/***  LEXER.C -- gets tokens from input, returns them to parse() in parser.c ***
*
*	Copyright (c) 1988-1990, Microsoft Corporation.  All rights reserved.
*
* Purpose:
*  This module contains the lexical routines of nmake
*
* Revision History:
*  15-Oct-1993 HV Use tchar.h instead of mbstring.h directly, change STR*() to _ftcs*()
*  01-Jun-1993 HV Use UngetTxtChr() instead of ungetc()
*  01-Jun-1993 HV Change #ifdef KANJI to _MBCS
*  10-May-1993 HV Add include file mbstring.h
*                 Change the str* functions to STR*
*  06-Apr-1993 HV Change createDosTmp() to use _makepath()
*  22-Mar-1993 HV Rewrite getPath() to make use of the new _splitpath() and
*		  _makepath() functions.
*  04-Aug-1992 SS CAVIAR 2266: expand INCLUDE macro in processIncludeFile()
*  08-Jun-1992 SS Port to DOSX32
*  02-Feb-1990 SB change fopen() to FILEOPEN()
*  01-Dec-1989 SB Changed realloc() to REALLOC()
*  22-Nov-1989 SB Changed free() to FREE()
*  19-Oct-1989 SB searchHandle passed around as extra param
*  08-Oct-1989 SB handle OS/2 1.2 quoted filenames
*  04-Sep-1989 SB temporary filename generated has a trailing '.' for LINK.EXE
*  24-Aug-1989 SB Allow $* and $@ in dependency lines
*  18-Aug-1989 SB Added fclose() return code check
*  31-Jul-1989 SB Added lookahead to the lexer for \ on dependency lines
*  06-Jul-1989 SB Remove escaping abilities of '^' in command lines totally
*  29-Jun-1989 SB Add duplicateInline() to detect duplicate inline filenames
*		  and issue error if duplicates are found
*  26-Jun-1989 SB Modify ParseScriptFileList() and add nextInlineFile() to
*		  handle complex syntax of Inline file command line.
*  15-Jun-1989 SB issue error for usage of inline file in an inference rule
*  18-May-1989 SB Added getPath(), changed processIncludeFile() to have C like
*		  processing of include files
*  16-May-1989 SB expand macros in include file names; handle '\' processing
*		  in same way for macros and dependency lines
*  15-May-1989 SB Changed nameStates to 16x14
*  13-May-1989 SB don't remove ESCH on reading cmd block
*  24-Apr-1989 SB made FILEINFO as void * and corrected regression in parsing
*		  inline file names
*  14-Apr-1989 SB inline file names are correctly expanded now
*  06-Apr-1989 SB ren removeFirstLtLt() as delInlineSymbol().
*  05-Apr-1989 SB made all funcs NEAR; Reqd to make all function calls NEAR
*  22-Mar-1989 SB removed unlinkTmpFiles() function; not needed
*  19-Jan-1989 SB added function removeFirstLtLt() to remove '<<' appearing
*		  in -n output
*  30-Dec-1988 SB Fixed GP fault for KEEP/NOKEEP in parseScriptFileList()
*		  and makeScriptFileList()
*  21-Dec-1988 SB Added parseScriptFileList() and appendScript() to allow
*		  handling of multiple script files inside a makefile
*		  Improved KEEP/NOKEEP so that each file can have its own
*		  action
*  16-Dec-1988 SB Added to makeScriptFile() for KEEP/NOKEEP
*  14-Dec-1988 SB Added tmpScriptFile so that a delete command can be
*		  added for unnamed script files for Z option
*  13-Dec-1988 SB Added processEschIn() to improve response files
*   5-Oct-1988 RB Strip trailing whitespace from macro defs, build lines.
*  22-Sep-1988 RB Fix skipComments() to not parse \\nl.
*  20-Sep-1988 RB Error if named script file creation fails.
*		  Count line numbers in script files.
*  18-Sep-1988 RB Handle mktemp() small limit.
*  17-Aug-1988 RB Clean up.
*  14-Jul-1988 rj Fixed handling of ^ before !, @, or -.
*   8-Jul-1988 rj Added handler to ignore ^ inside quotes.
*		  Made \ carry comments over lines.
*		  Made ^ carry comments over lines.
*  27-Jun-1988 rj Fixed bug with handling of response files.
*  16-Jun-1988 rj Finished up ESCH.
*  15-Jun-1988 rj Added support for ESCH escape:  modified skipWhiteSpace
*		  (adding some redundancy in setting colZero), getString,
*		  getName; removed \\nl escape.
*  13-Jun-1988 rj Fixed backslashes to work as in nmake, with addition of
*		  double-backslash escape.  (v1.5)
*
*******************************************************************************/

#include "nmake.h"
#include "nmmsg.h"
#include "proto.h"
#include "globals.h"
#include "grammar.h"

#define COMMENT(A,B,C)	    (((A) == ';' && B && C) || ((A) == '#'))
#ifdef _MBCS
#define GET(A)		    A ? GetTxtChr(file) : lgetc()
#else
#define GET(A)		    A ? getc(file) : lgetc()
#endif

extern char	  * NEAR makeInlineFiles(char*, char**, char**);
extern void	    NEAR removeTrailChars(char *);

LOCAL void	    NEAR skipComments(UCHAR);
LOCAL void	    NEAR getString(UCHAR,char*,char*);
LOCAL void	    NEAR getName(char*,char*);
LOCAL UCHAR	    NEAR determineTokenFor(int,char*,char*);
LOCAL void	    NEAR popFileStack(void);
LOCAL UCHAR	    NEAR include(int);
LOCAL char	  * NEAR getPath(const char *);

extern UCHAR	    NEAR nameStates[18][15];
extern UCHAR	    NEAR stringStates[13][14];
extern STRINGLIST * NEAR targetList;


/*  --------------------------------------------------------------------
 *  getToken()
 *
 *  arguments:	    init	global boolean value -- TRUE if tools.ini is the
 *				    file being lexed
 *		    n		size of s[]
 *		    expected	kind of token expected by parser -- only
 *				needed when parser wants a whole string
 *				(meaning everything left on the current line)
 *				-- this way getToken() doesn't break strings
 *				into their separate tokens
 *
 *  actions:	    if no tokens have been read from current file,
 *			returns some kind of newline to initialize the parser
 *			(if 1st char in file is whitespace, returns NEWLINESPACE
 *			else returns NEWLINE -- w/o actually getting a token
 *			from the input)
 *		    if the parser wants a whole string, reads rest of line
 *			into s and returns STRING
 *		    if at end of file, return ACCEPT (which is the last
 *			symbol on the parser's stack)
 *		    if input char is newline
 *			if followed by whitespace, return NEWLINESPACE
 *			if the next char is [ and we're reading tools.ini
 *			    pretend that we've reached end of file and
 *			    return ACCEPT
 *			otherwise return NEWLINE
 *		    if input char is colon
 *			if following char is also colon,
 *			    (put both chars in s) return DOUBLECOLON
 *			otherwise return SINGLECOLON
 *		    if input char is semicolon return SEMICOLON
 *		    if input char is equals return EQUALS
 *		    if input char is exclamation handle directives
 *			(not yet implemented)
 *		    otherwise char must be part of a name, so gather
 *			the rest of the identifier and return NAME
 *
 *  returns:	    token type: NEWLINE NEWLINESPACE NAME EQUALS COLON
 *			SEMICOLON STRING ACCEPT
 *
 *  modifies:	    buf     by modifying *s, which points somewhere into buf
 *		    line    global line count
 *		    fname   will change when !include is handled
 *		    colZero global flag set if at column zero of a file
 *
 *  The lexer has to keep track of whether or not it is at the beginning
 *  of a line in the makefile (i.e. in column zero) so that it will know
 *  whether to ignore comments.  If init is TRUE, meaning that we are
 *  lexing tools.ini, then we have to treat lines beginning with ';' as
 *  comment lines.  If the parser expects a string, only comments beginning
 *  in column zero are ignored; all others are returned as part of the
 *  string.  Comments are stripped from macro values (strings that are
 *  part of macro definitions).
 *
 *  The user can specify a macro definition or a build line that
 *  spans several lines (using the \<newline> to "continue" the lines) while
 *  interspersing comment lines with the text.
 */

UCHAR NEAR
getToken(n,expected)
unsigned n;					    /* size of s[]	      */
UCHAR expected; 				    /* STRING means get line  */
{						    /*	w/o checking for #;:= */
    char *s;
    char *end;
    int c;

    s = buf;
    end = buf + n;
    if (firstToken) {					    /* global var     */
	++line;
	firstToken = FALSE;				    /* parser needs to*/
							    /*	see some kind */
	c = lgetc();					    /*	of newline to */
	if (colZero = (BOOL) !WHITESPACE(c)) {		    /*	initialize it */
	    if (c == EOF)
		return(determineTokenFor(c,s,end));
	    else
		UngetTxtChr(c,file);
	    return(NEWLINE);
	}
	return(NEWLINESPACE);
    }
    if (expected == STRING || expected == VALUE) {	    /* get everything */
	getString(expected,s,end);			    /*	up to \n      */
	return(expected);
    }							    /* were/are we    */
    c = skipWhiteSpace(FROMLOCAL);			    /*	past col 0?   */
    *s++ = (char) c;					    /* save the letter*/
    *s = '\0';						    /* terminate s    */
    return(determineTokenFor(c,s,end));
}

/*  -----------------------------------------------------------------
 *  determineTokenFor()
 *
 *  arguments:	c        current input character
 *		s        buffer to place token in for return to parser
 *		end      end of the token return buffer
 *
 *  returns:	token type: NEWLINE NEWLINESPACE NAME EQUALS COLON
 *			    SEMICOLON ACCEPT
 *
 *  modifies:	    buf     by modifying *s, which points somewhere into buf
 *		    line    global line count
 *		    fname   will change when include is handled
 *		    init    global flag - set if parsing tools.ini
 *		    colZero global flag set if at column zero of a file
 *
 */


LOCAL UCHAR NEAR
determineTokenFor(c,s,end)
int c;
char *s;
char *end;
{
    switch (c) {
	case EOF:   if (!feof(file))
			makeError(line,LEXER+FATAL_ERR);
		    if (incTop) popFileStack();
		    else if (ifTop >= 0)  /* all directives not processed*/
			     makeError(line,SYNTAX_EOF_NO_DIRECTIVE);
	            else return(ACCEPT);
	case '\n':  ++line;
		    colZero = TRUE;
		    c = lgetc();
		    if (COMMENT(c,TRUE,init)) {
			skipComments(FROMLOCAL);
			++line;
		        colZero = TRUE;     /* manis - 11/13/87 */
			c = lgetc();
		    }
		    if (colZero = (BOOL) !WHITESPACE(c)) {
			if (c == EOF)
			    return(determineTokenFor(c,s,end));
			else
			    //save for next token
			    UngetTxtChr(c,file);
			return(NEWLINE);
		    }
		    return(NEWLINESPACE);
	case ':':   colZero = FALSE;
		    if ((c = lgetc()) == ':') {
			*s++ = (char) c;
			*s = '\0';
			return(DOUBLECOLON);
		    }
		    UngetTxtChr(c,file);
		    return(COLON);
	case ';':   colZero = FALSE;
		    return(SEMICOLON);
	case '=':   colZero = FALSE;
		    return(EQUALS);
	case '[':   if (init && colZero)
			return(ACCEPT);
	case ESCH:  UngetTxtChr(c, file);		   /* getName has to get esch */
		    s--;		      /* so we don't double the caret */
	default:    getName(s,end);
		    if (colZero && !_ftcsicmp(buf, "include")) {
			colZero = FALSE;
			if ((c = skipWhiteSpace(FROMLOCAL)) != ':'
			    && c != '=')  {
			    if (init)
				makeError(line, SYNTAX_UNEXPECTED_TOKEN, s);
			    return(include(c));
			}
			UngetTxtChr(c,file);
		    }
		    else colZero = FALSE;
		    return(NAME);
    }
}

/*  ----------------------------------------------------------------------------
 *  skipWhiteSpace()
 *
 *  arguments:	    c	    current input character
 *		    init    global boolean value -- TRUE if we're lexing tools.ini
 *		    colZero global boolean value -- TRUE if the current
 *			    input char is at the beginning of the line
 *
 *  actions:	    reads and discards characters until it gets a
 *			non-whitespace char that isn't part of a comment
 *			or hits the end of the line (NEWLINE and NEWLINESPACE
 *			are valid tokens and shouldn't be skipped w/ whitespace)
 *		    backslash-newline ('\\''\n') is treated as whitespace
 *		    comments are treated as whitespace
 *		    escaped whitespace is treated as whitespace (v1.5)
 *
 *  modifies:	    colZero global boolean value to :
 *			    TRUE if by skipping whitespace and comments we're
 *				at the beginning of a line
 *			    else if we skipped characters and are not at the
 *				beginning of a line, FALSE
 *			    else if we did not skip any characters, leave
 *				colZero unchanged
 *
 *  returns:	    c	    the current non-whitespace input char
 */

int NEAR
skipWhiteSpace(stream)
UCHAR stream;
{
    int c;

    do {
	c = GET(stream);
	if (WHITESPACE(c) || c == ESCH) {
	    if (c == ESCH) {
		c = GET(stream);
		if (!WHITESPACE(c)) {	    /* push char back out, return esch*/
		    UngetTxtChr(c, file);
		    c = ESCH;
		    break;
		}
	    }
	    colZero = FALSE;			    /* we've moved past col 0 */
	}
	if (c == '\\')
	    c = skipBackSlash(c, stream);
    } while(WHITESPACE(c));

    if (COMMENT(c,colZero,init)) {
	skipComments(stream);			    /* current char is always */
	c = '\n';				    /*	\n after comments     */
	colZero = TRUE; 			    /* always in col 0 after  */
    }						    /*	a comment	      */
    return(c);					    /* true if we're in col 0 */
}

/*  ----------------------------------------------------------------------------
 *  skipComments()
 *
 *  arguments:	    c	    pointer to current input character
 *		    init    global boolean value -- TRUE if tools.ini is the
 *				file being lexed
 *
 *  actions:	    reads and discards characters until it hits the end of
 *			the line
 *		    checks to see if 1st char on next line is comment,
 *			and if so, discards that line, too
 *		    DO NOT parse backslash-newline.  That would break our
 *		    precedence of comments over escaped newlines, the reverse
 *		    of Xenix.
 *
 *  modifies:       line      global line count
 *		    colZero
 *
 */


LOCAL void NEAR
skipComments(stream)
UCHAR stream;
{
    int c;

    for (;;) {
	colZero = FALSE;	       /* manis 11/13/87 */
	do
	    c = GET(stream);
	while (c != EOF && c != '\n');
	if (c == EOF) return;
	colZero = TRUE;
	c = GET(stream);
	if (!COMMENT(c,TRUE,init)) {		    /* if next line comment,  */
	    UngetTxtChr(c,file);			    /*	go around again       */
	    return;
	}
	++line;
    }
}

/*  -------------------------------------------------------------------------
 *  skipBackSlash()  - skips backslash-newline sequences
 *
 *
 *  arguments:  c            current input char
 *		stream       flag to determine if chars are to be got
 *			     from the raw stream or thru' lgetc()
 *
 *
 */

int NEAR
skipBackSlash(c,stream)
int c;
UCHAR stream;
{
    while (c == '\\') { 			    /* treat \newline as space*/
	if ((c = GET(stream)) == '\n') { 	    /*	and consume it too    */
	    colZero = TRUE;                         /* manis - 11/13-87 */
	    ++line;				    /* adjust line count      */
	    c = GET(stream);			    /* skip over newline      */
	    if (COMMENT(c,TRUE,init)) { 	    /* skip comment line after*/
		skipComments(stream);	            /*	continuation char     */
		++line;                             /* manis - 11/13/87       */
		c = GET(stream);
	    }
	}
	else {
	    UngetTxtChr(c,file);
	    c = '\\';
	    return(c);
	}
    }
    return(c);
}


/*  ----------------------------------------------------------------------------
 *  getString()
 *
 *  arguments:	    type    says which kind of token we're getting,
 *				a build STRING, or macro VALUE
 *				(we strip comments from VALUEs, but not
 *				from STRINGs)
 *		    s	    pointer to buffer that will hold string
 *		    init    global boolean value -- TRUE if tools.ini is the
 *				file being lexed
 *		    colZero global boolean value -- true if we 're in
 *			    1st position of line when invoked
 *		    end     pointer to end of s[]
 *
 *  actions:	    gets all chars up to the end of line or end of file
 *			and stores them in s[]
 *		    backslash followed by newline is replaced by a single
 *			space, and getString() continues getting characters
 *		    comments beginning in column 0 are ignored, as are
 *			comments anywhere on a VALUE line
 *
 *  modifies:	    buf     by modifying *s
 *		    line    global line count
 *		    colZero thru' calls to lgetc()
 *
 *  When build strings or macro values are continued on the next line w/
 *  a backslash before the newline, leading whitespace after the newline
 *  is omitted.  This is for xmake compatibility.
 *
 *  The continuation character is backslash immediately before newline.
 *
 *  The only difference between build strings and macro values is that
 *  comments are stripped from macro values and not from build strings.
 *
 *  Modifications:
 *
 *	06-Jul-1989 SB remove escaping in command lines
 *	15-Jun-1988 rj Added escape functionality.  Escape char., before
 *		       certain characters, causes those characters to bypass
 *		       the normal mechanism determining their type; they are
 *		       placed directly into the string.  Some characters cause
 *		       the escape character itself to be placed into the
 *		       string.
 */

LOCAL void NEAR
getString(type,s,end)
UCHAR type;						    /* build string or*/
char *s;						    /*	macro value?  */
char *end;
{
    int c;					   /*	   buffer    */
    UCHAR state, input;
    int tempC;
    unsigned size;					    /* whenever state */
    char *begin;					    /*	is 0, we're in*/
							    /*	column zero   */
    BOOL parsechar;				/* flag to examine char. type */
    BOOL inQuotes = (BOOL) FALSE;	      /* flag when inside quote marks */

    begin = s;
    c = lgetc();
    if (type == STRING) state = (UCHAR) 2;
    else if (WHITESPACE(c)) {
	state = (UCHAR) 2;
	c = skipWhiteSpace(FROMLOCAL);
    }
    else if (c == ESCH) {
	c = lgetc();
	if (WHITESPACE(c)) {
	    state = (UCHAR) 2;
	    c = skipWhiteSpace(FROMLOCAL);
	}
	else {
	    UngetTxtChr(c, file);
	    c = ESCH;
	}
    }
    else state = (UCHAR) 1;				    /* default state  */
    for (;;c = lgetc()) {
	if (c == '\"') inQuotes = (BOOL) !inQuotes;
	parsechar = 1;			     /* Default is examine character. */
	if (c == ESCH && !inQuotes && type == VALUE) {
	    c = lgetc();
	    switch (c) {
		case '$': case ESCH:		/* Special characters; must   */
		case '{': case '}':		/* not elide esch from string */
		case '(': case ')':
		case '!': case '-': case '@':
		    *s++ = ESCH;
		    if (s == end) {
			if (string == NULL) {			/* Increase size of s */
			    string = allocate(MAXBUF<<1);
			    _ftcsncpy(string,begin,MAXBUF);
			    s = string + MAXBUF;
			    size = MAXBUF << 1;
			    end = string + size;
			}
			else {
			    if ((size + MAXBUF < size)	    /* overflow error */
				|| !(string = REALLOC(string,size+MAXBUF)))
				makeError(line, MACRO_TOO_LONG);
			    s = string + size;
			    size += MAXBUF;
			    end = string + size;
			}
			begin = string;
		    }
		case '#': case '\n':		    /* elide esch right now!  */
		case '\\': case '\"':
		    input = DEFAULT_;
		    parsechar = 0;		    /* DON'T examine character*/
		    break;
		default:
		    break;			    /* DO examine character.  */
	    }
	}
	else if (c == ESCH) {
	    c = lgetc();
	    UngetTxtChr(c, file);
	    c = ESCH;
	}
	if (parsechar) {
	    switch (c) {
		case '#':   input = COMMENT_;	    break;
		case '=':   input = EQUALS_;	    break;
		case ':':   input = COLON_;	    break;
		case '$':   input = DOLLAR_;	    break;
		case '(':   input = OPENPAREN_;     break;
		case ')':   input = CLOSEPAREN_;    break;
		case '\\':  input = BACKSLASH_;     break;
		case '\n':
		case EOF:   input = NEWLINE_;	    break;
		case ' ':
		case '\t':  input = WHITESPACE_;    break;
		case '*':   input = STAR_;	    break;
		case '@':
		case '<':
		case '?':   input = SPECIAL1_;	    break;
		case 'F':
		case 'D':
		case 'B':
		case 'R':   input = SPECIAL2_;	    break;
		case ';':   input = (UCHAR) (!state && init ? COMMENT_ : DEFAULT_);
			    break;		  /* Handle comments in tools.ini */

		default:    input = (UCHAR) (MACRO_CHAR(c) ? MACROCHAR_:DEFAULT_);
			    break;
	    }
	}
	if (input == SPECIAL1_ && type == STRING && c == '<') {
	    if ((tempC = lgetc()) == '<') {		    /* << means start */
		s = makeInlineFiles(s, &begin, &end);	   /* an inline file  */
		input = NEWLINE_;
		c = '\n';
	    }
	    else {
		UngetTxtChr(tempC,file);
	    }
	    state = stringStates[state][input];
	}
	else if (input == COMMENT_) {			    /* Handle comments*/
	    if (!state) {
		inQuotes = (BOOL) FALSE;
		skipComments(FROMLOCAL);
		++line;
		continue;
	    }
	    else if (type == VALUE) state = OK;   /* don't elide from command */
	    else state = stringStates[state][input];
	}
	else state = stringStates[state][input];
	if (state == OK) {			    /* Accept end of string   */
	    inQuotes = (BOOL) FALSE;
	    UngetTxtChr(c,file);
	    /*
	     * Strip trailing whitespace from string.  Easier to do it here,
	     * else we have to treat a multi-string value (OBJS=a b c) as
	     * separate tokens.  [RB]
	     */
	    while (s > begin && _istspace(s[-1]))
		--s;
	    *s = '\0';
	    if (string) {
		if (s = REALLOC(string,s-string+1))
		    string = s;
	    }
	    else string = makeString(begin);
	    return;
	}
	else if (ON(state,ERROR_MASK))		    /* Error code from table  */
	    makeError(line,(state&~ERROR_MASK)+FATAL_ERR,c);
	if (!state) {				    /* Col 0; we just hit \nl */
	    *--s = ' '; 			    /* so treat it like white-*/
	    ++s;  ++line;			    /* space; overwrite the   */
	    colZero = TRUE;			    /* backslash with a space.*/
	    c = lgetc();
	    colZero = FALSE;
	    if (WHITESPACE(c)) {
		state = 2;
		do {
		    c = lgetc();
		} while (WHITESPACE(c));
	    }
	    UngetTxtChr(c,file);
	}
	else {					    /* Keep storing string    */
	    *s++ = (char) c;
	    if (s == end) {
		if (!string) {			    /* Increase size of s     */
		    string = allocate(MAXBUF<<1);
		    _ftcsncpy(string,begin,MAXBUF);
		    s = string + MAXBUF;
		    size = MAXBUF << 1;
		    end = string + size;
		}
		else {
		    if ((size + MAXBUF < size)		    /* overflow error */
			|| !(string = REALLOC(string,size+MAXBUF)))
			makeError(line, MACRO_TOO_LONG);
		    s = string + size;
		    size += MAXBUF;
		    end = string + size;
		}
	    }
	}
    }
}

/*  ----------------------------------------------------------------------------
 *  getName()
 *
 *  arguments:	    s	    pointer into buffer that will hold string
 *			    (s is pointing to buf+1 when passed, because
 *			    the caller, getToken(), has already seen and
 *			    saved one char)
 *		    init    global boolean value -- TRUE if tools.ini is the
 *				file being lexed
 *			    used by routine called - lgetc()
 *		    end     pointer to end of s[]
 *
 *  actions:	    gets all chars up to first token delimiter and stores
 *			them in s[] (delimiters are ' ', '\t', '\n' and (when
 *			not inside a macro invocation) ':' and '='
 *			note that backslash-newline is treated as a space,
 *			which is a delimiter
 *		    if the current input char is '$' this must be a macro
 *			invocation
 *			if the macro name is in parentheses
 *			    get all chars up to and including close paren
 *				(if ')' not found, error)
 *
 *  We check the syntax within the name here -- thus errors in macro
 *  invocation syntax will be caught.  Special macros cannot be used
 *  as part of names, with the exception of the dynamic dependency macros.
 *
 *  We can probably never overrun our buffer, because it would be extremely
 *  difficult for the user to get a name with 1024 characters or more into
 *  his makefile.
 *
 *  we never end up in column zero, because we push the delimiter back
 *  out on the input
 *
 *  uses state table defined in table.h, defs from grammar.h
 *
 *    modifies:   line    (possibly) thru' call to lgetc()
 *		  file    (possibly) if lgetc() finds a !include
 *		  fName   (possibly) if lgetc() finds a !include
 */

LOCAL void NEAR
getName(s,end)
char *s;
char *end;						    /* pts to end of s*/
{
    int c;
    UCHAR state,
	  input;
    BOOL seenBackSlash = FALSE;
    BOOL fQuoted = FALSE;
    char *beg = s - 1;
    BOOL parsechar;				/* flag to examine char. type */

    switch (*(s-1)) {
	case '$':   state = (UCHAR) 2;	break;
	case '{':   state = (UCHAR) 8;	break;
	case '"':   fQuoted = TRUE; state = (UCHAR)16; break;
	default:    state = (UCHAR) 0;	break;
    }
    for (;;) {
	c = lgetc();
	parsechar = 1;				  /* Default is examine char. */
	if (c == ESCH) {
	    c = lgetc();
	    switch (c) {
		case '{': case '}':		/* Special characters; must   */
		case '(': case ')':		/* not elide esch from string */
		case '$': case ESCH:
		    *s++ = ESCH;
		case '#': case '\n': case '\\':     /* elide esch right now!  */
		    input = DEFAULT_;
		    parsechar = 0;		    /* DON'T examine character*/
		    break;
		default:
		    break;			    /* DO examine character.  */
	    }
	}
	if (parsechar) {
	    switch (c) {
		case '#' :  input = COMMENT_;	    break;
		case '=' :  input = EQUALS_;	    break;
		case ';' :  input = SEMICOLON_;     break;
		case ':' :  input = COLON_;	    break;
		case '$' :  input = DOLLAR_;	    break;
		case '(' :  input = OPENPAREN_;     break;
		case ')' :  input = CLOSEPAREN_;    break;
		case '{' :  input = OPENCURLY_;     break;
		case '}' :  input = CLOSECURLY_;    break;
		case ' ' :
		case '\t':  input = (UCHAR)((fQuoted)
				       ? DEFAULT_ : WHITESPACE_);
			    break;
		case '\n':
		case EOF :  input = NEWLINE_;	    break;
		case '\\':  input = BKSLSH_;	    break;
		case '"' :  input = QUOTE_;	   break;
		//Add support for $* and $@ on the dependency line
		default  :  if (ON(actionFlags, A_DEPENDENT))
				input = (UCHAR)((MACRO_CHAR(c) || c == '*' || c == '@')
					 ?MACROCHAR_:DEFAULT_);
			    else
				input = (UCHAR)(MACRO_CHAR(c)?MACROCHAR_:DEFAULT_);
			    break;
	    }
	}
	state = nameStates[state][input];
	//Cheat lex table to think that you are handling quoted string case
	if (fQuoted && state == 1)
	    state = 16;
	//seenBackSlash is used to provide lookahead when \ is seen on a
	//dependency line
	if (seenBackSlash)
	    //if \ followed by \n then use it as a continuation
	    if (input == NEWLINE_) {
		++line;
		colZero = TRUE;
		c = lgetc();
		colZero = FALSE;
		if (WHITESPACE(c)) {
		    state = OK;
		    do {
			c = lgetc();
		    } while (WHITESPACE(c));
		}
		else
		    state = (UCHAR)((s == buf + 1) ? BEG : DEF);
	    }
	    else
		*s++ = '\\';
	seenBackSlash = FALSE;
	if (state == OK) {
	    if (s >= end)
		makeError(line,NAME_TOO_LONG);
	    UngetTxtChr(c,file);
	    *s = '\0';
	    removeTrailChars(beg);
	    return;
	}
	else if (ON(state,ERROR_MASK))
	    makeError(line,(state&~ERROR_MASK)+FATAL_ERR,c);
	if (state == BKS) {
	    //set lookahead flag
	    seenBackSlash = TRUE;
	}
	else
	    *s++ = (char) c;
    }
}

/*** createDosTmp -- Creates a unique temporary file. ******************
*
* Scope:
*  Global.
*
* Purpose:
*  To create a unique temporary file by calling _mktemp() but it gets
*  over _mktemp() limitation to be able to create more files.
*
* Input:
*  path -- The buffer initially contain the directory to store the temp
*          file.  On exit, if success, the temp file is appended to it.
*          In case of failure, the its contents is undetermined.
*
* Output:
*  If successful, temporary file name is appended to path and
*  the function returns the file pointer, else NULL.
*
* Errors/Warnings:
*
* Assumes:
*
* Modifies Globals:
*  None.
*
* Uses Globals:
*  None.
*
* Notes:
*
* History:
*  06-Apr-1993 HV Change createDosTmp() to use _makepath()
*
*******************************************************************************/
FILE * NEAR
createDosTmp(char *path)
{
    FILE 		*fd;
    static char 	template[] = "nmXXXXXX";
    static char		szExtension[] = ".";	// just a dot to make Mr.
						// linker happy.

    // CONSIDER: is the size of szDir too big for the stack?  We can
    //           make it a little smaller if we take the risk.
    char		szDir[_MAX_PATH];

    // CONSIDER: The path supplied by the caller might contain both
    //           the drive and probably some level of directories
    //           (e.g. c:\win\tmp)  Right now, _makepath happily takes
    //           the whole thing as the directory component, but that
    //           might change in the future.  In such case, we should
    //           first break up path to drive/dir compents before we
    //           construct the full pathname of the template.  There
    //           is something to watch out: for "c:\win\tmp", _splitpath
    //           will split as: "c:", "\win\", and "tmp", which is not
    //           what we want.  To fix it, append a backslash to the
    //           end before calling _splitpath.  "c:\win\tmp\" will
    //           be broken up correctly to "c:" and "\win\tmp\"

    if (!path || !*path)	// If path is empty, use "."
	_ftcscpy(szDir, ".");
    else
	_ftcscpy(szDir, path);

    // Construct the full pathname.  _mktemp() doesn't seem to like
    // template with trailing dot (".") so instead of specifying "."
    // for the extension, we defer it later and _ftcscat the dot to
    // its tail.
    _makepath(path, NULL, szDir, template, NULL);

    if(_mktemp(path) == NULL)
    {
	/*
	 * Mktemp() has a limit of 27 files per template.  If it fails, assume
	 * the limit has overflowed and increment the second letter of the
	 * template.
	 */
	if (template[1] == 'z')
	    template[1] = 'a';
	else
	    ++template[1]; // ASSUMPTION: that this will work with DBCS

	_makepath(path, NULL, szDir, template, NULL);

	if(_mktemp(path) == NULL)
	    return(NULL);
    }

    // add a trailing "." for the linker's sake
    _ftcscat(path, szExtension);

    // Open the file and return the file's descriptor.
    return(fd = FILEOPEN(path, "w"));
} // createDosTmp


LOCAL void NEAR
popFileStack()
{
    if (fclose(file) == EOF)
	makeError(0, ERROR_CLOSING_FILE, fName);
    FREE(fName);
    file = incStack[--incTop].file;
    fName = incStack[incTop].name;
    line = incStack[incTop].line;
}

/*  -----------------------------------------------------------------
 *  include()  -- handle include files
 *
 *  arguments:  c       first non-whitespace char after the string
 *			INCLUDE on the line...
 *		colZero global boolean value, set if currently at
 *			column zero of a file.
 *
 *  modifies:   line     global line count - if include file opened
 *		file     global pointer to current file
 *		fName    global pointer to name of current file
 *		colZero  global boolean value, changed if include
 *			 file opened and char from colZero is returned
 */


LOCAL UCHAR NEAR
include(c)
int c;
{
    unsigned n;
    char *s;

    if (c == '\n' || c == EOF)
	makeError(line,SYNTAX_NO_NAME);
    *buf = (char) c;
    if (!fgets(buf+1,MAXBUF - 1,file)) {
	if (feof(file))
	    makeError(line,SYNTAX_UNEXPECTED_TOKEN,"EOF");
	makeError(line,CANT_READ_FILE);
    }
    n = _ftcslen(buf) - 1;
    if (buf[n] == '\n') buf[n] = '\0';
    s = buf;
    while (WHITESPACE(*s)) ++s;
    return(processIncludeFile(s));
}

/*  -------------------------------------------------------------------
 *  processIncludeFile()  -- checks for include file and switches state
 *
 *  arguments:  s       buffer that has include file name
 *		colZero global boolean value, set if currently at
 *			column zero of a file.
 *              init    global boolean - set if tools.ini is being lexed
 *			used by lgetc() which is called from here...
 *
 *
 *  modifies:   line     global line count - if include file opened
 *		file     global pointer to current file
 *		fName    global pointer to name of current file
 *		colZero  global boolean value, changed if include
 *			 file opened and char from colZero is returned
 */

UCHAR NEAR
processIncludeFile(s)
char *s;
{
    MACRODEF *m;
    void *findBuf = _alloca(resultbuf_size);
    NMHANDLE searchHandle;
    char *t,
	 *p,
	 *u;
    int c = 0;
    int i;


    if (!*s || *s == '#')
	makeError(line,SYNTAX_NO_NAME);
    if (t = _ftcspbrk(s," \t#")) {
	if (*t == '#') c = *t;
	*t = '\0';
	if (!c) {
	    for (u = t; *++u;) {			    /* check for extra*/
		if (*u == '#') break;			    /*	text on line  */
		else if (!WHITESPACE(*u))
		    makeError(line,SYNTAX_UNEXPECTED_TOKEN,u);
	    }
	}
    }
    else t = s + _ftcslen(s);
    if (*s == '<' && *(t-1) == '>') {
	char * pt;

	*--t = '\0';
	p = removeMacros(++s);
	p = p == s ? makeString(s) : p;
	t = (m = findMacro("INCLUDE")) ? m->values->text : (char*) NULL;
	if (t != NULL) //expand INCLUDE macro before passing it on
	    {
	    char * pt1;

	    pt1= makeString(t);
	    pt = removeMacros(pt1);
	    if (pt != pt1)
		FREE (pt1);  //we've got a new string, free old one
	    }
	else
	    pt = NULL;
	if (!(u = searchPath(pt, p, findBuf, &searchHandle)))
	    makeError(line, CANT_OPEN_FILE, p);
	if (pt) FREE (pt);
	FREE(p);
	s = u;
    }
    else {
	if (*s == '"' && *(t-1) == '"') {
	    *--t = '\0';
	    ++s;
	}
	p = removeMacros(s);
	p = p == s ? makeString(s) : p;
	if (!findFirst(p,&findBuf, &searchHandle))
	    if (!_ftcspbrk(p, "\\/:")) {
		//use C sematics for include
		for (i = incTop;i >= 0;i--) {
		    t = (i == incTop) ? fName : incStack[i].name;
		    if (!(t = getPath(t)))
			continue;
		    u = (char *)allocate(_ftcslen(t) + 1 + _ftcslen(p) + 1);
		    _ftcscat(_ftcscat(_ftcscpy(u, t), "\\"), p);
		    if (findFirst(u, &findBuf, &searchHandle)) {
			s = u;
			FREE(t);
			break;
		    }
		    FREE(t);
		    FREE(u);
		}
		FREE(p);
		if (i < 0)
		    makeError(line,CANT_OPEN_FILE,s);
	    }
	    else
		makeError(line,CANT_OPEN_FILE,p);
    }
    for (i = 0; i < incTop; ++i) { 			    /* test for cycles*/
	if (!_ftcsicmp(s,incStack[i].name))
	    makeError(line,CYCLE_IN_INCLUDES,s);
    }
    incStack[incTop].file = file;			    /* push info on   */
    incStack[incTop].line = line;			    /*	stack	      */
    incStack[incTop++].name = fName;
    currentLine = 0;
    if (!(file = FILEOPEN(s,"rt")))			       /* read, text mode*/
	makeError(line,CANT_OPEN_FILE,s);
    fName = makeString(s);
    line = 1;
    colZero = TRUE;					    /* parser needs to*/
    c = lgetc();					    /*	see some kind */
    if (colZero = (BOOL) !WHITESPACE(c)) {		    /*	of newline to */
	UngetTxtChr(c,file); 				    /*	initialize it */
	return(NEWLINE);				    /*	for this file */
    }
    return(NEWLINESPACE);
}

#if 0 // The old getPath() is bracketted between the #if 0/#endif pair
LOCAL char * NEAR
getPath(s)
char *s;
{
    char *path = (char *)allocate(_ftcslen(s));
    char *t = _ftcsrchr(s, '\\'),
	 *u;
    int n;

    if (t && (u = _ftcsrchr(s, '/')) > t)
	t = u;
    if (!t)
	n = s[1] == ':' ? 2 : 0;
    else
	n = t - s;
    _ftcsncpy(path, s, n);
    path[n] = '\0';
    return(path);
}
#endif

/*** getPath -- return the drive/directory parts of a full path ***************
*
* Scope:
*  Local
*
* Purpose:
*  This function returns the drive/directory parts of a full path.  Space is
*  allocated for the resulting string, so the caller is responsible for freeing
*  it after use.
*
* Input:
*  pszFullPath -- The full pathname.
*
* Output:
*
* Errors/Warnings:
*
* Assumes:
*  Pathnames use MS-DOS file naming convension.
*
* Modifies Globals:
*  None.
*
* Uses Globals:
*  None.
*
* Notes:
*  To allocate temporary memory for the drive and path components, I have used
*  _MAX_DRIVE and _MAX_DIR.  Under Windows NT there are two possibilities:
*  1. These two parameters can be so large that the stack will be overflow
*  2. They are not large enough (?)
*
* History:
*  22-Mar-1993 HV Rewrite getPath() to make use of the new _splitpath() and
*		  _makepath() functions.
*******************************************************************************/
LOCAL char * NEAR
getPath(const char *pszFullPath)
{
    // HV: Are _MAX_DRIVE and _MAX_DIR good size under NT?  Does it
    // work under Windows NT?
    char szDrive[_MAX_DRIVE];
    char szDir[_MAX_DIR];
    char *szPath;
    char *pszSlash;

    // Separate the components of the fullpath
    _splitpath(pszFullPath, szDrive, szDir, NULL, NULL);

    // Allocate just enough memory to hold the drive/path combo then
    // Glue just the drive and dir component back together.
    szPath = (char *) rallocate(_ftcslen(szDrive) + _ftcslen(szDir) + 1);
    _makepath(szPath, szDrive, szDir, NULL, NULL);

    // Eliminate the trailing slash/blackslash to retain compatibility with
    // the older version of getPath()
    pszSlash = szPath + _ftcslen(szPath) - 1;
    if ('\\' == *pszSlash || '/' == *pszSlash)
	*pszSlash = '\0';

    return szPath;
}