslash(src, strlen(src), &numRead, dst);
     }
     if (readPtr != NULL) {
 	*readPtr = numRead;
diff --git a/generic/tclUtil.c b/generic/tclUtil.c
index 61c1973..41b3481 100644
--- a/generic/tclUtil.c
+++ b/generic/tclUtil.c
@@ -1649,7 +1649,7 @@ Tcl_Backslash(
     int *readPtr)		/* Fill in with number of characters read from
 				 * src, unless NULL. */
 {
-    char buf[TCL_UTF_MAX] = "";
+    char buf[4] = "";
     Tcl_UniChar ch = 0;
 
     Tcl_UtfBackslash(src, readPtr, buf);