marshal.c

这是一个同样来自贝尔实验室的和UNIX有着渊源的操作系统, 其简洁的设计和实现易于我们学习和理解

#include "common.h"
#include <ctype.h>

typedef struct Attach Attach;
typedef struct Alias Alias;
typedef struct Addr Addr;
typedef struct Ctype Ctype;

struct Attach {
	Attach	*next;
	char	*path;
	char	*type;
	int	ainline;
	Ctype	*ctype;
};

struct Alias
{
	Alias	*next;
	int	n;
	Addr	*addr;
};

struct Addr
{
	Addr	*next;
	char	*v;
};

enum {
	Hfrom,
	Hto,
	Hcc,
	Hbcc,
	Hsender,
	Hreplyto,
	Hinreplyto,
	Hdate,
	Hsubject,
	Hmime,
	Hpriority,
	Hmsgid,
	Hcontent,
	Hx,
	Hprecedence,
	Nhdr,
};

enum {
	PGPsign = 1,
	PGPencrypt = 2,
};

char *hdrs[Nhdr] = {
[Hfrom]		"from:",
[Hto]		"to:",
[Hcc]		"cc:",
[Hbcc]		"bcc:",
[Hreplyto]	"reply-to:",
[Hinreplyto]	"in-reply-to:",
[Hsender]	"sender:",
[Hdate]		"date:",
[Hsubject]	"subject:",
[Hpriority]	"priority:",
[Hmsgid]	"message-id:",
[Hmime]		"mime-",
[Hcontent]	"content-",
[Hx]		"x-",
[Hprecedence]	"precedence",
};

struct Ctype {
	char	*type;
	char 	*ext;
	int	display;
};

Ctype ctype[] = {
	{ "text/plain",			"txt",	1,	},
	{ "text/html",			"html",	1,	},
	{ "text/html",			"htm",	1,	},
	{ "text/tab-separated-values",	"tsv",	1,	},
	{ "text/richtext",		"rtx",	1,	},
	{ "message/rfc822",		"txt",	1,	},
	{ "", 				0,	0,	},
};

Ctype *mimetypes;

int pid = -1;
int pgppid = -1;

Attach*	mkattach(char*, char*, int);
int	readheaders(Biobuf*, int*, String**, Addr**, int);
void	body(Biobuf*, Biobuf*, int);
char*	mkboundary(void);
int	printdate(Biobuf*);
int	printfrom(Biobuf*);
int	printto(Biobuf*, Addr*);
int	printcc(Biobuf*, Addr*);
int	printsubject(Biobuf*, char*);
int	printinreplyto(Biobuf*, char*);
int	sendmail(Addr*, Addr*, int*, char*);
void	attachment(Attach*, Biobuf*);
int	cistrncmp(char*, char*, int);
int	cistrcmp(char*, char*);
char*	waitforsubprocs(void);
int	enc64(char*, int, uchar*, int);
Addr*	expand(int, char**);
Alias*	readaliases(void);
Addr*	expandline(String**, Addr*);
void	Bdrain(Biobuf*);
void	freeaddr(Addr *);
int	pgpopts(char*);
int	pgpfilter(int*, int, int);
void	readmimetypes(void);
char*	estrdup(char*);
void*	emalloc(int);
void*	erealloc(void*, int);
void	freeaddr(Addr*);
void	freeaddrs(Addr*);
void	freealias(Alias*);
void	freealiases(Alias*);
int	doublequote(Fmt*);
int	rfc2047fmt(Fmt*);
char*	mksubject(char*);

int rflag, lbflag, xflag, holding, nflag, Fflag, eightflag, dflag;
int pgpflag = 0;
char *user;
char *login;
Alias *aliases;
int rfc822syntaxerror;
char lastchar;
char *replymsg;

enum
{
	Ok = 0,
	Nomessage = 1,
	Nobody = 2,
	Error = -1,
};

#pragma varargck	type	"Z"	char*
#pragma varargck	type	"U"	char*

void
usage(void)
{
	fprint(2, "usage: %s [-Fr#xn] [-s subject] [-c ccrecipient] [-t type] [-aA attachment] [-p[es]] [-R replymsg] -8 | recipient-list\n",
		argv0);
	exits("usage");
}

void
fatal(char *fmt, ...)
{
	char buf[1024];
	va_list arg;

	if(pid >= 0)
		postnote(PNPROC, pid, "die");
	if(pgppid >= 0)
		postnote(PNPROC, pgppid, "die");

	va_start(arg, fmt);
	vseprint(buf, buf+sizeof(buf), fmt, arg);
	va_end(arg);
	fprint(2, "%s: %s\n", argv0, buf);
	holdoff(holding);
	exits(buf);
}

void
main(int argc, char **argv)
{
	Attach *first, **l, *a;
	char *subject, *type, *boundary;
	int flags, fd;
	Biobuf in, out, *b;
	Addr *to;
	Addr *cc;
	String *file, *hdrstring;
	int noinput, headersrv;
	int ccargc;
	char *ccargv[32];

	noinput = 0;
	subject = nil;
	first = nil;
	l = &first;
	type = nil;
	hdrstring = nil;
	ccargc = 0;

	quotefmtinstall();
	fmtinstall('Z', doublequote);
	fmtinstall('U', rfc2047fmt);

	ARGBEGIN{
	case 't':
		type = ARGF();
		if(type == nil)
			usage();
		break;
	case 'a':
		flags = 0;
		goto aflag;
	case 'A':
		flags = 1;
	aflag:
		a = mkattach(ARGF(), type, flags);
		if(a == nil)
			exits("bad args");
		type = nil;
		*l = a;
		l = &a->next;
		break;
	case 'C':
		if(ccargc >= nelem(ccargv)-1)
			sysfatal("too many cc's");
		ccargv[ccargc] = ARGF();
		if(ccargv[ccargc] == nil)
			usage();
		ccargc++;
		break;
	case 'R':
		replymsg = ARGF();
		break;
	case 's':
		subject = ARGF();
		break;
	case 'F':
		Fflag = 1;		// file message
		break;
	case 'r':
		rflag = 1;		// for sendmail
		break;
	case 'd':
		dflag = 1;		// for sendmail
		break;
	case '#':
		lbflag = 1;		// for sendmail
		break;
	case 'x':
		xflag = 1;		// for sendmail
		break;
	case 'n':			// no standard input
		nflag = 1;
		break;
	case '8':			// read recipients from rfc822 header
		eightflag = 1;
		break;
	case 'p':			// pgp flag: encrypt, sign, or both
		if(pgpopts(ARGF()) < 0)
			sysfatal("bad pgp options");
		break;
	default:
		usage();
		break;
	}ARGEND;

	login = getlog();
	user = getenv("upasname");
	if(user == nil || *user == 0)
		user = login;
	if(user == nil || *user == 0)
		sysfatal("can't read user name");

	if(Binit(&in, 0, OREAD) < 0)
		sysfatal("can't Binit 0: %r");

	if(nflag && eightflag)
		sysfatal("can't use both -n and -8");
	if(eightflag && argc >= 1)
		usage();
	else if(!eightflag && argc < 1)
		usage();

	aliases = readaliases();
	if(!eightflag){
		to = expand(argc, argv);
		cc = expand(ccargc, ccargv);
	} else {
		to = nil;
		cc = nil;
	}

	flags = 0;
	headersrv = Nomessage;
	if(!nflag && !xflag && !lbflag &&!dflag) {
		// pass through headers, keeping track of which we've seen,
		// perhaps building to list.
		holding = holdon();
		headersrv = readheaders(&in, &flags, &hdrstring, eightflag ? &to : nil, 1);
		if(rfc822syntaxerror){
			Bdrain(&in);
			fatal("rfc822 syntax error, message not sent");
		}
		if(to == nil){
			Bdrain(&in);
			fatal("no addresses found, message not sent");
		}

		switch(headersrv){
		case Error:		// error
			fatal("reading");
			break;
		case Nomessage:		// no message, just exit mimicking old behavior
			noinput = 1;
			if(first == nil)
				exits(0);
			break;
		}
	}

	fd = sendmail(to, cc, &pid, Fflag ? argv[0] : nil);
	if(fd < 0)
		sysfatal("execing sendmail: %r\n:");
	if(xflag || lbflag || dflag){
		close(fd);
		exits(waitforsubprocs());
	}
	
	if(Binit(&out, fd, OWRITE) < 0)
		fatal("can't Binit 1: %r");

	if(!nflag){
		if(Bwrite(&out, s_to_c(hdrstring), s_len(hdrstring)) != s_len(hdrstring))
			fatal("write error");
		s_free(hdrstring);
		hdrstring = nil;

		// read user's standard headers
		file = s_new();
		mboxpath("headers", user, file, 0);
		b = Bopen(s_to_c(file), OREAD);
		if(b != nil){
			switch(readheaders(b, &flags, &hdrstring, nil, 0)){
			case Error:	// error
				fatal("reading");
			}
			Bterm(b);
			if(Bwrite(&out, s_to_c(hdrstring), s_len(hdrstring)) != s_len(hdrstring))
				fatal("write error");
			s_free(hdrstring);
			hdrstring = nil;
		}
	}

	// add any headers we need
	if((flags & (1<<Hdate)) == 0)
		if(printdate(&out) < 0)
			fatal("writing");
	if((flags & (1<<Hfrom)) == 0)
		if(printfrom(&out) < 0)
			fatal("writing");
	if((flags & (1<<Hto)) == 0)
		if(printto(&out, to) < 0)
			fatal("writing");
	if((flags & (1<<Hcc)) == 0)
		if(printcc(&out, cc) < 0)
			fatal("writing");
	if((flags & (1<<Hsubject)) == 0 && subject != nil)
		if(printsubject(&out, subject) < 0)
			fatal("writing");
	if(replymsg != nil)
		if(printinreplyto(&out, replymsg) < 0)
			fatal("writing");
	Bprint(&out, "MIME-Version: 1.0\n");

	if(pgpflag){	// interpose pgp process between us and sendmail to handle body
		Bflush(&out);
		Bterm(&out);
		fd = pgpfilter(&pgppid, fd, pgpflag);
		if(Binit(&out, fd, OWRITE) < 0)
			fatal("can't Binit 1: %r");
	}

	// if attachments, stick in multipart headers
	boundary = nil;
	if(first != nil){
		boundary = mkboundary();
		Bprint(&out, "Content-Type: multipart/mixed;\n");
		Bprint(&out, "\tboundary=\"%s\"\n\n", boundary);
		Bprint(&out, "This is a multi-part message in MIME format.\n");
		Bprint(&out, "--%s\n", boundary);
		Bprint(&out, "Content-Disposition: inline\n");
	}

	if(!nflag){
		if(!noinput && headersrv == Ok){
			body(&in, &out, 1);
		}
	} else
		Bprint(&out, "\n");
	holdoff(holding);

	Bflush(&out);
	for(a = first; a != nil; a = a->next){
		if(lastchar != '\n')
			Bprint(&out, "\n");
		Bprint(&out, "--%s\n", boundary);
		attachment(a, &out);
	}

	if(first != nil){
		if(lastchar != '\n')
			Bprint(&out, "\n");
		Bprint(&out, "--%s--\n", boundary);
	}

	Bterm(&out);
	close(fd);
	exits(waitforsubprocs());
}

// evaluate pgp option string
int
pgpopts(char *s)
{
	if(s == nil || s[0] == '\0')
		return -1;
	while(*s){
		switch(*s++){
		case 's':  case 'S':
			pgpflag |= PGPsign;
			break;
		case 'e': case 'E':
			pgpflag |= PGPencrypt;
			break;
		default:
			return -1;
		}
	}
	return 0;
}

// read headers from stdin into a String, expanding local aliases,
// keep track of which headers are there, which addresses we have
// remove Bcc: line.
int
readheaders(Biobuf *in, int *fp, String **sp, Addr **top, int strict)
{
	Addr *to;
	String *s, *sline;
	char *p;
	int i, seen, hdrtype;

	s = s_new();
	sline = nil;
	to = nil;
	hdrtype = -1;
	seen = 0;
	for(;;) {
		if((p = Brdline(in, '\n')) != nil) {
			seen = 1;
			p[Blinelen(in)-1] = 0;

			// coalesce multiline headers
			if((*p == ' ' || *p == '\t') && sline){
				s_append(sline, "\n");
				s_append(sline, p);
				p[Blinelen(in)-1] = '\n';
				continue;
			}
		}

		// process the current header, it's all been read
		if(sline) {
			assert(hdrtype != -1);
			if(top){
				switch(hdrtype){
				case Hto:
				case Hcc:
				case Hbcc:
					to = expandline(&sline, to);
					break;
				}
			}
			if(hdrtype == Hsubject){
				s_append(s, mksubject(s_to_c(sline)));
				s_append(s, "\n");
			}else if(top==nil || hdrtype!=Hbcc){
				s_append(s, s_to_c(sline));
				s_append(s, "\n");
			}
			s_free(sline);
			sline = nil;
		}

		if(p == nil)
			break;

		// if no :, it's not a header, seek back and break
		if(strchr(p, ':') == nil){
			p[Blinelen(in)-1] = '\n';
			Bseek(in, -Blinelen(in), 1);
			break;
		}

		sline = s_copy(p);

		// classify the header.  If we don't recognize it, break.  This is
		// to take care of user's that start messages with lines that contain
		// ':'s but that aren't headers.  This is a bit hokey.  Since I decided
		// to let users type headers, I need some way to distinguish.  Therefore,
		// marshal tries to know all likely headers and will indeed screw up if
		// the user types an unlikely one. -- presotto
		hdrtype = -1;
		for(i = 0; i < nelem(hdrs); i++){
			if(cistrncmp(hdrs[i], p, strlen(hdrs[i])) == 0){
				*fp |= 1<<i;
				hdrtype = i;
				break;
			}
		}
		if(strict){
			if(hdrtype == -1){
				p[Blinelen(in)-1] = '\n';
				Bseek(in, -Blinelen(in), 1);
				break;
			}
		} else
			hdrtype = 0;
		p[Blinelen(in)-1] = '\n';
	}

	*sp = s;
	if(top)
		*top = to;

	if(seen == 0){
		if(Blinelen(in) == 0)
			return Nomessage;
		else
			return Ok;
	}
	if(p == nil)
		return Nobody;
	return Ok;
}

// pass the body to sendmail, make sure body starts and ends with a newline
void
body(Biobuf *in, Biobuf *out, int docontenttype)
{
	char *buf, *p;
	int i, n, len;

	n = 0;
	len = 16*1024;
	buf = emalloc(len);

	// first char must be newline
	i = Bgetc(in);
	if(i > 0){
		if(i != '\n')
			buf[n++] = '\n';
		buf[n++] = i;
	} else {
		buf[n++] = '\n';
	}

	// read into memory
	if(docontenttype){
		while(docontenttype){
			if(n == len){
				len += len>>2;
				buf = realloc(buf, len);
				if(buf == nil)
					sysfatal("%r");
			}
			p = buf+n;
			i = Bread(in, p, len - n);
			if(i < 0)
				fatal("input error2");
			if(i == 0)
				break;
			n += i;
			for(; i > 0; i--)
				if((*p++ & 0x80) && docontenttype){
					Bprint(out, "Content-Type: text/plain; charset=\"UTF-8\"\n");
					Bprint(out, "Content-Transfer-Encoding: 8bit\n");
					docontenttype = 0;
					break;
				}
		}
		if(docontenttype){
			Bprint(out, "Content-Type: text/plain; charset=\"US-ASCII\"\n");
			Bprint(out, "Content-Transfer-Encoding: 7bit\n");
		}
	}

	// write what we already read
	if(Bwrite(out, buf, n) < 0)
		fatal("output error");
	if(n > 0)
		lastchar = buf[n-1];
	else
		lastchar = '\n';


	// pass the rest
	for(;;){
		n = Bread(in, buf, len);
		if(n < 0)
			fatal("input error2");
		if(n == 0)
			break;
		if(Bwrite(out, buf, n) < 0)
			fatal("output error");
		lastchar = buf[n-1];
	}
}

// pass the body to sendmail encoding with base64
//
//  the size of buf is very important to enc64.  Anything other than
//  a multiple of 3 will cause enc64 to output a termination sequence.
//  To ensure that a full buf corresponds to a multiple of complete lines,
//  we make buf a multiple of 3*18 since that's how many enc64 sticks on
//  a single line.  This avoids short lines in the output which is pleasing
//  but not necessary.
//
void
body64(Biobuf *in, Biobuf *out)
{
	uchar buf[3*18*54];
	char obuf[3*18*54*2];
	int m, n;

	Bprint(out, "\n");
	for(;;){
		n = Bread(in, buf, sizeof(buf));
		if(n < 0)