TclTrim() can also demand NUL-terminated arguments, and be simplified.

1 files changed, 21 insertions, 64 deletions

diff --git a/generic/tclUtil.c b/generic/tclUtil.c
index b6eebdc..82ef9b7 100644
--- a/generic/tclUtil.c
+++ b/generic/tclUtil.c
@@ -1516,42 +1516,6 @@ Tcl_Backslash(
 /*
  *----------------------------------------------------------------------
  *
- * UtfWellFormedEnd --
- *	Checks the end of utf string is malformed, if yes - wraps bytes
- *	to the given buffer (as well-formed NTS string).  The buffer
- *	argument should be initialized by the caller and ready to use.
- *
- * Results:
- *	The bytes with well-formed end of the string.
- *
- * Side effects:
- *	Buffer (DString) may be allocated, so must be released.
- *
- *----------------------------------------------------------------------
- */
-
-static inline const char*
-UtfWellFormedEnd(
-    Tcl_DString *buffer,	/* Buffer used to hold well-formed string. */
-    CONST char *bytes,		/* Pointer to the beginning of the string. */
-    int length)			/* Length of the string. */
-{
-    CONST char *l = bytes + length;
-    CONST char *p = Tcl_UtfPrev(l, bytes);
-
-    if (Tcl_UtfCharComplete(p, l - p)) {
-	return bytes;
-    }
-    /* 
-     * Malformed utf-8 end, be sure we've NTS to safe compare of end-character,
-     * avoid segfault by access violation out of range.
-     */
-    Tcl_DStringAppend(buffer, bytes, length);
-    return Tcl_DStringValue(buffer);
-}
-/*
- *----------------------------------------------------------------------
- *
  * TclTrimRight --
  *	Takes two counted strings in the Tcl encoding.  Conceptually
  *	finds the sub string (offset) to trim from the right side of the
@@ -1712,39 +1676,32 @@ TclTrim(
     int numTrim,	/* ...and its length in bytes */
 			/* Calls in this routine
 			 * rely on (trim[numTrim] == '\0'). */
-    int *trimRight)	/* Offset from the end of the string. */
+    int *trimRightPtr)	/* Offset from the end of the string. */
 {
-    int trimLeft;
-    Tcl_DString bytesBuf, trimBuf;
+    int trimLeft = 0, trimRight = 0;
 
-    *trimRight = 0;
     /* Empty strings -> nothing to do */
-    if ((numBytes == 0) || (numTrim == 0)) {
-	return 0;
-    }
-
-    Tcl_DStringInit(&bytesBuf);
-    Tcl_DStringInit(&trimBuf);
-    bytes = UtfWellFormedEnd(&bytesBuf, bytes, numBytes);
-    trim = UtfWellFormedEnd(&trimBuf, trim, numTrim);
-
-    trimLeft = TclTrimLeft(bytes, numBytes, trim, numTrim);
-    if (trimLeft > numBytes) {
-	trimLeft = numBytes;
-    }
-    numBytes -= trimLeft;
-    /* have to trim yet (first char was already verified within TrimLeft) */
-    if (numBytes > 1) {
-	bytes += trimLeft;
-	*trimRight = TclTrimRight(bytes, numBytes, trim, numTrim);
-	if (*trimRight > numBytes) {
-	    *trimRight = numBytes;
+    if ((numBytes > 0) && (numTrim > 0)) {
+
+	/* When bytes is NUL-terminated, returns 0 <= trimLeft <= numBytes */
+	trimLeft = TclTrimLeft(bytes, numBytes, trim, numTrim);
+	numBytes -= trimLeft;
+
+	/* If we did not trim the whole string, it starts with a character
+	 * that we will not trim. Skip over it. */
+	if (numBytes > 0) {
+	    const char *first = bytes + trimLeft;
+	    bytes = Tcl_UtfNext(first);
+	    numBytes -= (bytes - first);
+
+	    if (numBytes > 0) {
+		/* When bytes is NUL-terminated, returns
+		 * 0 <= trimRight <= numBytes */
+		trimRight = TclTrimRight(bytes, numBytes, trim, numTrim);
+	    }
 	}
     }
-
-    Tcl_DStringFree(&bytesBuf);
-    Tcl_DStringFree(&trimBuf);
-
+    *trimRightPtr = trimRight;
     return trimLeft;
 }