Attachment "3173086-2.patch" to
ticket [3173086fff]
added by
dgp
2011-02-18 04:09:50.
Index: generic/tclDictObj.c
===================================================================
RCS file: /cvsroot/tcl/tcl/generic/tclDictObj.c,v
retrieving revision 1.56.2.4
diff -u -r1.56.2.4 tclDictObj.c
--- generic/tclDictObj.c 19 May 2010 21:47:49 -0000 1.56.2.4
+++ generic/tclDictObj.c 17 Feb 2011 21:04:23 -0000
@@ -461,19 +461,27 @@
Tcl_Obj *dictPtr)
{
#define LOCAL_SIZE 20
- int localFlags[LOCAL_SIZE], *flagPtr;
+ int localFlags[LOCAL_SIZE], *flagPtr = NULL;
Dict *dict = dictPtr->internalRep.otherValuePtr;
ChainEntry *cPtr;
Tcl_Obj *keyPtr, *valuePtr;
- int numElems, i, length;
+ int i, length, bytesNeeded = 0;
char *elem, *dst;
+ const int maxFlags = UINT_MAX / sizeof(int);
/*
* This field is the most useful one in the whole hash structure, and it
* is not exposed by any API function...
*/
- numElems = dict->table.numEntries * 2;
+ int numElems = dict->table.numEntries * 2;
+
+ /* Handle empty list case first, simplifies what follows */
+ if (numElems == 0) {
+ dictPtr->bytes = tclEmptyStringRep;
+ dictPtr->length = 0;
+ return;
+ }
/*
* Pass 1: estimate space, gather flags.
@@ -481,55 +489,63 @@
if (numElems <= LOCAL_SIZE) {
flagPtr = localFlags;
+ } else if (numElems > maxFlags) {
+ Tcl_Panic("UpdateStringOfDict: size requirement exceeds limits");
} else {
flagPtr = (int *) ckalloc((unsigned) numElems*sizeof(int));
}
- dictPtr->length = 1;
for (i=0,cPtr=dict->entryChainHead; i<numElems; i+=2,cPtr=cPtr->nextPtr) {
/*
* Assume that cPtr is never NULL since we know the number of array
* elements already.
*/
+ flagPtr[i] = ( i ? TCL_DONT_QUOTE_HASH : 0 );
keyPtr = (Tcl_Obj *) Tcl_GetHashKey(&dict->table, &cPtr->entry);
elem = TclGetStringFromObj(keyPtr, &length);
- dictPtr->length += Tcl_ScanCountedElement(elem, length,
- &flagPtr[i]) + 1;
+ bytesNeeded += TclScanElement(elem, length, flagPtr+i);
+ if (bytesNeeded < 0) {
+ Tcl_Panic("UpdateStringOfDict: size requirement exceeds limits");
+ }
+ flagPtr[i+1] = TCL_DONT_QUOTE_HASH;
valuePtr = Tcl_GetHashValue(&cPtr->entry);
elem = TclGetStringFromObj(valuePtr, &length);
- dictPtr->length += Tcl_ScanCountedElement(elem, length,
- &flagPtr[i+1]) + 1;
+ bytesNeeded += TclScanElement(elem, length, flagPtr+i+1);
+ if (bytesNeeded < 0) {
+ Tcl_Panic("UpdateStringOfDict: size requirement exceeds limits");
+ }
}
+ if (bytesNeeded > INT_MAX - numElems + 1) {
+ Tcl_Panic("UpdateStringOfDict: size requirement exceeds limits");
+ }
+ bytesNeeded += numElems;
/*
* Pass 2: copy into string rep buffer.
*/
- dictPtr->bytes = ckalloc((unsigned) dictPtr->length);
+ dictPtr->length = bytesNeeded - 1;
+ dictPtr->bytes = ckalloc((unsigned) bytesNeeded);
dst = dictPtr->bytes;
for (i=0,cPtr=dict->entryChainHead; i<numElems; i+=2,cPtr=cPtr->nextPtr) {
+ flagPtr[i] |= ( i ? TCL_DONT_QUOTE_HASH : 0 );
keyPtr = (Tcl_Obj *) Tcl_GetHashKey(&dict->table, &cPtr->entry);
elem = TclGetStringFromObj(keyPtr, &length);
- dst += Tcl_ConvertCountedElement(elem, length, dst,
- flagPtr[i] | (i==0 ? 0 : TCL_DONT_QUOTE_HASH));
- *(dst++) = ' ';
+ dst += TclConvertElement(elem, length, dst, flagPtr[i]);
+ *dst++ = ' ';
+ flagPtr[i+1] |= TCL_DONT_QUOTE_HASH;
valuePtr = Tcl_GetHashValue(&cPtr->entry);
elem = TclGetStringFromObj(valuePtr, &length);
- dst += Tcl_ConvertCountedElement(elem, length, dst,
- flagPtr[i+1] | TCL_DONT_QUOTE_HASH);
- *(dst++) = ' ';
+ dst += TclConvertElement(elem, length, dst, flagPtr[i+1]);
+ *dst++ = ' ';
}
+ dictPtr->bytes[dictPtr->length] = '\0';
+
if (flagPtr != localFlags) {
ckfree((char *) flagPtr);
}
- if (dst == dictPtr->bytes) {
- *dst = 0;
- } else {
- *(--dst) = 0;
- }
- dictPtr->length = dst - dictPtr->bytes;
}
/*
Index: generic/tclIndexObj.c
===================================================================
RCS file: /cvsroot/tcl/tcl/generic/tclIndexObj.c,v
retrieving revision 1.38
diff -u -r1.38 tclIndexObj.c
--- generic/tclIndexObj.c 13 Dec 2007 15:23:18 -0000 1.38
+++ generic/tclIndexObj.c 17 Feb 2011 21:04:23 -0000
@@ -536,12 +536,13 @@
} else {
elementStr = TclGetStringFromObj(origObjv[i], &elemLen);
}
- len = Tcl_ScanCountedElement(elementStr, elemLen, &flags);
+ flags = 0;
+ len = TclScanElement(elementStr, elemLen, &flags);
if (MAY_QUOTE_WORD && len != elemLen) {
char *quotedElementStr = TclStackAlloc(interp, (unsigned)len);
- len = Tcl_ConvertCountedElement(elementStr, elemLen,
+ len = TclConvertElement(elementStr, elemLen,
quotedElementStr, flags);
Tcl_AppendToObj(objPtr, quotedElementStr, len);
TclStackFree(interp, quotedElementStr);
@@ -590,12 +591,13 @@
*/
elementStr = TclGetStringFromObj(objv[i], &elemLen);
- len = Tcl_ScanCountedElement(elementStr, elemLen, &flags);
+ flags = 0;
+ len = TclScanElement(elementStr, elemLen, &flags);
if (MAY_QUOTE_WORD && len != elemLen) {
char *quotedElementStr = TclStackAlloc(interp,(unsigned) len);
- len = Tcl_ConvertCountedElement(elementStr, elemLen,
+ len = TclConvertElement(elementStr, elemLen,
quotedElementStr, flags);
Tcl_AppendToObj(objPtr, quotedElementStr, len);
TclStackFree(interp, quotedElementStr);
Index: generic/tclInt.h
===================================================================
RCS file: /cvsroot/tcl/tcl/generic/tclInt.h,v
retrieving revision 1.362.2.17
diff -u -r1.362.2.17 tclInt.h
--- generic/tclInt.h 30 Nov 2010 20:59:27 -0000 1.362.2.17
+++ generic/tclInt.h 17 Feb 2011 21:04:23 -0000
@@ -2560,6 +2560,8 @@
MODULE_SCOPE ContLineLoc* TclContinuationsGet(Tcl_Obj *objPtr);
MODULE_SCOPE void TclContinuationsCopy(Tcl_Obj *objPtr,
Tcl_Obj *originObjPtr);
+MODULE_SCOPE int TclConvertElement(CONST char *src, int length,
+ char *dst, int flags);
MODULE_SCOPE void TclDeleteNamespaceVars(Namespace *nsPtr);
/* TIP #280 - Modified token based evulation, with line information. */
MODULE_SCOPE int TclEvalEx(Tcl_Interp *interp, const char *script,
@@ -2768,6 +2770,8 @@
MODULE_SCOPE void TclRemoveScriptLimitCallbacks(Tcl_Interp *interp);
MODULE_SCOPE int TclReToGlob(Tcl_Interp *interp, const char *reStr,
int reStrLen, Tcl_DString *dsPtr, int *flagsPtr);
+MODULE_SCOPE int TclScanElement(CONST char *string, int length,
+ int *flagPtr);
MODULE_SCOPE void TclSetBgErrorHandler(Tcl_Interp *interp,
Tcl_Obj *cmdPrefix);
MODULE_SCOPE void TclSetBignumIntRep(Tcl_Obj *objPtr,
Index: generic/tclListObj.c
===================================================================
RCS file: /cvsroot/tcl/tcl/generic/tclListObj.c,v
retrieving revision 1.49.2.4
diff -u -r1.49.2.4 tclListObj.c
--- generic/tclListObj.c 19 May 2010 22:04:48 -0000 1.49.2.4
+++ generic/tclListObj.c 17 Feb 2011 21:04:23 -0000
@@ -1844,74 +1844,73 @@
Tcl_Obj *listPtr) /* List object with string rep to update. */
{
# define LOCAL_SIZE 20
- int localFlags[LOCAL_SIZE], *flagPtr;
+ int localFlags[LOCAL_SIZE], *flagPtr = NULL;
List *listRepPtr = (List *) listPtr->internalRep.twoPtrValue.ptr1;
int numElems = listRepPtr->elemCount;
- register int i;
+ int i, length, bytesNeeded = 0;
char *elem, *dst;
- int length;
Tcl_Obj **elemPtrs;
+ const int maxFlags = UINT_MAX / sizeof(int);
/*
- * Convert each element of the list to string form and then convert it to
- * proper list element form, adding it to the result buffer.
+ * Mark the list as being canonical; although it will now have a string
+ * rep, it is one we derived through proper "canonical" quoting and so
+ * it's known to be free from nasties relating to [concat] and [eval].
*/
+ listRepPtr->canonicalFlag = 1;
+
+ /* Handle empty list case first, so rest of the routine is simpler */
+
+ if (numElems == 0) {
+ listPtr->bytes = tclEmptyStringRep;
+ listPtr->length = 0;
+ return;
+ }
+
/*
* Pass 1: estimate space, gather flags.
*/
if (numElems <= LOCAL_SIZE) {
flagPtr = localFlags;
+ } else if (numElems > maxFlags) {
+ Tcl_Panic("UpdateStringOfList: size requirement exceeds limits");
} else {
flagPtr = (int *) ckalloc((unsigned) numElems * sizeof(int));
}
- listPtr->length = 1;
elemPtrs = &listRepPtr->elements;
for (i = 0; i < numElems; i++) {
+ flagPtr[i] = ( i ? TCL_DONT_QUOTE_HASH : 0 );
elem = TclGetStringFromObj(elemPtrs[i], &length);
- listPtr->length += Tcl_ScanCountedElement(elem, length, flagPtr+i)+1;
-
- /*
- * Check for continued sanity. [Bug 1267380]
- */
-
- if (listPtr->length < 1) {
- Tcl_Panic("string representation size exceeds sane bounds");
+ bytesNeeded += TclScanElement(elem, length, flagPtr+i);
+ if (bytesNeeded < 0) {
+ Tcl_Panic("UpdateStringOfList: size requirement exceeds limits");
}
}
+ if (bytesNeeded > INT_MAX - numElems + 1) {
+ Tcl_Panic("UpdateStringOfList: size requirement exceeds limits");
+ }
+ bytesNeeded += numElems;
/*
* Pass 2: copy into string rep buffer.
*/
- listPtr->bytes = ckalloc((unsigned) listPtr->length);
+ listPtr->length = bytesNeeded - 1;
+ listPtr->bytes = ckalloc((unsigned) bytesNeeded);
dst = listPtr->bytes;
for (i = 0; i < numElems; i++) {
+ flagPtr[i] |= ( i ? TCL_DONT_QUOTE_HASH : 0 );
elem = TclGetStringFromObj(elemPtrs[i], &length);
- dst += Tcl_ConvertCountedElement(elem, length, dst,
- flagPtr[i] | (i==0 ? 0 : TCL_DONT_QUOTE_HASH));
- *dst = ' ';
- dst++;
+ dst += TclConvertElement(elem, length, dst, flagPtr[i]);
+ *dst++ = ' ';
}
+ listPtr->bytes[listPtr->length] = '\0';
+
if (flagPtr != localFlags) {
ckfree((char *) flagPtr);
}
- if (dst == listPtr->bytes) {
- *dst = 0;
- } else {
- dst--;
- *dst = 0;
- }
- listPtr->length = dst - listPtr->bytes;
-
- /*
- * Mark the list as being canonical; although it has a string rep, it is
- * one we derived through proper "canonical" quoting and so it's known to
- * be free from nasties relating to [concat] and [eval].
- */
-
- listRepPtr->canonicalFlag = 1;
}
/*
Index: generic/tclUtil.c
===================================================================
RCS file: /cvsroot/tcl/tcl/generic/tclUtil.c,v
retrieving revision 1.97.2.10
diff -u -r1.97.2.10 tclUtil.c
--- generic/tclUtil.c 15 Jan 2011 19:07:01 -0000 1.97.2.10
+++ generic/tclUtil.c 17 Feb 2011 21:04:25 -0000
@@ -41,8 +41,6 @@
* USE_BRACES - 1 means the string contains a special
* character that can be handled simply by
* enclosing the entire argument in braces.
- * BRACES_UNMATCHED - 1 means that braces aren't properly matched in
- * the argument.
* TCL_DONT_QUOTE_HASH - 1 means the caller insists that a leading hash
* character ('#') should *not* be quoted. This
* is appropriate when the caller can guarantee
@@ -52,7 +50,6 @@
*/
#define USE_BRACES 2
-#define BRACES_UNMATCHED 4
/*
* The following key is used by Tcl_PrintDouble and TclPrecTraceProc to
@@ -650,6 +647,15 @@
*----------------------------------------------------------------------
*/
+#define COMPAT 1
+
+#define CONVERT_NONE 0
+#define CONVERT_BRACE 2
+#define CONVERT_ESCAPE 4
+#define CONVERT_SAFE 16
+#define CONVERT_MASK (CONVERT_BRACE | CONVERT_ESCAPE)
+
+
int
Tcl_ScanCountedElement(
CONST char *string, /* String to convert to Tcl list element. */
@@ -657,74 +663,73 @@
int *flagPtr) /* Where to store information to guide
* Tcl_ConvertElement. */
{
- int flags, nestingLevel;
- register CONST char *p, *lastChar;
+ int flags = CONVERT_SAFE;
+ int numBytes = TclScanElement(string, length, &flags);
- /*
- * This function and Tcl_ConvertElement together do two things:
- *
- * 1. They produce a proper list, one that will yield back the argument
- * strings when evaluated or when disassembled with Tcl_SplitList. This
- * is the most important thing.
- *
- * 2. They try to produce legible output, which means minimizing the use
- * of backslashes (using braces instead). However, there are some
- * situations where backslashes must be used (e.g. an element like
- * "{abc": the leading brace will have to be backslashed. For each
- * element, one of three things must be done:
- *
- * (a) Use the element as-is (it doesn't contain any special
- * characters). This is the most desirable option.
- *
- * (b) Enclose the element in braces, but leave the contents alone.
- * This happens if the element contains embedded space, or if it
- * contains characters with special interpretation ($, [, ;, or \),
- * or if it starts with a brace or double-quote, or if there are no
- * characters in the element.
- *
- * (c) Don't enclose the element in braces, but add backslashes to
- * prevent special interpretation of special characters. This is a
- * last resort used when the argument would normally fall under
- * case (b) but contains unmatched braces. It also occurs if the
- * last character of the argument is a backslash or if the element
- * contains a backslash followed by newline.
- *
- * The function figures out how many bytes will be needed to store the
- * result (actually, it overestimates). It also collects information about
- * the element in the form of a flags word.
- *
- * Note: list elements produced by this function and
- * Tcl_ConvertCountedElement must have the property that they can be
- * enclosing in curly braces to make sub-lists. This means, for example,
- * that we must not leave unmatched curly braces in the resulting list
- * element. This property is necessary in order for functions like
- * Tcl_DStringStartSublist to work.
- */
-
- nestingLevel = 0;
- flags = 0;
- if (string == NULL) {
- string = "";
- }
- if (length == -1) {
- length = strlen(string);
- }
- lastChar = string + length;
- p = string;
- if ((p == lastChar) || (*p == '{') || (*p == '"')) {
- flags |= USE_BRACES;
+ *flagPtr = flags;
+ return numBytes;
+}
+
+int
+TclScanElement(
+ CONST char *string, /* String to convert to Tcl list element. */
+ int length, /* Number of bytes in string, or -1. */
+ int *flagPtr) /* Where to store information to guide
+ * Tcl_ConvertElement. */
+{
+ CONST char *p = string;
+ int nestingLevel = 0;
+ int forbidNone = 0;
+ int requireEscape = 0;
+ int extra = 0;
+ int bytesNeeded;
+#if COMPAT
+ int preferEscape = 0;
+ int preferBrace = 0;
+ int braceCount = 0;
+#endif
+
+ if ((p == NULL) || (length == 0)) {
+ *flagPtr = CONVERT_BRACE;
+ return 2;
+ }
+
+ if ((*p == '{') || (*p == '"') || ((*p == '\0') && (length == -1))) {
+ forbidNone = 1;
+#if COMPAT
+ preferBrace = 1;
+#endif
}
- for (; p < lastChar; p++) {
+
+ while (length) {
switch (*p) {
case '{':
+#if COMPAT
+ braceCount++;
+#endif
+ extra++;
nestingLevel++;
break;
case '}':
+#if COMPAT
+ braceCount++;
+#endif
+ extra++;
nestingLevel--;
if (nestingLevel < 0) {
- flags |= TCL_DONT_USE_BRACES|BRACES_UNMATCHED;
+ requireEscape = 1;
}
break;
+ case ']':
+ case '"':
+#if COMPAT
+ forbidNone = 1;
+ extra++;
+ preferEscape = 1;
+ break;
+#else
+ /* FLOW THROUGH */
+#endif
case '[':
case '$':
case ';':
@@ -734,32 +739,103 @@
case '\r':
case '\t':
case '\v':
- flags |= USE_BRACES;
+ forbidNone = 1;
+ extra++;
+#if COMPAT
+ preferBrace = 1;
+#endif
break;
case '\\':
- if ((p+1 == lastChar) || (p[1] == '\n')) {
- flags = TCL_DONT_USE_BRACES | BRACES_UNMATCHED;
- } else {
- int size;
-
- Tcl_UtfBackslash(p, &size, NULL);
- p += size-1;
- flags |= USE_BRACES;
+ extra++;
+ if ((length == 1) || ((length == -1) && (p[1] == '\0'))) {
+ requireEscape = 1;
+ break;
}
+ if (p[1] == '\n') {
+ extra++;
+ requireEscape = 1;
+ length -= (length > 0);
+ p++;
+ break;
+ }
+ if ((p[1] == '{') || (p[1] == '}')) {
+ extra++;
+ length -= (length > 0);
+ p++;
+ }
+ forbidNone = 1;
+#if COMPAT
+ preferBrace = 1;
+#endif
+ break;
+ case '\0':
+ if (length == -1) {
+ goto endOfString;
+ }
+ /* TODO: Panic on improper encoding? */
break;
}
+ length -= (length > 0);
+ p++;
}
+
+ endOfString:
if (nestingLevel != 0) {
- flags = TCL_DONT_USE_BRACES | BRACES_UNMATCHED;
+ requireEscape = 1;
}
- *flagPtr = flags;
- /*
- * Allow enough space to backslash every character plus leave two spaces
- * for braces.
- */
+ bytesNeeded = p - string;
- return 2*(p-string) + 2;
+ if (requireEscape) {
+ bytesNeeded += extra;
+ if ((*string == '#') && !(*flagPtr & TCL_DONT_QUOTE_HASH)) {
+ bytesNeeded++;
+ }
+ *flagPtr = CONVERT_ESCAPE;
+ goto overflowCheck;
+ }
+ if (*flagPtr & CONVERT_SAFE) {
+ if (extra < 2) {
+ extra = 2;
+ }
+ *flagPtr |= TCL_DONT_USE_BRACES;
+ }
+ if (forbidNone) {
+#if COMPAT
+ if (preferEscape && !preferBrace) {
+ bytesNeeded += (extra - braceCount);
+ if ((*string == '#') && !(*flagPtr & TCL_DONT_QUOTE_HASH)) {
+ bytesNeeded++;
+ }
+ if (*flagPtr & TCL_DONT_USE_BRACES) {
+ bytesNeeded += braceCount;
+ }
+ *flagPtr = CONVERT_MASK;
+ goto overflowCheck;
+ }
+#endif
+ if (*flagPtr & TCL_DONT_USE_BRACES) {
+ bytesNeeded += extra;
+ if ((*string == '#') && !(*flagPtr & TCL_DONT_QUOTE_HASH)) {
+ bytesNeeded++;
+ }
+ } else {
+ bytesNeeded += 2;
+ }
+ *flagPtr = CONVERT_BRACE;
+ goto overflowCheck;
+ }
+
+ if ((*string == '#') && !(*flagPtr & TCL_DONT_QUOTE_HASH)) {
+ bytesNeeded += 2;
+ }
+ *flagPtr = CONVERT_NONE;
+
+ overflowCheck:
+ if (bytesNeeded < 0) {
+ Tcl_Panic("TclScanElement: string length overflow");
+ }
+ return bytesNeeded;
}
/*
@@ -820,125 +896,139 @@
char *dst, /* Place to put list-ified element. */
int flags) /* Flags produced by Tcl_ScanElement. */
{
- register char *p = dst;
- register CONST char *lastChar;
+ int numBytes = TclConvertElement(src, length, dst, flags);
+ dst[numBytes] = '\0';
+ return numBytes;
+}
- /*
- * See the comment block at the beginning of the Tcl_ScanElement code for
- * details of how this works.
- */
+int TclConvertElement(
+ register CONST char *src, /* Source information for list element. */
+ int length, /* Number of bytes in src, or -1. */
+ char *dst, /* Place to put list-ified element. */
+ int flags) /* Flags produced by Tcl_ScanElement. */
+{
+ int conversion = flags & CONVERT_MASK;
+ char *p = dst;
- if (src && length == -1) {
- length = strlen(src);
+ if ((flags & TCL_DONT_USE_BRACES) && (conversion & CONVERT_BRACE)) {
+ conversion = CONVERT_ESCAPE;
}
if ((src == NULL) || (length == 0)) {
- p[0] = '{';
- p[1] = '}';
- p[2] = 0;
- return 2;
+ src = tclEmptyStringRep;
+ length = 0;
+ conversion = CONVERT_BRACE;
}
- lastChar = src + length;
if ((*src == '#') && !(flags & TCL_DONT_QUOTE_HASH)) {
- flags |= USE_BRACES;
+ if (conversion == CONVERT_ESCAPE) {
+ p[0] = '\\';
+ p[1] = '#';
+ p += 2;
+ src++;
+ length--;
+ } else {
+ conversion = CONVERT_BRACE;
+ }
}
- if ((flags & USE_BRACES) && !(flags & TCL_DONT_USE_BRACES)) {
+ if (conversion == CONVERT_NONE) {
+ if (length == -1) {
+ /* TODO: INT_MAX overflow? */
+ while (*src) {
+ *p++ = *src++;
+ }
+ return p - dst;
+ } else {
+ memcpy(dst, src, length);
+ return length;
+ }
+ }
+ if (conversion == CONVERT_BRACE) {
*p = '{';
p++;
- for (; src != lastChar; src++, p++) {
- *p = *src;
+ if (length == -1) {
+ /* TODO: INT_MAX overflow? */
+ while (*src) {
+ *p++ = *src++;
+ }
+ } else {
+ memcpy(p, src, length);
+ p += length;
}
*p = '}';
p++;
- } else {
- if (*src == '{') {
- /*
- * Can't have a leading brace unless the whole element is enclosed
- * in braces. Add a backslash before the brace. Furthermore, this
- * may destroy the balance between open and close braces, so set
- * BRACES_UNMATCHED.
- */
-
- p[0] = '\\';
- p[1] = '{';
- p += 2;
- src++;
- flags |= BRACES_UNMATCHED;
- } else if ((*src == '#') && !(flags & TCL_DONT_QUOTE_HASH)) {
- /*
- * Leading '#' could be seen by [eval] as the start of a comment,
- * if on the first element of a list, so quote it.
- */
-
- p[0] = '\\';
- p[1] = '#';
- p += 2;
- src++;
- }
- for (; src != lastChar; src++) {
- switch (*src) {
- case ']':
- case '[':
- case '$':
- case ';':
- case ' ':
- case '\\':
- case '"':
- *p = '\\';
- p++;
- break;
- case '{':
- case '}':
- /*
- * It may not seem necessary to backslash braces, but it is.
- * The reason for this is that the resulting list element may
- * actually be an element of a sub-list enclosed in braces
- * (e.g. if Tcl_DStringStartSublist has been invoked), so
- * there may be a brace mismatch if the braces aren't
- * backslashed.
- */
+ return p - dst;
+ }
+ /* conversion == CONVERT_ESCAPE or CONVERT_MASK */
- if (flags & BRACES_UNMATCHED) {
- *p = '\\';
- p++;
- }
- break;
- case '\f':
- *p = '\\';
- p++;
- *p = 'f';
- p++;
- continue;
- case '\n':
- *p = '\\';
- p++;
- *p = 'n';
- p++;
- continue;
- case '\r':
- *p = '\\';
- p++;
- *p = 'r';
- p++;
- continue;
- case '\t':
- *p = '\\';
- p++;
- *p = 't';
- p++;
- continue;
- case '\v':
+ for ( ; length; src++, length -= (length > 0)) {
+ switch (*src) {
+ case ']':
+ case '[':
+ case '$':
+ case ';':
+ case ' ':
+ case '\\':
+ case '"':
+ *p = '\\';
+ p++;
+ break;
+ case '{':
+ case '}':
+#if COMPAT
+ if (conversion == CONVERT_ESCAPE) {
+#endif
*p = '\\';
p++;
- *p = 'v';
- p++;
- continue;
+#if COMPAT
}
- *p = *src;
+#endif
+ break;
+ case '\f':
+ *p = '\\';
+ p++;
+ *p = 'f';
+ p++;
+ continue;
+ case '\n':
+ *p = '\\';
+ p++;
+ *p = 'n';
+ p++;
+ continue;
+ case '\r':
+ *p = '\\';
+ p++;
+ *p = 'r';
+ p++;
+ continue;
+ case '\t':
+ *p = '\\';
+ p++;
+ *p = 't';
+ p++;
+ continue;
+ case '\v':
+ *p = '\\';
+ p++;
+ *p = 'v';
p++;
+ continue;
+ case '\0':
+ if (length == -1) {
+ return p - dst;
+ }
+ /*
+ * If we reach this point, there's an embedded NULL in the
+ * string range being processed, which should not happen when
+ * the encoding rules for Tcl strings are properly followed.
+ * If the day ever comes when we stop tolerating such things,
+ * this is where to put the Tcl_Panic().
+ */
+ break;
}
+ *p = *src;
+ p++;
}
- *p = '\0';
- return p-dst;
+ return p - dst;
}
/*
@@ -967,11 +1057,20 @@
CONST char * CONST *argv) /* Array of string values. */
{
# define LOCAL_SIZE 20
- int localFlags[LOCAL_SIZE], *flagPtr;
- int numChars;
- char *result;
- char *dst;
- int i;
+ int localFlags[LOCAL_SIZE], *flagPtr = NULL;
+ int i, bytesNeeded = 0;
+ char *result, *dst;
+ const int maxFlags = UINT_MAX / sizeof(int);
+
+ if (argc == 0) {
+ /*
+ * Handle empty list case first, so logic of the general case
+ * can be simpler.
+ */
+ result = ckalloc(1);
+ result[0] = '\0';
+ return result;
+ }
/*
* Pass 1: estimate space, gather flags.
@@ -979,32 +1078,48 @@
if (argc <= LOCAL_SIZE) {
flagPtr = localFlags;
+ } else if (argc > maxFlags) {
+ /*
+ * We cannot allocate a large enough flag array to format this
+ * list in one pass. We could imagine converting this routine
+ * to a multi-pass implementation, but for sizeof(int) == 4,
+ * the limit is a max of 2^30 list elements and since each element
+ * is at least one byte formatted, and requires one byte space
+ * between it and the next one, that a minimum space requirement
+ * of 2^31 bytes, which is already INT_MAX. If we tried to format
+ * a list of > maxFlags elements, we're just going to overflow
+ * the size limits on the formatted string anyway, so just issue
+ * that same panic early.
+ */
+ Tcl_Panic("Tcl_Merge: size requirement exceeds limits");
} else {
flagPtr = (int *) ckalloc((unsigned) argc*sizeof(int));
}
- numChars = 1;
for (i = 0; i < argc; i++) {
- numChars += Tcl_ScanElement(argv[i], &flagPtr[i]) + 1;
+ flagPtr[i] = ( i ? TCL_DONT_QUOTE_HASH : 0 );
+ bytesNeeded += TclScanElement(argv[i], -1, &flagPtr[i]);
+ if (bytesNeeded < 0) {
+ Tcl_Panic("Tcl_Merge: size requirement exceeds limits");
+ }
}
+ if (bytesNeeded > INT_MAX - argc + 1) {
+ Tcl_Panic("Tcl_Merge: size requirement exceeds limits");
+ }
+ bytesNeeded += argc;
/*
* Pass two: copy into the result area.
*/
- result = (char *) ckalloc((unsigned) numChars);
+ result = ckalloc((unsigned) bytesNeeded);
dst = result;
for (i = 0; i < argc; i++) {
- numChars = Tcl_ConvertElement(argv[i], dst,
- flagPtr[i] | (i==0 ? 0 : TCL_DONT_QUOTE_HASH));
- dst += numChars;
+ flagPtr[i] |= ( i ? TCL_DONT_QUOTE_HASH : 0 );
+ dst += TclConvertElement(argv[i], -1, dst, flagPtr[i]);
*dst = ' ';
dst++;
}
- if (dst == result) {
- *dst = 0;
- } else {
- dst[-1] = 0;
- }
+ dst[-1] = 0;
if (flagPtr != localFlags) {
ckfree((char *) flagPtr);
@@ -1895,12 +2010,11 @@
CONST char *element) /* String to append. Must be
* null-terminated. */
{
- int newSize, flags, strSize;
- char *dst;
-
- strSize = ((element== NULL) ? 0 : strlen(element));
- newSize = Tcl_ScanCountedElement(element, strSize, &flags)
- + dsPtr->length + 1;
+ char *dst = dsPtr->string + dsPtr->length;
+ int needSpace = TclNeedSpace(dsPtr->string, dst);
+ int flags = needSpace ? TCL_DONT_QUOTE_HASH : 0;
+ int newSize = dsPtr->length + needSpace
+ + TclScanElement(element, -1, &flags);
/*
* Allocate a larger buffer for the string if the current one isn't large
@@ -1921,6 +2035,7 @@
dsPtr->string = (char *) ckrealloc((void *) dsPtr->string,
(size_t) dsPtr->spaceAvl);
}
+ dst = dsPtr->string + dsPtr->length;
}
/*
@@ -1928,8 +2043,7 @@
* the end, with a space, if needed.
*/
- dst = dsPtr->string + dsPtr->length;
- if (TclNeedSpace(dsPtr->string, dst)) {
+ if (needSpace) {
*dst = ' ';
dst++;
dsPtr->length++;
@@ -1942,7 +2056,8 @@
flags |= TCL_DONT_QUOTE_HASH;
}
- dsPtr->length += Tcl_ConvertCountedElement(element, strSize, dst, flags);
+ dsPtr->length += TclConvertElement(element, -1, dst, flags);
+ dsPtr->string[dsPtr->length] = '\0';
return dsPtr->string;
}
