asmain.c

<B>Digital的Unix操作系统VAX 4.2源码</B>

#ifndef lint
static	char	*sccsid = "@(#)asmain.c	4.1	(ULTRIX)	7/3/90";
#endif not lint

/************************************************************************
 *									*
 *			Copyright (c) 1985 by				*
 *		Digital Equipment Corporation, Maynard, MA		*
 *			All rights reserved.				*
 *									*
 *   This software is furnished under a license and may be used and	*
 *   copied  only  in accordance with the terms of such license and	*
 *   with the  inclusion  of  the  above  copyright  notice.   This	*
 *   software  or  any  other copies thereof may not be provided or	*
 *   otherwise made available to any other person.  No title to and	*
 *   ownership of the software is hereby transferred.			*
 *									*
 *   This software is  derived  from  software  received  from  the	*
 *   University    of   California,   Berkeley,   and   from   Bell	*
 *   Laboratories.  Use, duplication, or disclosure is  subject  to	*
 *   restrictions  under  license  agreements  with  University  of	*
 *   California and with AT&T.						*
 *									*
 *   The information in this software is subject to change  without	*
 *   notice  and should not be construed as a commitment by Digital	*
 *   Equipment Corporation.						*
 *									*
 *   Digital assumes no responsibility for the use  or  reliability	*
 *   of its software on equipment which is not supplied by Digital.	*
 *									*
 ************************************************************************/

/************************************************************************
 *
 *			Modification History
 *
 *	David L Ballenger, 25-Mar-1985
 * 003-	Remove error messages for G and H float opcodes, since the
 *	emulation code will be provided on all ULTRIX-32/32m systems.
 *
 *	Stephen Reilly, 14-May-84
 * 002- Base the removal of the g and g float opcodes on the symbol
 *	GFLOAT
 *
 *	Stephen Reilly, 3-Apr-84
 * 001- Remove error messages for g and h float opcodes.
 *
 ***********************************************************************/

/*
 *	Copyright (c) 1982 Regents of the University of California
 */

#include <stdio.h>
#include <ctype.h>
#include <signal.h>

#include "as.h"
#include "assyms.h"
#include "asscan.h"
#include "asexpr.h"

#include <sys/stat.h>

#define	unix_lang_name "VAX/UNIX Assembler V6/30/83 4.13"
/*
 *	variables to manage reading the assembly source files
 */
char	*dotsname;	/*the current file name; managed by the parser*/
int	lineno;		/*current line number; managed by the parser*/
char	**innames;	/*names of the files being assembled*/
int	ninfiles;	/*how many interesting files there are*/
/*
 *	Flags settable from the argv process argument list
 */
int	silent = 0;	/*don't complain about any errors*/
int	savelabels = 0;	/*write the labels to the a.out file*/
int 	d124 = 4;	/*default allocate 4 bytes for unknown pointers*/
int	anyerrs = 0;	/*no errors yet*/
int	anywarnings=0;	/*no warnings yet*/
int	orgwarn = 0;	/*Bad origins*/
int	passno = 1;	/* current pass*/
int	jxxxJUMP = 0;	/* in jxxxes that branch too far, use jmp instead of brw */
int	readonlydata = 0;	/* initialzed data -> text space */

int	nGHnumbers = 0;		/* GH numbers used */
int	nGHopcodes = 0;		/* GH opcodes used */
int	nnewopcodes = 0;	/* new opcodes used */

#ifdef DEBUG
int 	debug = 0;
int	toktrace = 0;
#endif

int	useVM =	0;

char	*endcore;	/*where to get more symbol space*/

/*
 *	Managers of the a.out file.
 */
struct	exec	hdr;
#define	MAGIC	0407
u_long	tsize;		/* total text size */
u_long	dsize;		/* total data size */
u_long	datbase;	/* base of the data segment */
u_long	trsize;		/* total text relocation size */
u_long	drsize;		/* total data relocation size */

/*
 *	Information about the current segment is accumulated in
 *	usedot; the most important information stored is the
 *	accumulated size of each of the text and data segments
 *
 *	dotp points to the correct usedot expression for the current segment
 */
struct	exp	usedot[NLOC+NLOC];	/* info about all segments */
struct	exp	*dotp;			/* data/text location pointer */
/*
 *	The inter pass temporary token file is opened and closed by stdio, but
 *	is written to using direct read/write, as the temporary file
 *	is composed of buffers exactly BUFSIZ long.
 */
FILE	*tokfile;			/* interpass communication file */
char	tokfilename[TNAMESIZE];
/*
 *	The string file is the string table
 *	cat'ed to the end of the built up a.out file
 */
FILE	*strfile;			/* interpass string file */
char	strfilename[TNAMESIZE];
int	strfilepos = 0;			/* position within the string file */
/*
 *	a.out is created during the second pass.
 *	It is opened by stdio, but is filled with the parallel
 *	block I/O library
 */
char	*outfile = "a.out";
FILE	*a_out_file;			
off_t	a_out_off;			/* cumulative offsets for segments */
/*
 *	The logical files containing the assembled data for each of
 *	the text and data segments are
 *	managed by the parallel block I/O library.
 *	a.out is logically opened in many places at once to
 *	receive the assembled data from the various segments as
 *	it all trickles in, but is physically opened only once
 *	to minimize file overhead.
 */
BFILE	*usefile[NLOC+NLOC];		/* text/data files */
BFILE	*txtfil;			/* current text/data file */
/*
 *	Relocation information is accumulated seperately for each
 *	segment.  This is required by the old loader (from BTL),
 *	but not by the new loader (Bill Joy).  
 *
 *	However, the size of the relocation information can not be computed
 *	during or after the 1st pass because the ''absoluteness' of values
 *	is unknown until all locally declared symbols have been seen.
 *	Thus, the size of the relocation information is only
 *	known after the second pass is finished.
 *	This obviates the use of the block I/O
 *	library, which requires knowing the exact offsets in a.out.
 *
 *	So, we save the relocation information internally (we don't
 *	go to internal files to minimize overhead).
 *
 *	Empirically, we studied 259 files composing the system,
 *	two compilers and a compiler generator: (all of which have
 *	fairly large source files)
 *	
 *	Number of files = 259
 *		Number of non zero text reloc files: 233
 *		Number of non zero data reloc files: 53
 *	Average text relocation = 889
 *	Average data relocation = 346
 *	Number of files > BUFSIZ text relocation = 71
 *	Number of files > BUFSIZ data relocation = 6
 *	
 *	For compiled C code, there is usually one text segment and two
 *	data segments; we see that allocating our own buffers and
 *	doing our internal handling of relocation information will,
 *	on the average, not use more memory than taken up by the buffers
 *	allocated for doing file I/O in parallel to a number of file.
 *	
 *	If we are assembling with the -V option, we
 *	use the left over token buffers from the 2nd pass,
 *	otherwise, we create our own.
 *
 *	When the 2nd pass is complete, closeoutrel flushes the token
 *	buffers out to a BFILE.
 *
 *	The internals to relbufdesc are known only in assyms.c
 *
 *	outrel constructs the relocation information.
 *	closeoutrel flushes the relocation information to relfil.
 */
struct	relbufdesc	*rusefile[NLOC+NLOC];	
struct	relbufdesc 	*relfil;	/* un concatnated relocation info */
BFILE	*relocfile;			/* concatnated relocation info */
/*
 *	Once the relocation information has been written,
 *	we can write out the symbol table using the Block I/O
 *	mechanisms, as we once again know the offsets into
 *	the a.out file.
 *
 *	We use relfil to output the symbol table information.
 */
char	*tmpdirprefix = "/tmp/";
int delexit();

main(argc, argv)
	int	argc;
	char 	**argv;
{
	char	*sbrk();

	tokfilename[0] = 0;
	strfilename[0] = 0;
	endcore = sbrk(0);

	argprocess(argc, argv);		/* process argument lists */
	if (anyerrs) exit(1);

	initialize();
	zeroorigins();			/* set origins to zero */
	zerolocals();			/* fix local label counters */

	i_pass1();			/* open temp files, etc */
	pass1();			/* first pass through .s files */
	testlocals();			/* check for undefined locals */
	if (anyerrs) delexit();

	pass1_5();			/* resolve jxxx */
	if (anyerrs) delexit();

	open_a_out();			/* open a.out */
	roundsegments();		/* round segments to FW */
	build_hdr();			/* build initial header, and output */
	
	i_pass2();			/* reopen temporary file, etc */
	pass2();			/* second pass through the virtual .s */
	if (anyerrs) delexit();

	fillsegments();			/* fill segments with 0 to FW */
	reloc_syms();			/* dump relocation and symbol table */

	delete();			/* remove tmp file */
	bflush();			/* close off block I/O view of a.out */
	fix_a_out();			/* add in text and data reloc counts */

	if (anyerrs == 0 && orgwarn)
		yyerror("Caution: absolute origins.\n");

	if (nnewopcodes) {
		yywarning("Caution: New Opcodes");
		yywarning("These are not defined for all implementations of the VAX architecture.\n");
	}

	exit(anyerrs != 0);
}

argprocess(argc, argv)
	int	argc;
	char	*argv[];
{
	register	char	*cp;

	ninfiles = 0;
	silent = 0;
#ifdef DEBUG
	debug = 0;
#endif
	innames = (char **)ClearCalloc(argc+1, sizeof (innames[0]));
	dotsname = "<argv error>";
	while (argc > 1) {
		if (argv[1][0] != '-')
			innames[ninfiles++] = argv[1];
		else {
			cp = argv[1] + 1;
			/*
			 *	We can throw away single minus signs, so
			 *	that make scripts for the PDP 11 assembler work
			 *	on this assembler too
			 */
			while (*cp){	
				switch(*cp++){
				 default:
					yyerror("Unknown flag: %c", *--cp);
					cp++;
					break;
				 case 'v':
					selfwhat(stdout);
					exit(1);
				 case 'd':
					d124 = *cp++ - '0';
					if ( (d124 != 1) && (d124 != 2) && 
					     (d124 != 4)){
						yyerror("-d[124] only");
						exit(1);
					}
					break;
				 case 'o':
					if (argc < 3){
						yyerror("-o what???");
						exit(1);
					}
					outfile = argv[2];
				   bumpone:
					argc -= 2;
					argv += 2;
					goto nextarg;

				 case 't':
					if (argc < 3){
						yyerror("-t what???");
						exit(1);
					}
					tmpdirprefix = argv[2];
					goto bumpone;
				 
				 case 'V':
					useVM = 1;
					break;
				 case 'W':
					silent = 1;
					break;
				 case 'L':
					savelabels = 1;
					break;
				 case 'J':
					jxxxJUMP = 1;
					break;
#ifdef DEBUG
				 case 'D':
					debug = 1;
					break;
				 case 'T':
					toktrace = 1;
					break;
#endif
				 case 'R':
					readonlydata = 1;
					break;
				}	/*end of the switch*/
			}	/*end of pulling out all arguments*/
		}	/*end of a flag argument*/
		--argc; ++argv;
	   nextarg:;
	}
	/* innames[ninfiles] = 0; */
}
/*
 *	poke through the data space and find all sccs identifiers.
 *	We assume:
 *	a) that extern char **environ; is the first thing in the bss
 *	segment (true, if one is using the new version of cmgt.crt0.c)
 *	b) that the sccsid's have not been put into text space.
 */
selfwhat(place)
	FILE	*place;
{
	extern	char **environ;
	register	char	*ub;
	register	char *cp;
	register	char	*pat;
	char	*sbrk();