Tcl Source Code

2011-03-17  Donal K. Fellows  <[email protected]>

	* generic/tclCompExpr.c (ParseExpr): Generate errorCode information on
	failure to parse expressions.

2011-03-17  Jan Nijtmans  <[email protected]>

	* generic/tkMain.c: [Patch 3124683]: Reorganize the platform-specific
	stuff in (tcl|tk)Main.c.

2011-03-16  Jan Nijtmans  <[email protected]>

	* generic/tclCkalloc.c: [Bug 3197864]: Pointer truncation on Win64
	TCL_MEM_DEBUG builds.

2011-03-16  Don Porter  <[email protected]>

	* generic/tclBasic.c:	Some rewrites to eliminate calls to
	* generic/tclParse.c:	isspace() and their /* INTL */ risk.
	* generic/tclProc.c:

    /* These variables control generation of the error message. */
    Tcl_Obj *msg = NULL;	/* The error message. */
    Tcl_Obj *post = NULL;	/* In a few cases, an additional postscript
				 * for the error message, supplying more
				 * information after the error msg and
				 * location have been reported. */
    const char *errCode = NULL;	/* The detail word of the errorCode list, or
				 * NULL to indicate that no changes to the
				 * errorCode are to be done. */
    const char *subErrCode = NULL;
				/* Extra information for use in generating the
				 * errorCode. */
    const char *mark = "_@_";	/* In the portion of the complete error
				 * message where the error location is
				 * reported, this "mark" substring is inserted
				 * into the string being parsed to aid in
				 * pinpointing the location of the syntax
				 * error in the expression. */
    int insertMark = 0;		/* A boolean controlling whether the "mark"
				 * should be inserted. */
    const int limit = 25;	/* Portions of the error message are
				 * constructed out of substrings of the
				 * original expression. In order to keep the
				 * error message readable, we impose this
				 * limit on the substring size we extract. */

    TclParseInit(interp, start, numBytes, parsePtr);

    nodes = attemptckalloc(nodesAvailable * sizeof(OpNode));
    if (nodes == NULL) {
	TclNewLiteralStringObj(msg, "not enough memory to parse expression");
	errCode = "NOMEM";
	goto error;
    }

    /*
     * Initialize the parse tree with the special "START" node.
     */

	    do {
		newPtr = attemptckrealloc(nodes, size * sizeof(OpNode));
	    } while ((newPtr == NULL)
		    && ((size -= (size - nodesUsed) / 2) > nodesUsed));
	    if (newPtr == NULL) {
		TclNewLiteralStringObj(msg,
			"not enough memory to parse expression");
		errCode = "NOMEM";
		goto error;
	    }
	    nodesAvailable = size;
	    nodes = newPtr;
	}
	nodePtr = nodes + nodesUsed;

	/*
	 * Skip white space between lexemes.
	 */

	scanned = TclParseAllWhiteSpace(start, numBytes);
	start += scanned;
	numBytes -= scanned;

	scanned = ParseLexeme(start, numBytes, &lexeme, &literal);

	/*
	 * Use context to categorize the lexemes that are ambiguous.
	 */

	if ((NODE_TYPE & lexeme) == 0) {
	    int b;

	    switch (lexeme) {
	    case INVALID:
		msg = Tcl_ObjPrintf("invalid character \"%.*s\"",
			scanned, start);
		errCode = "BADCHAR";
		goto error;
	    case INCOMPLETE:
		msg = Tcl_ObjPrintf("incomplete operator \"%.*s\"",
			scanned, start);
		errCode = "PARTOP";
		goto error;
	    case BAREWORD:

		/*
		 * Most barewords in an expression are a syntax error. The
		 * exceptions are that when a bareword is followed by an open
		 * paren, it might be a function call, and when the bareword

		     * When we compile the expression we'll need the function
		     * name, and there's no place in the parse tree to store
		     * it, so we keep a separate list of all the function
		     * names we've parsed in the order we found them.
		     */

		    Tcl_ListObjAppendElement(NULL, funcList, literal);


		} else if (Tcl_GetBooleanFromObj(NULL,literal,&b) == TCL_OK) {
		    lexeme = BOOLEAN;
		} else {
		    Tcl_DecrRefCount(literal);
		    msg = Tcl_ObjPrintf("invalid bareword \"%.*s%s\"",

			    (scanned < limit) ? scanned : limit - 3, start,
			    (scanned < limit) ? "" : "...");
		    post = Tcl_ObjPrintf(
			    "should be \"$%.*s%s\" or \"{%.*s%s}\"",
			    (scanned < limit) ? scanned : limit - 3,
			    start, (scanned < limit) ? "" : "...",
			    (scanned < limit) ? scanned : limit - 3,
			    start, (scanned < limit) ? "" : "...");
		    Tcl_AppendPrintfToObj(post, " or \"%.*s%s(...)\" or ...",

			    (scanned < limit) ? scanned : limit - 3,
			    start, (scanned < limit) ? "" : "...");
		    if (NotOperator(lastParsed)) {
			errCode = "BADNUMBER";
			if ((lastStart[0] == '0')
				&& ((lastStart[1] == 'o')
				|| (lastStart[1] == 'O'))
				&& (lastStart[2] >= '0')
				&& (lastStart[2] <= '9')) {
			    const char *end = lastStart + 2;
			    Tcl_Obj *copy;

			    while (isdigit(UCHAR(*end))) {
				end++;
			    }
			    copy = Tcl_NewStringObj(lastStart, end-lastStart);

			    if (TclCheckBadOctal(NULL, Tcl_GetString(copy))) {

				Tcl_AppendToObj(post,
					" (invalid octal number?)", -1);
				errCode = "BADNUMBER";
				subErrCode = "OCTAL";
			    }
			    Tcl_DecrRefCount(copy);
			}
			scanned = 0;
			insertMark = 1;
			parsePtr->errorType = TCL_PARSE_BAD_NUMBER;
		    } else {
			errCode = "BAREWORD";
		    }
		    goto error;

		}
		break;
	    case PLUS:
	    case MINUS:
		if (IsOperator(lastParsed)) {
		    /*
		     * A "+" or "-" coming just after another operator must be

	    /*
	     * A leaf operand appearing just after something that's not an
	     * operator is a syntax error.
	     */

	    if (NotOperator(lastParsed)) {
		msg = Tcl_ObjPrintf("missing operator at %s", mark);
		errCode = "MISSING";
		if (lastStart[0] == '0') {
		    Tcl_Obj *copy = Tcl_NewStringObj(lastStart,
			    start + scanned - lastStart);

		    if (TclCheckBadOctal(NULL, Tcl_GetString(copy))) {
			TclNewLiteralStringObj(post,
				"looks like invalid octal number");
			errCode = "BADNUMBER_OCTAL";
		    }
		    Tcl_DecrRefCount(copy);
		}
		scanned = 0;
		insertMark = 1;
		parsePtr->errorType = TCL_PARSE_BAD_NUMBER;

		code = Tcl_ParseQuotedString(NULL, start, numBytes,
			parsePtr, 1, &end);
		scanned = end - start;
		break;

	    case BRACED:
		code = Tcl_ParseBraces(NULL, start, numBytes,
			parsePtr, 1, &end);
		scanned = end - start;
		break;

	    case VARIABLE:
		code = Tcl_ParseVarName(NULL, start, numBytes, parsePtr, 1);

		/*
		 * Handle the quirk that Tcl_ParseVarName reports a successful
		 * parse even when it gets only a "$" with no variable name.
		 */

		tokenPtr = parsePtr->tokenPtr + wordIndex + 1;
		if (code == TCL_OK && tokenPtr->type != TCL_TOKEN_VARIABLE) {
		    TclNewLiteralStringObj(msg, "invalid character \"$\"");
		    errCode = "BADCHAR";
		    goto error;
		}
		scanned = tokenPtr->size;
		break;

	    case SCRIPT: {
		Tcl_Parse *nestedPtr =
			TclStackAlloc(interp, sizeof(Tcl_Parse));

		tokenPtr = parsePtr->tokenPtr + parsePtr->numTokens;
		tokenPtr->type = TCL_TOKEN_COMMAND;
		tokenPtr->start = start;
		tokenPtr->numComponents = 0;

		end = start + numBytes;
		start++;
		while (1) {
		    code = Tcl_ParseCommand(interp, start, end - start, 1,
			    nestedPtr);
		    if (code != TCL_OK) {
			parsePtr->term = nestedPtr->term;
			parsePtr->errorType = nestedPtr->errorType;
			parsePtr->incomplete = nestedPtr->incomplete;
			break;
		    }
		    start = nestedPtr->commandStart + nestedPtr->commandSize;
		    Tcl_FreeParse(nestedPtr);
		    if ((nestedPtr->term < end) && (nestedPtr->term[0] == ']')
			    && !nestedPtr->incomplete) {
			break;
		    }

		    if (start == end) {
			TclNewLiteralStringObj(msg, "missing close-bracket");
			parsePtr->term = tokenPtr->start;
			parsePtr->errorType = TCL_PARSE_MISSING_BRACKET;
			parsePtr->incomplete = 1;
			code = TCL_ERROR;
			errCode = "UNBALANCED";
			break;
		    }
		}
		TclStackFree(interp, nestedPtr);
		end = start;
		start = tokenPtr->start;
		scanned = end - start;
		tokenPtr->size = scanned;
		parsePtr->numTokens++;
		break;
	    }			/* SCRIPT case */
	    }
	    if (code != TCL_OK) {
		/*
		 * Here we handle all the syntax errors generated by the
		 * Tcl_Token generating parsing routines called in the switch
		 * just above. If the value of parsePtr->incomplete is 1, then
		 * the error was an unbalanced '[', '(', '{', or '"' and
		 * parsePtr->term is pointing to that unbalanced character. If
		 * the value of parsePtr->incomplete is 0, then the error is
		 * one of lacking whitespace following a quoted word, for
		 * example: expr {[an error {foo}bar]}, and parsePtr->term
		 * points to where the whitespace is missing. We reset our
		 * values of start and scanned so that when our error message
		 * is constructed, the location of the syntax error is sure to
		 * appear in it, even if the quoted expression is truncated.
		 */

		start = parsePtr->term;
		scanned = parsePtr->incomplete;
		if (parsePtr->incomplete) {
		    errCode = "UNBALANCED";
		}
		goto error;
	    }

	    tokenPtr = parsePtr->tokenPtr + wordIndex;
	    tokenPtr->size = scanned;
	    tokenPtr->numComponents = parsePtr->numTokens - wordIndex - 1;
	    if (!parseOnly && ((lexeme == QUOTED) || (lexeme == BRACED))) {

	     * operand of an operator that doesn't take one.
	     */

	    if (NotOperator(lastParsed)) {
		msg = Tcl_ObjPrintf("missing operator at %s", mark);
		scanned = 0;
		insertMark = 1;
		errCode = "MISSING";
		goto error;
	    }

	    /* Create an OpNode for the unary operator */
	    nodePtr->lexeme = lexeme;
	    nodePtr->precedence = prec[lexeme];
	    nodePtr->mark = MARK_RIGHT;

			scanned = 0;
			complete = lastParsed = OT_EMPTY;
			break;
		    }
		    msg = Tcl_ObjPrintf("empty subexpression at %s", mark);
		    scanned = 0;
		    insertMark = 1;
		    errCode = "EMPTY";
		    goto error;
		}

		if (nodePtr[-1].precedence > precedence) {
		    if (nodePtr[-1].lexeme == OPEN_PAREN) {
			TclNewLiteralStringObj(msg, "unbalanced open paren");
			parsePtr->errorType = TCL_PARSE_MISSING_PAREN;
			errCode = "UNBALANCED";
		    } else if (nodePtr[-1].lexeme == COMMA) {
			msg = Tcl_ObjPrintf(
				"missing function argument at %s", mark);
			scanned = 0;
			insertMark = 1;
			errCode = "MISSING";
		    } else if (nodePtr[-1].lexeme == START) {
			TclNewLiteralStringObj(msg, "empty expression");
			errCode = "EMPTY";
		    }

		} else if (lexeme == CLOSE_PAREN) {
		    TclNewLiteralStringObj(msg, "unbalanced close paren");
		    errCode = "UNBALANCED";
		} else if ((lexeme == COMMA)
			&& (nodePtr[-1].lexeme == OPEN_PAREN)
			&& (nodePtr[-2].lexeme == FUNCTION)) {
		    msg = Tcl_ObjPrintf("missing function argument at %s",
			    mark);
		    scanned = 0;
		    insertMark = 1;

		    errCode = "UNBALANCED";
		}
		if (msg == NULL) {
		    msg = Tcl_ObjPrintf("missing operand at %s", mark);
		    scanned = 0;
		    insertMark = 1;
		    errCode = "MISSING";
		}
		goto error;
	    }

	    /*
	     * Here is where the tree comes together. At this point, we have a
	     * stack of incomplete trees corresponding to substrings that are

		 */

		/* Parens must balance */
		if ((incompletePtr->lexeme == OPEN_PAREN)
			&& (lexeme != CLOSE_PAREN)) {
		    TclNewLiteralStringObj(msg, "unbalanced open paren");
		    parsePtr->errorType = TCL_PARSE_MISSING_PAREN;
		    errCode = "UNBALANCED";
		    goto error;
		}

		/* Right operand of "?" must be ":" */
		if ((incompletePtr->lexeme == QUESTION)
			&& (NotOperator(complete)
			|| (nodes[complete].lexeme != COLON))) {

		    msg = Tcl_ObjPrintf("missing operator \":\" at %s", mark);
		    scanned = 0;
		    insertMark = 1;
		    errCode = "MISSING";
		    goto error;
		}

		/* Operator ":" may only be right operand of "?" */
		if (IsOperator(complete)
			&& (nodes[complete].lexeme == COLON)
			&& (incompletePtr->lexeme != QUESTION)) {
		    TclNewLiteralStringObj(msg,
			    "unexpected operator \":\" "
			    "without preceding \"?\"");
		    errCode = "SURPRISE";
		    goto error;
		}

		/*
		 * Attach complete tree as right operand of most recent
		 * incomplete tree.
		 */

	     * More syntax checks...
	     */

	    /* Parens must balance. */
	    if (lexeme == CLOSE_PAREN) {
		if (incompletePtr->lexeme != OPEN_PAREN) {
		    TclNewLiteralStringObj(msg, "unbalanced close paren");
		    errCode = "UNBALANCED";
		    goto error;
		}
	    }

	    /* Commas must appear only in function argument lists. */
	    if (lexeme == COMMA) {
		if  ((incompletePtr->lexeme != OPEN_PAREN)
			|| (incompletePtr[-1].lexeme != FUNCTION)) {
		    TclNewLiteralStringObj(msg,
			    "unexpected \",\" outside function argument list");
		    errCode = "SURPRISE";
		    goto error;
		}
	    }

	    /* Operator ":" may only be right operand of "?" */
	    if (IsOperator(complete) && (nodes[complete].lexeme == COLON)) {
		TclNewLiteralStringObj(msg,
			"unexpected operator \":\" without preceding \"?\"");
		errCode = "SURPRISE";
		goto error;
	    }

	    /*
	     * Create no node for a CLOSE_PAREN lexeme.
	     */

	 */

	numBytes = parsePtr->end - parsePtr->string;
	Tcl_AppendObjToErrorInfo(interp, Tcl_ObjPrintf(
		"\n    (parsing expression \"%.*s%s\")",
		(numBytes < limit) ? numBytes : limit - 3,
		parsePtr->string, (numBytes < limit) ? "" : "..."));
	if (errCode) {
	    Tcl_SetErrorCode(interp, "TCL", "PARSE", "EXPR", errCode,
		    subErrCode, NULL);
	}
    }

    return TCL_ERROR;
}

/*
 *----------------------------------------------------------------------