From f9fb8c8daf918e57d3a50a0428623a5e5c260e70 Mon Sep 17 00:00:00 2001
From: dgp <dgp@users.sourceforge.net>
Date: Thu, 4 Dec 2014 18:26:31 +0000
Subject: Stop using isalnum(.).  Its results are not portable.  Replace with
 our own private routine TclIsBareword() that does exactly what we want.

---
 generic/tclCompExpr.c | 15 ++++++++++-----
 generic/tclInt.h      |  1 +
 generic/tclParse.c    | 47 +++++++++++++++++++++++++++++++++++++++++++----
 3 files changed, 54 insertions(+), 9 deletions(-)

diff --git a/generic/tclCompExpr.c b/generic/tclCompExpr.c
index dde4e56..2470931 100644
--- a/generic/tclCompExpr.c
+++ b/generic/tclCompExpr.c
@@ -1920,8 +1920,7 @@ ParseLexeme(
     literal = Tcl_NewObj();
     if (TclParseNumber(NULL, literal, NULL, start, numBytes, &end,
 	    TCL_PARSE_NO_WHITESPACE) == TCL_OK) {
-	if (end < start + numBytes && !isalnum(UCHAR(*end))
-		&& UCHAR(*end) != '_') {
+	if (end < start + numBytes && !TclIsBareword(*end)) {
 	
 	number:
 	    TclInitStringRep(literal, start, end-start);
@@ -1945,7 +1944,7 @@ ParseLexeme(
 	    if (literal->typePtr == &tclDoubleType) {
 		const char *p = start;
 		while (p < end) {
-		    if (!isalnum(UCHAR(*p++))) {
+		    if (!TclIsBareword(*p++)) {
 			/*
 			 * The number has non-bareword characters, so we 
 			 * must treat it as a number.
@@ -1969,7 +1968,13 @@ ParseLexeme(
 	}
     }
 
-    if (!isalnum(UCHAR(*start))) {
+    /*
+     * We reject leading underscores in bareword.  No sensible reason why.
+     * Might be inspired by reserved identifier rules in C, which of course
+     * have no direct relevance here.
+     */  
+
+    if (!TclIsBareword(*start) || *start == '_') {
 	if (Tcl_UtfCharComplete(start, numBytes)) {
 	    scanned = Tcl_UtfToUniChar(start, &ch);
 	} else {
@@ -1983,7 +1988,7 @@ ParseLexeme(
 	return scanned;
     }
     end = start;
-    while (numBytes && (isalnum(UCHAR(*end)) || (UCHAR(*end) == '_'))) {
+    while (numBytes && TclIsBareword(*end)) {
 	end += 1;
 	numBytes -= 1;
     }
diff --git a/generic/tclInt.h b/generic/tclInt.h
index dd66d76..255ee23 100644
--- a/generic/tclInt.h
+++ b/generic/tclInt.h
@@ -2608,6 +2608,7 @@ MODULE_SCOPE void	TclInitSubsystems(void);
 MODULE_SCOPE int	TclInterpReady(Tcl_Interp *interp);
 MODULE_SCOPE int	TclIsLocalScalar(const char *src, int len);
 MODULE_SCOPE int	TclIsSpaceProc(char byte);
+MODULE_SCOPE int	TclIsBareword(char byte);
 MODULE_SCOPE int	TclJoinThread(Tcl_ThreadId id, int *result);
 MODULE_SCOPE void	TclLimitRemoveAllHandlers(Tcl_Interp *interp);
 MODULE_SCOPE Tcl_Obj *	TclLindexList(Tcl_Interp *interp,
diff --git a/generic/tclParse.c b/generic/tclParse.c
index 90ec43d..025304c 100644
--- a/generic/tclParse.c
+++ b/generic/tclParse.c
@@ -628,6 +628,47 @@ TclIsSpaceProc(
 /*
  *----------------------------------------------------------------------
  *
+ * TclIsBareword--
+ *
+ *	Report whether byte is one that can be part of a "bareword".
+ *	This concept is named in expression parsing, where it determines
+ *	what can be a legal function name, but is the same definition used
+ *	in determining what variable names can be parsed as variable
+ *	substitutions without the benefit of enclosing braces.  The set of
+ *	ASCII chars that are accepted are the numeric chars ('0'-'9'),
+ *	the alphabetic chars ('a'-'z', 'A'-'Z')	and underscore ('_').
+ *
+ * Results:
+ *	Returns 1, if byte is in the accepted set of chars, 0 otherwise.
+ *
+ * Side effects:
+ *	None.
+ *
+ *----------------------------------------------------------------------
+ */
+
+int
+TclIsBareword(
+    char byte)
+{
+    if (byte < '0' || byte > 'z') {
+	return 0;
+    }
+    if (byte <= '9' || byte >= 'a') {
+	return 1;
+    }
+    if (byte == '_') {
+	return 1;
+    }
+    if (byte < 'A' || byte > 'Z') {
+	return 0;
+    }
+    return 1;
+}
+
+/*
+ *----------------------------------------------------------------------
+ *
  * ParseWhiteSpace --
  *
  *	Scans up to numBytes bytes starting at src, consuming white space
@@ -1343,7 +1384,6 @@ Tcl_ParseVarName(
 {
     Tcl_Token *tokenPtr;
     register const char *src;
-    unsigned char c;
     int varIndex;
     unsigned array;
 
@@ -1427,13 +1467,12 @@ Tcl_ParseVarName(
 	tokenPtr->numComponents = 0;
 
 	while (numBytes) {
-	    c = UCHAR(*src);
-	    if (isalnum(c) || (c == '_')) {	/* INTL: ISO only, UCHAR. */
+	    if (TclIsBareword(*src)) {
 		src += 1;
 		numBytes -= 1;
 		continue;
 	    }
-	    if ((c == ':') && (numBytes != 1) && (src[1] == ':')) {
+	    if ((src[0] == ':') && (numBytes != 1) && (src[1] == ':')) {
 		src += 2;
 		numBytes -= 2;
 		while (numBytes && (*src == ':')) {
-- 
cgit v0.12