20090430-1331

6 files changed, 208 insertions, 35 deletions

diff --git a/appl/lib/deflate.b b/appl/lib/deflate.b
index ccfeafbe..500b3797 100644
--- a/appl/lib/deflate.b
+++ b/appl/lib/deflate.b
@@ -42,6 +42,7 @@ GZCRCPOLY:	con int 16redb88320;
 GZOSINFERNO:	con GZOSUNIX;
 
 
+Hnone, Hgzip, Hzlib: con iota;  # LZstate.headers
 LZstate: adt
 {
 	hist:		array of byte;		# [HistSize];
@@ -56,11 +57,13 @@ LZstate: adt
 	prevoff:	int;
 	maxchars:	int;			# compressor tuning
 	maxdefer:	int;
+	level:		int;
 
-	crctab: array of int;
+	crctab: array of int;			# for gzip trailer
 	crc:		int;
 	tot:		int;
-	headers:	int;
+	sum:		big;			# for zlib trailer
+	headers:	int;			# which header to print, if any
 
 	outbuf:		array of byte;		# current output buffer;
 	out:		int;			# current position in the output buffer
@@ -287,22 +290,27 @@ start(param: string): chan of ref Rq
 {
 	# param contains flags:
 	# [0-9] - compression level
+	# h gzip header/trailer
+	# z zlib header/trailer
 	# v verbose
 	# d debug
 	lz := ref LZstate;
-	level := 6;
+	lz.level = 6;
 	lz.verbose = lz.debug = 0;
-	lz.headers = 0;
+	lz.headers = Hnone;
 	lz.crc = lz.tot = 0;
+	lz.sum = big 1;
 	# XXX could also put filename and modification time in param
 	for (i := 0; i < len param; i++) {
 		case param[i] {
 		'0' to '9' =>
-			level = param[i] - '0';
+			lz.level = param[i] - '0';
 		'v' =>
 			lz.verbose = 1;
 		'h' =>
-			lz.headers = 1;
+			lz.headers = Hgzip;
+		'z' =>
+			lz.headers = Hzlib;
 		'd' =>
 			lz.debug = 1;
 		}
@@ -345,17 +353,17 @@ start(param: string): chan of ref Rq
 	lz.dot = lz.me;
 	lz.bits = 0;
 	lz.nbits = 0;
-	if(level < 5) {
+	if(lz.level < 5) {
 		lz.maxchars = 1;
 		lz.maxdefer = 0;
-	} else if(level == 9) {
+	} else if(lz.level == 9) {
 		lz.maxchars = 4000;
 		lz.maxdefer = MaxMatch;
 	} else {
 		lz.maxchars = 200;
 		lz.maxdefer = MaxMatch / 4;
 	}
-	if (lz.headers)
+	if (lz.headers == Hgzip)
 		lz.crctab = mkcrctab(GZCRCPOLY);
 	lz.c = chan of ref Rq;
 	lz.rc = chan of int;
@@ -383,8 +391,7 @@ deflate(lz: ref LZstate)
 {
 	lz.c <-= ref Rq.Start(sys->pctl(0, nil));
 
-	if (lz.headers)
-		header(lz);
+	header(lz);
 	buf := array[DeflateBlock] of byte;
 	out := array[DeflateBlock + DeflateOut] of byte;
 	eof := 0;
@@ -393,8 +400,7 @@ deflate(lz: ref LZstate)
 		nbuf := 0;
 		if (!eof) {
 			(eof, nbuf) = fillbuf(lz, buf);
-			if (lz.headers)	# added by Roman Joel Pacheco
-				inblock(lz, buf[0:nbuf]);
+			inblock(lz, buf[0:nbuf]);
 		}
 		if(eof && nbuf == 0 && nslop == 0) {
 			if(lz.nbits) {
@@ -405,8 +411,7 @@ deflate(lz: ref LZstate)
 					exit;
 				continue;
 			}
-			if (lz.headers)
-				footer(lz);
+			footer(lz);
 			lz.c <-= ref Rq.Finished(nil);
 			exit;
 		}
@@ -538,7 +543,7 @@ deflate(lz: ref LZstate)
 	}
 }
 
-header(lz: ref LZstate)
+headergzip(lz: ref LZstate)
 {
 	buf := array[20] of byte;
 	i := 0;
@@ -569,7 +574,44 @@ header(lz: ref LZstate)
 		exit;
 }
 
-footer(lz: ref LZstate)
+headerzlib(lz: ref LZstate)
+{
+	CIshift:	con 12;
+	CMdeflate:	con 8;
+	CMshift:	con 8;
+	LVshift:	con 6;
+	LVfastest, LVfast, LVnormal, LVbest: con iota;
+
+	level := LVnormal;
+	if(lz.level < 6)
+		level = LVfastest;
+	else if(lz.level >= 9)
+		level = LVbest;
+
+	h := 0;
+	h |= 7<<CIshift; # value is: (log2 of window size)-8
+	h |= CMdeflate<<CMshift;
+	h |= level<<LVshift;
+	h += 31-(h%31);
+
+	buf := array[2] of byte;
+	buf[0] = byte (h>>8);
+	buf[1] = byte (h>>0);
+
+	lz.c <-= ref Rq.Result(buf, lz.rc);
+	if (<-lz.rc == -1)
+		exit;
+}
+
+header(lz: ref LZstate)
+{
+	case lz.headers {
+	Hgzip =>	headergzip(lz);
+	Hzlib =>	headerzlib(lz);
+	}
+}
+
+footergzip(lz: ref LZstate)
 {
 	buf := array[8] of byte;
 	i := 0;
@@ -587,6 +629,28 @@ footer(lz: ref LZstate)
 		exit;
 }
 
+footerzlib(lz: ref LZstate)
+{
+        buf := array[4] of byte;
+	i := 0;
+        buf[i++] = byte (lz.sum>>24);
+        buf[i++] = byte (lz.sum>>16);
+        buf[i++] = byte (lz.sum>>8);
+        buf[i++] = byte (lz.sum>>0);
+
+	lz.c <-= ref Rq.Result(buf, lz.rc);
+	if(<-lz.rc == -1)
+		exit;
+}
+
+footer(lz: ref LZstate)
+{
+	case lz.headers {
+	Hgzip =>	footergzip(lz);
+	Hzlib =>	footerzlib(lz);
+	}
+}
+
 lzput(lz: ref LZstate, bits, nbits: int): int
 {
 	bits = (bits << lz.nbits) | lz.bits;
@@ -1349,7 +1413,7 @@ mkcrctab(poly: int): array of int
 	return crctab;
 }
 
-inblock(lz: ref LZstate, buf: array of byte)
+inblockcrc(lz: ref LZstate, buf: array of byte)
 {
 	crc := lz.crc;
 	n := len buf;
@@ -1360,6 +1424,28 @@ inblock(lz: ref LZstate, buf: array of byte)
 	lz.tot += n;
 }
 
+inblockadler(lz: ref LZstate, buf: array of byte)
+{
+	ZLADLERBASE:	con big 65521;
+
+	s1 := lz.sum & big 16rffff;
+	s2 := (lz.sum>>16) & big 16rffff;
+
+	for(i := 0; i < len buf; i++) {
+		s1 = (s1 + big buf[i]) % ZLADLERBASE;
+		s2 = (s2 + s1) % ZLADLERBASE;
+	}
+	lz.sum = (s2<<16) + s1;
+}
+
+inblock(lz: ref LZstate, buf: array of byte)
+{
+	case lz.headers {
+	Hgzip =>	inblockcrc(lz, buf);
+	Hzlib =>	inblockadler(lz, buf);
+	}
+}
+
 fatal(lz: ref LZstate, s: string)
 {
 	lz.c <-= ref Rq.Error(s);
diff --git a/appl/lib/inflate.b b/appl/lib/inflate.b
index 0f4d6ba7..3ee2503a 100644
--- a/appl/lib/inflate.b
+++ b/appl/lib/inflate.b
@@ -135,6 +135,7 @@ offdec: ref DeHuff;
 revtab: array of byte;	# bit reversal for endian swap of huffman codes
 mask: array of int;		# for masking low-order n bits of an int
 
+Hnone, Hgzip, Hzlib: con iota;  # State.headers
 State: adt {
 	ibuf, obuf: array of byte;
 	c: chan of ref Rq;
@@ -145,7 +146,8 @@ State: adt {
 	hist: array of byte;	# history buffer for lempel-ziv backward references
 	usehist: int;		# == 1 if 'hist' is valid
 	crctab: array of int;
-	crc, tot: int;
+	crc, tot: int;		# for gzip trailer
+	sum:	big;		# for zlib trailer
 	
 	reg: int;		# 24-bit shift register
 	nbits: int;		# number of valid bits in reg
@@ -226,9 +228,16 @@ start(params: string): chan of ref Rq
 	s.nbits = 0;
 	s.crc = 0;
 	s.tot = 0;
+	s.sum = big 1;
 	s.hist = array[Blocksize] of byte;
-	s.headers = (params != nil && params[0] == 'h');
-	if (s.headers)
+	s.headers = Hnone;
+	if(params != nil) {
+		if(params[0] == 'h')
+			s.headers = Hgzip;
+		if(params[0] == 'z')
+			s.headers = Hzlib;
+	}
+	if (s.headers == Hgzip)
 		s.crctab = mkcrctab(GZCRCPOLY);
 	spawn inflate(s);
 	return s.c;
@@ -237,8 +246,7 @@ start(params: string): chan of ref Rq
 inflate(s: ref State)
 {
 	s.c <-= ref Rq.Start(sys->pctl(0, nil));
-	if (s.headers)
-		header(s);
+	header(s);
 
 	for(;;) {
 		bfinal := getn(s, 1, 0);
@@ -268,23 +276,21 @@ inflate(s: ref State)
 		}
 		if(bfinal) {
 			if(s.out) {
-				if (s.headers)
-					outblock(s);
+				outblock(s);
 				s.c <- = ref Rq.Result(s.obuf[0:s.out], s.rc);
 				flag := <- s.rc;
 				if (flag == -1)
 					exit;
 			}
 			flushbits(s);
-			if (s.headers)
-				footer(s);
+			footer(s);
 			s.c <-= ref Rq.Finished(s.ibuf[s.in - s.nbits/8:s.ein]);
 			exit;
 		}
 	}
 }
 
-header(s: ref State)
+headergzip(s: ref State)
 {
 	if(byte getb(s) != GZMAGIC1 || byte getb(s) != GZMAGIC2)
 		fatal(s, "not a gzip file");
@@ -335,7 +341,33 @@ header(s: ref State)
 	}
 }
 
-footer(s: ref State)
+headerzlib(s: ref State)
+{
+	Fdict:		con 1<<5;
+	CMshift:	con 8;
+	CMmask:		con (1<<4)-1;
+	CMdeflate:	con 8;
+
+	h := 0;
+	h |= getb(s)<<8;
+	h |= getb(s);
+	if(h % 31 != 0)
+		fatal(s, "invalid zlib header");
+	if(h&Fdict)
+		fatal(s, "preset dictionary not supported");
+	if(((h>>CMshift)&CMmask) != CMdeflate)
+		fatal(s, "zlib compression method not deflate");
+}
+
+header(s: ref State)
+{
+	case s.headers {
+	Hgzip =>	headergzip(s);
+	Hzlib =>	headerzlib(s);
+	}
+}
+
+footergzip(s: ref State)
 {
 	fcrc := getword(s);
 	if(s.crc != fcrc)
@@ -345,6 +377,25 @@ footer(s: ref State)
 		fatal(s, sys->sprint("byte count mismatch: computed %d, expected %d", s.tot, ftot));
 }
 
+footerzlib(s: ref State)
+{
+	sum := big 0;
+	sum = (sum<<8)|big getb(s);
+	sum = (sum<<8)|big getb(s);
+	sum = (sum<<8)|big getb(s);
+	sum = (sum<<8)|big getb(s);
+	if(sum != s.sum)
+		fatal(s, sys->sprint("adler32 mismatch: computed %bux, expected %bux", s.sum, sum));
+}
+
+footer(s: ref State)
+{
+	case s.headers {
+	Hgzip =>	footergzip(s);
+	Hzlib =>	footerzlib(s);
+	}
+}
+
 getword(s: ref State): int
 {
 	n := 0;
@@ -694,8 +745,7 @@ flushbits(s: ref State)
 #
 flushout(s: ref State)
 {
-	if (s.headers)
-		outblock(s);
+	outblock(s);
 	s.c <-= ref Rq.Result(s.obuf[0:s.out], s.rc);
 	flag := <- s.rc;
 	if (flag == -1)
@@ -723,7 +773,7 @@ mkcrctab(poly: int): array of int
 	return crctab;
 }
 
-outblock(s: ref State)
+outblockgzip(s: ref State)
 {
 	buf := s.obuf;
 	n := s.out;
@@ -735,6 +785,31 @@ outblock(s: ref State)
 	s.tot += n;
 }
 
+outblockzlib(s: ref State)
+{
+	ZLADLERBASE:	con big 65521;
+
+	buf := s.obuf;
+	n := s.out;
+
+	s1 := s.sum & big 16rffff;
+	s2 := (s.sum>>16) & big 16rffff;
+
+	for(i := 0; i < n; i++) {
+		s1 = (s1 + big buf[i]) % ZLADLERBASE;
+		s2 = (s2 + s1) % ZLADLERBASE;
+	}
+	s.sum = (s2<<16) + s1;
+}
+
+outblock(s: ref State)
+{
+	case s.headers {
+	Hgzip =>	outblockgzip(s);
+	Hzlib =>	outblockzlib(s);
+	}
+}
+
 #
 # irrecoverable error; invariably denotes data corruption
 #
diff --git a/dis/lib/deflate.dis b/dis/lib/deflate.dis
index c3941ac7..4c0624ac 100644
--- a/dis/lib/deflate.dis
+++ b/dis/lib/deflate.dis
diff --git a/dis/lib/inflate.dis b/dis/lib/inflate.dis
index 89c844bb..4c316b68 100644
--- a/dis/lib/inflate.dis
+++ b/dis/lib/inflate.dis
diff --git a/include/version.h b/include/version.h
index 2ef40a47..61c9db13 100644
--- a/include/version.h
+++ b/include/version.h
@@ -1 +1 @@
-#define VERSION	"Fourth Edition (20090427)"
+#define VERSION	"Fourth Edition (20090430)"
diff --git a/man/2/filter-deflate b/man/2/filter-deflate
index 34b320fb..6de01912 100644
--- a/man/2/filter-deflate
+++ b/man/2/filter-deflate
@@ -44,6 +44,10 @@ Add a gzip header and footer to the data. With this flag,
 the data after filtering will be in exactly the same
 format as a gzip file, with accompanying checksum.
 .TP
+.RB ` z '
+Add a zlib header and footer to the data.  The footer contains a
+checksum.
+.TP
 .RB ` 0 '\ to\ ` 9 '
 Specifies the level of compression to be used (9 highest). See
 .IR gzip (1).
@@ -59,8 +63,10 @@ argument to
 begins with the character
 .RB ` h '
 then the input to the filter is assumed to be in the
-standard gzip file format; the output will be checked
-for integrity. While processing, the
+standard gzip file format; if it starts with the character
+.RB ` z '
+it is assumed to be in zlib format; the output will be checked
+for integrity in both cases. While processing a gzip stream, the
 .B Rq.Info
 message is used to transmit some information; the type
 of information is determined by the first word of
@@ -87,3 +93,9 @@ original file before compression.
 .SH SEE ALSO
 .IR gzip (1),
 .IR filter (2)
+.br
+Internet RFCs
+.IR RFC1950 ,
+.IR RFC1951 ,
+and
+.IR RFC1952 .