Subversion Repositories tendra.SVN

/*

    		 Crown Copyright (c) 1997

    This TenDRA(r) Computer Program is subject to Copyright

    owned by the United Kingdom Secretary of State for Defence

    acting through the Defence Evaluation and Research Agency

    (DERA).  It is made available to Recipients with a

    royalty-free licence for its use, reproduction, transfer

    to other parties and amendment for any purpose not excluding

    product development provided that any such use et cetera

    shall be deemed to be acceptance of the following conditions:-

        (1) Its Recipients shall ensure that this Notice is

        reproduced upon any copies or amended versions of it;

        (2) Any amended version of it shall be clearly marked to

        show both the nature of and the organisation responsible

        for the relevant amendment or amendments;

        (3) Its onward transfer from a recipient to another

        party shall be deemed to be that party's acceptance of

        these conditions;

        (4) DERA gives no warranty or assurance as to its

        quality or suitability for any purpose and DERA accepts

        no liability whatsoever in relation to any use to which

        it may be put.

*/

#include "config.h"

#include "system.h"

#include "version.h"

#include "c_types.h"

#include "hashid_ops.h"

#include "id_ops.h"

#include "nspace_ops.h"

#include "error.h"

#include "catalog.h"

#include "option.h"

#include "buffer.h"

#include "char.h"

#include "dump.h"

#include "file.h"

#include "hash.h"

#include "lex.h"

#include "literal.h"

#include "macro.h"

#include "namespace.h"

#include "predict.h"

#include "preproc.h"

#include "print.h"

#include "symbols.h"

#include "syntax.h"

#include "ustring.h"

#include "xalloc.h"

/*

    LIST OF FREE LEXICAL TOKENS

    All the free lexical tokens are formed into a list.

*/

PPTOKEN *free_tokens = NULL ;

static LIST ( PPTOKEN_P ) alloc_tokens = NULL_list ( PPTOKEN_P ) ;

/*

    ALLOCATE A NEW TOKEN

    This routine allocates a new token from the list free_tokens.

*/

PPTOKEN *new_pptok

    PROTO_Z ()

{

    PPTOKEN *p = free_tokens ;

    if ( p == NULL ) {

	PPTOKEN *q ;

	int i, n = 2000 ;

	p = xmalloc_nof ( PPTOKEN, n ) ;

	CONS_pptok ( p, alloc_tokens, alloc_tokens ) ;

	q = p ;

	for ( i = 1 ; i < n ; i++ ) {

	    q->next = q + 1 ;

	    q++ ;

	}

	q->next = NULL ;

    }

    free_tokens = p->next ;

    p->pp_opts = real_opts ;

    return ( p ) ;

}

/*

    FREE A SINGLE TOKEN

    This macro frees the single token P by adding it to the list of all

    free tokens.

*/

#define free_pptok( P )\

    {\

	( P )->next = free_tokens ;\

	free_tokens = ( P ) ;\

    }

/*

    FREE A LIST OF TOKENS

    This routine adds the list of tokens p to the list of all free tokens.

*/

void free_tok_list

    PROTO_N ( ( p ) )

    PROTO_T ( PPTOKEN *p )

{

    PPTOKEN *q = p ;

    if ( q == NULL ) return ;

    while ( q->next ) q = q->next ;

    q->next = free_tokens ;

    free_tokens = p ;

    return ;

}

/*

    FREE ALL ALLOCATED PREPROCESSING TOKENS

    This routine frees all the space allocated for preprocessing tokens.

    It should only be called after the input has been completely processed.

*/

void term_macros

    PROTO_Z ()

{

    LIST ( PPTOKEN_P ) p = alloc_tokens ;

    while ( !IS_NULL_list ( p ) ) {

	PPTOKEN *q ;

	DESTROY_CONS_pptok ( destroy, q, p, p ) ;

	xfree_nof ( q ) ;

    }

    alloc_tokens = p ;

    free_tokens = NULL ;

    return ;

}

/*

    COPY A TOKEN

    This macro copies the contents of the token with token value T and data

    Q into P.

*/

#define copy_pptok( P, T, Q )\

    {\

	( P )->tok = ( T ) ;\

	( P )->pp_data = ( Q )->pp_data ;\

	( P )->pp_opts = ( Q )->pp_opts ;\

	( P )->pp_space = ( Q )->pp_space ;\

    }

/*

    ASSIGN TOKEN COMPONENTS

    This routine assigns the token components for the token t, which has

    just been read from the input file (or faked on occasions - these are

    indicated) into p.  It is only necessary to call this routine is T is

    less than or equal to LAST_COMPLEX_TOKEN (defined in symbols.h).  If any

    cases are added to this routine then it may be necessary to change the

    value of this macro.

*/

void token_parts

    PROTO_N ( ( t, p ) )

    PROTO_T ( int t X PPTOKEN *p )

{

    switch ( t ) {

	case lex_identifier : {

	    /* Identifiers */

	    HASHID nm = token_hashid ;

	    IDENTIFIER id = DEREF_id ( hashid_id ( nm ) ) ;

	    p->pp_data.id.hash = nm ;

	    p->pp_data.id.use = id ;

	    break ;

	}

	case lex_char_Hlit :

	case lex_string_Hlit :

	case lex_wchar_Hlit :

	case lex_wstring_Hlit : {

	    /* String and character literals */

	    string s1 = token_buff.start ;

	    gen_size n = ( gen_size ) ( token_buff.posn - s1 ) ;

	    string s2 ;

	    if ( n < 2 ) {

		/* Optimise for small strings */

		s2 = xustrcpy ( s1 ) ;

	    } else {

		s2 = xustr ( n + 1 ) ;

		xumemcpy ( s2, s1, n ) ;

		s2 [n] = 0 ;

	    }

	    p->pp_data.str.start = s2 ;

	    p->pp_data.str.end = s2 + n ;

	    break ;

	}

	case lex_integer_Hlit : {

	    /* Integer and floating-point literals */

	    p->pp_data.text = xustrcpy ( token_buff.start ) ;

	    break ;

	}

	case lex_hash_Hif :

	case lex_hash_Helif : {

	    /* Target dependent conditionals */

	    p->pp_data.exp = crt_hash_if_exp ;

	    break ;

	}

	case lex_unknown : {

	    /* Unknown characters */

	    int i ;

	    string s1 = token_buff.start ;

	    string s2 = p->pp_data.buff ;

	    ASSERT ( MULTI_WIDTH <= sizeof ( p->pp_data.buff ) ) ;

	    for ( i = 0 ; i < MULTI_WIDTH ; i++ ) s2 [i] = s1 [i] ;

	    break ;

	}

    }

    return ;

}

/*

    REMOVE ANY IGNORED TOKENS FROM A LIST

    This routine removes any ignored tokens from the list tok, returning

    the result.

*/

PPTOKEN *clean_tok_list

    PROTO_N ( ( toks ) )

    PROTO_T ( PPTOKEN *toks )

{

    unsigned long sp = 0 ;

    PPTOKEN p0, *p = &p0 ;

    PPTOKEN *q ;

    p->next = toks ;

    while ( q = p->next, q != NULL ) {

	if ( q->tok == lex_ignore_token ) {

	    sp |= q->pp_space ;

	    p->next = q->next ;

	    free_pptok ( q ) ;

	    q = p->next ;

	    if ( q == NULL ) break ;

	} else {

	    if ( sp ) {

		q->pp_space |= sp ;

		sp = 0 ;

	    }

	}

	p = q ;

    }

    return ( p0.next ) ;

}

/*

    READ A LINE OF TOKENS

    This routine reads the sequence of preprocessing tokens comprising a

    preprocessing directive (for example, a macro definition).  If t1 is

    not lex_ignore_token then it is taken to be the first token in the

    definition, similarly tn gives the last token.

*/

PPTOKEN *read_line

    PROTO_N ( ( t1, tn ) )

    PROTO_T ( int t1 X int tn )

{

    int t = t1 ;

    unsigned long sp = 0 ;

    PPTOKEN dummy_tok, *this_tok = &dummy_tok ;

    if ( t == lex_ignore_token ) {

	t = read_token () ;

	update_column () ;

	if ( in_preproc_dir ) preproc_loc = crt_loc ;

    }

    while ( t != lex_newline && t != lex_eof ) {

	this_tok->next = new_pptok () ;

	this_tok = this_tok->next ;

	this_tok->tok = t ;

	if ( t <= LAST_COMPLEX_TOKEN ) token_parts ( t, this_tok ) ;

	this_tok->pp_space = ( sp & WHITE_MASK ) ;

	sp = skip_white ( 0 ) ;

	t = read_token () ;

	update_column () ;

	if ( in_preproc_dir ) preproc_loc = crt_loc ;

    }

    if ( tn != lex_ignore_token ) {

	this_tok->next = new_pptok () ;

	this_tok = this_tok->next ;

	this_tok->tok = tn ;

	token_parts ( tn, this_tok ) ;

	this_tok->pp_space = ( sp & WHITE_MASK ) ;

    }

    this_tok->next = NULL ;

    if ( in_preproc_dir ) IGNORE skip_to_end () ;

    return ( dummy_tok.next ) ;

}

/*

    COPY A LIST OF TOKENS

    This routine copies the list of tokens toks, excluding any ignored

    tokens.

*/

static PPTOKEN *copy_tok_list

    PROTO_N ( ( toks ) )

    PROTO_T ( PPTOKEN *toks )

{

    PPTOKEN *ptr_tok ;

    PPTOKEN dummy_tok, *this_tok = &dummy_tok ;

    for ( ptr_tok = toks ; ptr_tok != NULL ; ptr_tok = ptr_tok->next ) {

	int t = ptr_tok->tok ;

	if ( t != lex_ignore_token ) {

	    this_tok->next = new_pptok () ;

	    this_tok = this_tok->next ;

	    copy_pptok ( this_tok, t, ptr_tok ) ;

	}

    }

    this_tok->next = NULL ;

    return ( dummy_tok.next ) ;

}

/*

    STRINGISE A LIST OF TOKENS

    This routine turns the list of tokens toks into a string.  The result

    is built up in token_buff.  If esc is true then any '"' (or whatever

    the value of quote is) and '\' characters in string and character

    literals (including the initial and terminating quotes) are preceded

    by a '\'.  This routine is used in the implementation of the # operator,

    in macro #include directives and a couple of other preprocessing

    directives.  It returns 1 to indicate a valid string.

*/

int quote_tok_list

    PROTO_N ( ( toks, esc, quote ) )

    PROTO_T ( PPTOKEN *toks X int esc X int quote )

{

    int res = 1 ;

    string st, se ;

    int started = 0 ;

    int escaped = 0 ;

    PPTOKEN *ptr_tok ;

    character qo = ( character ) quote ;

    BUFFER *bf = clear_buffer ( &token_buff, NIL ( FILE ) ) ;

    /* Scan through tokens */

    for ( ptr_tok = toks ; ptr_tok != NULL ; ptr_tok = ptr_tok->next ) {

	character p, q ;

	int t = ptr_tok->tok ;

	if ( t == lex_ignore_token ) continue ;

	/* Print initial space if necessary */

	if ( ptr_tok->pp_space && started ) bfputc ( bf, char_space ) ;

	/* Find the token name */

	switch ( t ) {

	    case lex_identifier : {

		/* Identifiers */

		HASHID nm = ptr_tok->pp_data.id.hash ;

		st = DEREF_string ( hashid_name_etc_text ( nm ) ) ;

		bfputs ( bf, st ) ;

		break ;

	    }

	    case lex_integer_Hlit : {

		/* Integer and floating-point literals */

		st = ptr_tok->pp_data.text ;

		bfputs ( bf, st ) ;

		break ;

	    }

	    case lex_char_Hlit : {

		/* Character literals */

		p = 0 ;

		q = char_single_quote ;

		string_label : {

		    st = ptr_tok->pp_data.str.start ;

		    se = ptr_tok->pp_data.str.end ;

		    /* Prefix and opening quote */

		    if ( p ) bfputc ( bf, ( int ) p ) ;

		    if ( esc && q == qo ) bfputc ( bf, char_backslash ) ;

		    bfputc ( bf, ( int ) q ) ;

		    /* Copy string */

		    while ( st != se ) {

			character c = *( st++ ) ;

			if ( c == qo || c == char_backslash ) {

			    /* Escaped characters */

			    if ( esc ) bfputc ( bf, char_backslash ) ;

			}

			bfputc ( bf, ( int ) c ) ;

		    }

		    /* Closing quote */

		    if ( esc && q == qo ) bfputc ( bf, char_backslash ) ;

		    bfputc ( bf, ( int ) q ) ;

		}

		break ;

	    }

	    case lex_wchar_Hlit : {

		/* Wide character literals */

		p = char_L ;

		q = char_single_quote ;

		goto string_label ;

	    }

	    case lex_string_Hlit : {

		/* String literals */

		p = 0 ;

		q = char_quote ;

		goto string_label ;

	    }

	    case lex_wstring_Hlit : {

		/* Wide string literals */

		p = char_L ;

		q = char_quote ;

		goto string_label ;

	    }

	    case lex_unknown : {

		/* Unknown characters */

		unsigned long u ;

		int ch = CHAR_SIMPLE ;

		u = get_multi_char ( ptr_tok->pp_data.buff, &ch ) ;

		if ( ch == CHAR_SIMPLE ) {

		    bfputc ( bf, ( int ) u ) ;

		} else {

		    print_char ( u, ch, 0, bf ) ;

		}

		break ;

	    }

	    case lex_macro_Harg : {

		/* Macro parameters */

		HASHID nm = ptr_tok->pp_data.par.hash ;

		st = DEREF_string ( hashid_name_etc_text ( nm ) ) ;

		bfputs ( bf, st ) ;

		break ;

	    }

	    default : {

		/* Symbols */

		st = token_name ( t ) ;

		bfputs ( bf, st ) ;

		break ;

	    }

	}

	started = 1 ;

    }

    /* End of string */

    bfputc ( bf, 0 ) ;

    bf->posn-- ;

    /* Check for legal strings */

    st = bf->start ;

    se = bf->posn ;

    while ( st != se ) {

	if ( escaped ) {

	    escaped = 0 ;

	} else {

	    character c = *st ;

	    if ( c == qo ) res = 0 ;

	    if ( c == char_backslash ) escaped = 1 ;

	}

	st++ ;

    }

    if ( escaped ) res = 0 ;

    return ( res ) ;

}

/*

    CONCATENATE TWO TOKENS

    This routine concatenates the two tokens p and q into a single token.

    This is used to implement the ## operator.  If the result is a valid

    preprocessing token then p is overwritten by the result and 1 is

    returned.  Otherwise p and q are unchanged and 0 is returned.

*/

static int concat_pptoks

    PROTO_N ( ( p, q ) )

    PROTO_T ( PPTOKEN *p X PPTOKEN *q )

{

    int a = p->tok ;

    int b = q->tok ;

    unsigned long sa = p->pp_space ;

    unsigned long sb = q->pp_space ;

    p->pp_space = ( sa | sb ) ;

    q->pp_space = 0 ;

    if ( a >= FIRST_SYMBOL && a <= LAST_SYMBOL ) {

	if ( b >= FIRST_SYMBOL && b <= LAST_SYMBOL ) {

	    /* Two symbols may combine to give another symbol */

	    int c ;

	    string s = token_buff.start ;

	    ustrcpy_v ( s, token_name ( a ) ) ;

	    ustrcpy_v ( s + ustrlen ( s ), token_name ( b ) ) ;

	    for ( c = FIRST_SYMBOL ; c <= LAST_SYMBOL ; c++ ) {

		if ( ustreq ( s, token_name ( c ) ) ) {

		    /* Token found - check options */

		    p->tok = c ;

		    if ( c >= FIRST_C_SYMBOL && c <= LAST_C_SYMBOL ) {

			return ( 1 ) ;

		    }

#if LANGUAGE_CPP

		    if ( c >= FIRST_CPP_SYMBOL && c <= LAST_CPP_SYMBOL ) {

			return ( 1 ) ;

		    }

#endif

		    if ( c >= FIRST_EXTRA_SYMBOL && c <= LAST_EXTRA_SYMBOL ) {

			if ( allow_extra_symbols ) return ( 1 ) ;

		    }

		    if ( c >= FIRST_DIGRAPH && c <= LAST_DIGRAPH ) {

			if ( allow_digraphs ) return ( 1 ) ;

		    }

		    p->tok = a ;

		}

	    }

	    return ( 0 ) ;

	} else if ( a == lex_dot && b == lex_integer_Hlit ) {

	    /* A dot may start a number */

	    string s = q->pp_data.text ;

	    if ( s [0] == char_dot ) return ( 0 ) ;

	    p->tok = lex_integer_Hlit ;

	    p->pp_data.text = xustrcat ( token_name ( a ), s ) ;

	    return ( 1 ) ;

	} else if ( a == lex_backslash && b == lex_identifier ) {

	    /* A backslash may start a universal character */

	    /* NOT YET IMPLEMENTED */

	    /* EMPTY */

	}

    } else if ( a == lex_identifier ) {

	HASHID nm = p->pp_data.id.hash ;

	string s = DEREF_string ( hashid_name_etc_text ( nm ) ) ;

	if ( b == lex_identifier ) {

	    /* Two identifiers give another identifier */

	    HASHID nm2 = q->pp_data.id.hash ;

	    string s2 = DEREF_string ( hashid_name_etc_text ( nm2 ) ) ;

	    s = xustrcat ( s, s2 ) ;

	    nm = lookup_name ( s, hash ( s ), 2, lex_identifier ) ;

	    p->pp_data.id.hash = nm ;

	    p->pp_data.id.use = DEREF_id ( hashid_id ( nm ) ) ;

	    return ( 1 ) ;

	} else if ( b == lex_integer_Hlit ) {

	    /* An identifier and a number may give an identifier */

	    character c ;

	    string n = q->pp_data.text ;

	    while ( c = *( n++ ), c != 0 ) {

		if ( c == char_dot || c == char_plus || c == char_minus ) {

		    /* The number must be entirely alphanumeric */

		    return ( 0 ) ;

		}

	    }

	    s = xustrcat ( s, q->pp_data.text ) ;

	    nm = lookup_name ( s, hash ( s ), 2, lex_identifier ) ;

	    p->pp_data.id.hash = nm ;

	    p->pp_data.id.use = DEREF_id ( hashid_id ( nm ) ) ;

	    return ( 1 ) ;

	} else if ( s [0] == char_L && s [1] == 0 ) {

	    /* An L may start a wide character or string */

	    if ( b == lex_char_Hlit ) {

		p->tok = lex_wchar_Hlit ;

		p->pp_data.str.start = q->pp_data.str.start ;

		p->pp_data.str.end = q->pp_data.str.end ;

		return ( 1 ) ;

	    } else if ( b == lex_string_Hlit ) {

		p->tok = lex_wstring_Hlit ;

		p->pp_data.str.start = q->pp_data.str.start ;

		p->pp_data.str.end = q->pp_data.str.end ;

		return ( 1 ) ;

	    }

	}

    } else if ( a == lex_integer_Hlit ) {

	string s = p->pp_data.text ;

	if ( b == lex_identifier ) {

	    /* A number followed by an identifier is a number */

	    HASHID nm = q->pp_data.id.hash ;

	    string s2 = DEREF_string ( hashid_name_etc_text ( nm ) ) ;

	    p->pp_data.text = xustrcat ( s, s2 ) ;

	    return ( 1 ) ;

	} else if ( b == lex_integer_Hlit ) {

	    /* Two numbers form another number */

	    string s2 = q->pp_data.text ;

	    p->pp_data.text = xustrcat ( s, s2 ) ;

	    return ( 1 ) ;

	} else if ( b == lex_dot || b == lex_ellipsis ) {

	    /* A number followed by a sequence of dots is a number */

	    p->pp_data.text = xustrcat ( s, token_name ( b ) ) ;

	    return ( 1 ) ;

	} else if ( b == lex_plus || b == lex_minus ) {

	    /* A sign may terminate a number after e or E */

	    unsigned n = ( unsigned ) ustrlen ( s ) - 1 ;

	    if ( s [n] == char_e || s [n] == char_E ) {

		p->pp_data.text = xustrcat ( s, token_name ( b ) ) ;

		return ( 1 ) ;

	    }

	}

    }

    return ( 0 ) ;

}

/*

    DUMMY LOCATION FOR INPUT FILE

    This dummy location represents tokens read directly from the input file.

    If present, it will always be the last element of a list of token

    locations.

*/

static PPTOKEN *dummy_loc_toks = NULL ;

static TOKEN_LOC dummy_loc = { &dummy_loc_toks, NULL } ;

TOKEN_LOC *file_loc = &dummy_loc ;

/*

    FORWARD DECLARATION

    The functions expand_macro, expand_toks and expand_tok_list are defined

    recursively.  This gives the necessary forward declarations.

*/

static PPTOKEN *expand_toks PROTO_S ( ( PPTOKEN *, TOKEN_LOC *, int ) ) ;

/*

    HANDLE OLD STYLE STRINGISING

    This routine handles the old style stringising for the definition defn

    for the given macro.  Argument replacement has already been performed

    on defn.  If this facility is enabled then in macro definitions of the

    form:

		#define f( X )	"X"

    quotes are classified as unknown characters rather than string

    terminators.  This means that the X is recognised as a macro parameter

    and is replaced during argument replacement.  The job of this routine

    is to spot these unrecognised quotes and turn them into proper strings.

*/

PPTOKEN *recognise_strings

    PROTO_N ( ( defn, macro, act ) )

    PROTO_T ( PPTOKEN *defn X HASHID macro X int act )

{

    PPTOKEN *this_tok = defn ;

    PPTOKEN *last_tok = defn ;

    while ( this_tok != NULL ) {

	if ( this_tok->tok == lex_unknown ) {

	    unsigned long u ;

	    int ch = CHAR_SIMPLE ;

	    character qo = char_question ;

	    u = get_multi_char ( this_tok->pp_data.buff, &ch ) ;

	    if ( ch == CHAR_SIMPLE ) qo = ( character ) u ;

	    if ( qo == char_quote || qo == char_single_quote ) {

		/* Start of string */

		int t ;

		int escaped = 0 ;

		PPTOKEN *next_tok = this_tok->next ;

		PPTOKEN *ptr_tok = next_tok ;

		while ( ptr_tok != NULL ) {

		    t = ptr_tok->tok ;

		    if ( t == lex_macro_Harg ) {

			HASHID nm = ptr_tok->pp_data.par.hash ;

			ERROR err = ERR_cpp_stringize_old ( nm, macro ) ;

			report ( preproc_loc, err ) ;

		    }

		    if ( escaped ) {

			escaped = 0 ;

		    } else if ( t == lex_unknown ) {

			character qc = char_question ;

			u = get_multi_char ( ptr_tok->pp_data.buff, &ch ) ;

			if ( ch == CHAR_SIMPLE ) qc = ( character ) u ;

			if ( qc == qo ) break ;

			if ( qc == char_backslash ) escaped = 1 ;

		    }

		    ptr_tok = ptr_tok->next ;

		}

		if ( act ) {

		    if ( ptr_tok == NULL ) {

			/* No closing quote */

			report ( crt_loc, ERR_cpp_stringize_bad ( macro ) ) ;

			this_tok->next = NULL ;

		    } else {

			ptr_tok->tok = lex_ignore_token ;

			this_tok->next = ptr_tok->next ;

			ptr_tok->next = NULL ;

		    }

		    /* Form the string */

		    if ( !quote_tok_list ( next_tok, 0, ( int ) qo ) ) {

			report ( crt_loc, ERR_cpp_stringize_bad ( macro ) ) ;

		    }

		    t = ( qo == char_quote ? lex_string_Hlit : lex_char_Hlit ) ;

		    this_tok->tok = t ;

		    token_parts ( t, this_tok ) ;

		    free_tok_list ( next_tok ) ;

		    /* Check for wide strings */

		    if ( last_tok->tok == lex_identifier ) {

			string s ;

			HASHID nm = last_tok->pp_data.id.hash ;

			s = DEREF_string ( hashid_name_etc_text ( nm ) ) ;

			if ( s [0] == char_L && s [1] == 0 ) {

			    if ( t == lex_string_Hlit ) {

				t = lex_wstring_Hlit ;

			    } else {

				t = lex_wchar_Hlit ;

			    }

			    copy_pptok ( last_tok, t, this_tok ) ;

			    last_tok->next = this_tok->next ;

			    free_pptok ( this_tok ) ;

			    this_tok = last_tok ;

			}

		    }

		}

	    }

	}

	last_tok = this_tok ;

	this_tok = this_tok->next ;

    }

    return ( defn ) ;

}

/*

    HANDLE TOKEN CONCATENATION

    This routine handles any ## operators in the definition defn of the

    given macro.  Note that any initial or terminal ## operators have

    already been reported.

*/

static PPTOKEN *process_concat

    PROTO_N ( ( defn, macro ) )

    PROTO_T ( PPTOKEN *defn X HASHID macro )

{

    PPTOKEN *this_tok ;

    while ( defn && defn->tok == lex_hash_Hhash_Hop ) {

	/* Check for initial ## */

	this_tok = defn ;

	defn = defn->next ;

	free_pptok ( this_tok ) ;

    }

    this_tok = defn ;

    while ( this_tok != NULL ) {

	PPTOKEN *next_tok = this_tok->next ;

	if ( next_tok == NULL ) break ;

	if ( next_tok->tok == lex_hash_Hhash_Hop ) {

	    /* Delete the ## */

	    this_tok->next = next_tok->next ;

	    free_pptok ( next_tok ) ;

	    /* Check for terminal ## */

	    if ( this_tok->next == NULL ) break ;

	    /* Do the token concatenation */

	    if ( concat_pptoks ( this_tok, this_tok->next ) ) {

		/* Delete the second argument if successful */

		next_tok = this_tok->next ;

		this_tok->next = next_tok->next ;

		free_pptok ( next_tok ) ;

	    } else {

		report ( crt_loc, ERR_cpp_concat_bad ( macro ) ) ;

	    }

	    /* Now reprocess this_tok */

	} else {

	    this_tok = next_tok ;

	}

    }

    return ( defn ) ;

}

/*

    MAXIMUM NUMBER OF MACRO PARAMETERS

    This macro defines the maximum number of macro parameters which

    expand_macro can handle without having to allocate temporary space

    to hold them.  With allocation the number of parameters is unlimited.

*/

#define MAX_MACRO_PARAMS	256

/*

    EXPAND A MACRO DEFINITION

    This routine expands the macro given by the hash table entry macro.

    The argument locs gives a list of locations where macro arguments can

    be read from.  locs will never be NULL.  The argument complete is true

    to indicate that this is a complete macro expansion, and that any

    argument errors should be reported.  If locs contains file_loc then

    complete will always be true.  When reading from file_loc we always

    set in_preproc_dir to 2 to make read_token return lex_eof at the end

    of each file, rather than automatically reverting to the including

    file, and to cause it to ignore any preprocessing directives.

    Note that the entry for the macro in the hash table is marked during

    expansion to prevent recursive expansions.  Several points concerning

    macro expansion are undefined; in this implementation:

	1.  Firstly, # operators are evaluated from left to right;

	2.  Secondly, ## operators are evaluated from left to right;

	3.  If a ## b is not a valid preprocessing token then it is

	    resolved to a b;

	4.  A # operator in a function-like macro which is not followed

	    by a macro argument is ignored (it is left as # in object-like

	    macros of course);

	5.  A ## operator at the start or end of a macro is ignored;

	6.  Any preprocessing directives in the macro arguments are treated

	    as normal sequences of preprocessing tokens.

    A further undefined area concerns the ban on recursive macro expansions.

    This is extended from the macro definition itself to any extra tokens

    which are read during the expansion of the macro definition.  For

    example, in:

		    #define f( a )	a * g

		    #define g( a )	f ( a )

		    f ( 2 ) ( 9 )

    the result is '2 * f ( 9 )', rather than '2 * 9 * g'.

*/

PPTOKEN *expand_macro

    PROTO_N ( ( macro, locs, complete ) )

    PROTO_T ( HASHID macro X TOKEN_LOC *locs X int complete )

{

    LOCATION loc ;

    int state = 0 ;

    PPTOKEN *defn ;

    unsigned long sp = 0 ;

    unsigned no_pars = 0 ;

    int have_unknown = 0 ;

    int have_hash_hash = 0 ;

    unsigned long ws = crt_spaces ;

    PPTOKEN dummy_tok, *this_tok = &dummy_tok ;

    PPTOKEN *arg_array_base [ MAX_MACRO_PARAMS + 1 ] ;

    PPTOKEN **arg_array = arg_array_base ;

    /* Get the macro identifier */

    IDENTIFIER id = DEREF_id ( hashid_id ( macro ) ) ;

    unsigned tag = TAG_id ( id ) ;