gzip/patches/gzip-1.3.13-rsync.patch

Index: gzip-1.5/deflate.c
===================================================================
--- gzip-1.5.orig/deflate.c
+++ gzip-1.5/deflate.c
@@ -131,6 +131,14 @@
 #endif
 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
 
+#ifndef RSYNC_WIN
+#  define RSYNC_WIN 4096
+#endif
+/* Size of rsync window, must be < MAX_DIST */
+
+#define RSYNC_SUM_MATCH(sum) ((sum) % RSYNC_WIN == 0)
+/* Whether window sum matches magic value */
+
 /* ===========================================================================
  * Local data used by the "longest match" routines.
  */
@@ -212,6 +220,8 @@ local int compr_level;
 unsigned good_match;
 /* Use a faster search when the previous match is longer than this */
 
+local ulg rsync_sum;  /* rolling sum of rsync window */
+local ulg rsync_chunk_end; /* next rsync sequence point */
 
 /* Values for max_lazy_match, good_match and max_chain_length, depending on
  * the desired pack level (0..9). The values given below have been tuned to
@@ -314,6 +324,10 @@ void lm_init (pack_level, flags)
 #endif
     /* prev will be initialized on the fly */
 
+    /* rsync params */
+    rsync_chunk_end = 0xFFFFFFFFUL;
+    rsync_sum = 0;
+
     /* Set the default configuration parameters:
      */
     max_lazy_match   = configuration_table[pack_level].max_lazy;
@@ -550,6 +564,8 @@ local void fill_window()
         memcpy((char*)window, (char*)window+WSIZE, (unsigned)WSIZE);
         match_start -= WSIZE;
         strstart    -= WSIZE; /* we now have strstart >= MAX_DIST: */
+        if (rsync_chunk_end != 0xFFFFFFFFUL)
+            rsync_chunk_end -= WSIZE;
 
         block_start -= (long) WSIZE;
 
@@ -579,13 +595,46 @@ local void fill_window()
     }
 }
 
+local void rsync_roll(start, num)
+    unsigned start;
+    unsigned num;
+{
+    unsigned i;
+
+    if (start < RSYNC_WIN) {
+       /* before window fills. */
+       for (i = start; i < RSYNC_WIN; i++) {
+           if (i == start + num) return;
+           rsync_sum += (ulg)window[i];
+       }
+       num -= (RSYNC_WIN - start);
+       start = RSYNC_WIN;
+    }
+
+    /* buffer after window full */
+    for (i = start; i < start+num; i++) {
+       /* New character in */
+       rsync_sum += (ulg)window[i];
+       /* Old character out */
+       rsync_sum -= (ulg)window[i - RSYNC_WIN];
+       if (rsync_chunk_end == 0xFFFFFFFFUL && RSYNC_SUM_MATCH(rsync_sum))
+           rsync_chunk_end = i;
+    }
+}
+
+/* ===========================================================================
+ * Set rsync_chunk_end if window sum matches magic value.
+ */
+#define RSYNC_ROLL(s, n) \
+   do { if (rsync) rsync_roll((s), (n)); } while(0)
+
 /* ===========================================================================
  * Flush the current block, with given end-of-file flag.
  * IN assertion: strstart is set to the end of the current match.
  */
 #define FLUSH_BLOCK(eof) \
    flush_block(block_start >= 0L ? (char*)&window[(unsigned)block_start] : \
-                (char*)NULL, (long)strstart - block_start, (eof))
+                (char*)NULL, (long)strstart - block_start, flush-1, (eof)) 
 
 /* ===========================================================================
  * Processes a new input file and return its compressed length. This
@@ -596,7 +645,7 @@ local void fill_window()
 local off_t deflate_fast()
 {
     IPos hash_head; /* head of the hash chain */
-    int flush;      /* set if current block must be flushed */
+    int flush;      /* set if current block must be flushed, 2=>and padded  */ 
     unsigned match_length = 0;  /* length of best match */
 
     prev_length = MIN_MATCH-1;
@@ -625,7 +674,8 @@ local off_t deflate_fast()
             flush = ct_tally(strstart-match_start, match_length - MIN_MATCH);
 
             lookahead -= match_length;
-
+            
+            RSYNC_ROLL(strstart, match_length);
             /* Insert new strings in the hash table only if the match length
              * is not too large. This saves time but degrades compression.
              */
@@ -654,9 +704,14 @@ local off_t deflate_fast()
             /* No match, output a literal byte */
             Tracevv((stderr,"%c",window[strstart]));
             flush = ct_tally (0, window[strstart]);
+            RSYNC_ROLL(strstart, 1);
             lookahead--;
             strstart++;
         }
+        if (rsync && strstart > rsync_chunk_end) {                                                        
+            rsync_chunk_end = 0xFFFFFFFFUL;                                                               
+            flush = 2;                                                                                    
+       }  
         if (flush) FLUSH_BLOCK(0), block_start = strstart;
 
         /* Make sure that we always have enough lookahead, except
@@ -730,6 +785,7 @@ off_t deflate()
              */
             lookahead -= prev_length-1;
             prev_length -= 2;
+            RSYNC_ROLL(strstart, prev_length+1);
             do {
                 strstart++;
                 INSERT_STRING(strstart, hash_head);
@@ -742,24 +798,40 @@ off_t deflate()
             match_available = 0;
             match_length = MIN_MATCH-1;
             strstart++;
-            if (flush) FLUSH_BLOCK(0), block_start = strstart;
 
+            if (rsync && strstart > rsync_chunk_end) {
+                rsync_chunk_end = 0xFFFFFFFFUL;
+                flush = 2;
+            }
+            if (flush) FLUSH_BLOCK(0), block_start = strstart;
         } else if (match_available) {
             /* If there was no match at the previous position, output a
              * single literal. If there was a match but the current match
              * is longer, truncate the previous match to a single literal.
              */
             Tracevv((stderr,"%c",window[strstart-1]));
-            if (ct_tally (0, window[strstart-1])) {
-                FLUSH_BLOCK(0), block_start = strstart;
+            flush = ct_tally (0, window[strstart-1]);
+            if (rsync && strstart > rsync_chunk_end) {
+                rsync_chunk_end = 0xFFFFFFFFUL;
+                flush = 2;
             }
+            if (flush) FLUSH_BLOCK(0), block_start = strstart;
+            RSYNC_ROLL(strstart, 1);
             strstart++;
             lookahead--;
         } else {
             /* There is no previous match to compare with, wait for
              * the next step to decide.
              */
+            if (rsync && strstart > rsync_chunk_end) {
+                /* Reset huffman tree */
+                rsync_chunk_end = 0xFFFFFFFFUL;
+                flush = 2;
+                FLUSH_BLOCK(0), block_start = strstart;
+            }             
+             
             match_available = 1;
+            RSYNC_ROLL(strstart, 1);  
             strstart++;
             lookahead--;
         }
Index: gzip-1.5/doc/gzip.texi
===================================================================
--- gzip-1.5.orig/doc/gzip.texi
+++ gzip-1.5/doc/gzip.texi
@@ -353,6 +353,14 @@ specified on the command line are direct
 into the directory and compress all the files it finds there (or
 decompress them in the case of @command{gunzip}).
 
+@item --rsyncable
+While compressing, synchronize the output occasionally based on the
+input.  This reduces compression by about 1 percent most cases, but
+means that the @code{rsync} program can take advantage of similarities
+in the uncompressed input when syncronizing two files compressed with
+this flag.  @code{gunzip} cannot tell the difference between a
+compressed file created with this option, and one created without it.
+
 @item --suffix @var{suf}
 @itemx -S @var{suf}
 Use suffix @var{suf} instead of @samp{.gz}.  Any suffix can be
Index: gzip-1.5/gzip.c
===================================================================
--- gzip-1.5.orig/gzip.c
+++ gzip-1.5/gzip.c
@@ -213,6 +213,7 @@ int  ofd;                  /* output fil
 unsigned insize;           /* valid bytes in inbuf */
 unsigned inptr;            /* index of next byte to be processed in inbuf */
 unsigned outcnt;           /* bytes in output buffer */
+int rsync = 0;             /* make ryncable chunks */
 
 static int handled_sig[] =
   {
@@ -270,7 +271,7 @@ static const struct option longopts[] =
     {"best",       0, 0, '9'}, /* compress better */
     {"lzw",        0, 0, 'Z'}, /* make output compatible with old compress */
     {"bits",       1, 0, 'b'}, /* max number of bits per code (implies -Z) */
-
+    {"rsyncable",  0, 0, 'R'}, /* make rsync-friendly archive */ 
     { 0, 0, 0, 0 }
 };
 
@@ -353,6 +354,7 @@ local void help()
  "  -Z, --lzw         produce output compatible with old compress",
  "  -b, --bits=BITS   max number of bits per code (implies -Z)",
 #endif
+ "    --rsyncable   Make rsync-friendly archive",  
  "",
  "With no FILE, or when FILE is -, read standard input.",
  "",
@@ -482,6 +484,9 @@ int main (int argc, char **argv)
             recursive = 1;
 #endif
             break;
+            
+        case 'R':
+            rsync = 1; break;
         case 'S':
 #ifdef NO_MULTIPLE_DOTS
             if (*optarg == '.') optarg++;
Index: gzip-1.5/gzip.h
===================================================================
--- gzip-1.5.orig/gzip.h
+++ gzip-1.5/gzip.h
@@ -140,6 +140,7 @@ EXTERN(uch, window);         /* Sliding
 extern unsigned insize; /* valid bytes in inbuf */
 extern unsigned inptr;  /* index of next byte to be processed in inbuf */
 extern unsigned outcnt; /* bytes in output buffer */
+extern int rsync;  /* deflate into rsyncable chunks */ 
 
 extern off_t bytes_in;   /* number of input bytes */
 extern off_t bytes_out;  /* number of output bytes */
@@ -287,7 +288,7 @@ extern off_t deflate (void);
         /* in trees.c */
 extern void ct_init     (ush *attr, int *method);
 extern int  ct_tally    (int dist, int lc);
-extern off_t flush_block (char *buf, ulg stored_len, int eof);
+extern off_t flush_block (char *buf, ulg stored_len, int pad, int eof);
 
         /* in bits.c */
 extern void     bi_init    (file_t zipfile);
Index: gzip-1.5/trees.c
===================================================================
--- gzip-1.5.orig/trees.c
+++ gzip-1.5/trees.c
@@ -856,9 +856,10 @@ local void send_all_trees(lcodes, dcodes
  * trees or store, and output the encoded block to the zip file. This function
  * returns the total compressed length for the file so far.
  */
-off_t flush_block(buf, stored_len, eof)
+off_t flush_block(buf, stored_len, pad, eof)  
     char *buf;        /* input block, or NULL if too old */
     ulg stored_len;   /* length of input block */
+    int pad;          /* pad output to byte boundary */ 
     int eof;          /* true if this is the last block for a file */
 {
     ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
@@ -951,6 +952,10 @@ off_t flush_block(buf, stored_len, eof)
         Assert (input_len == bytes_in, "bad input size");
         bi_windup();
         compressed_len += 7;  /* align on byte boundary */
+    } else if (pad && (compressed_len % 8) != 0) {
+        send_bits((STORED_BLOCK<<1)+eof, 3);  /* send block type */
+        compressed_len = (compressed_len + 3 + 7) & ~7L;
+        copy_block(buf, 0, 1); /* with header */
     }
 
     return compressed_len >> 3;
1	Index: gzip-1.5/deflate.c
2	===================================================================
3	--- gzip-1.5.orig/deflate.c
4	+++ gzip-1.5/deflate.c
5	@@ -131,6 +131,14 @@
6	#endif
7	/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
8
9	+#ifndef RSYNC_WIN
10	+# define RSYNC_WIN 4096
11	+#endif
12	+/* Size of rsync window, must be < MAX_DIST */
13	+
14	+#define RSYNC_SUM_MATCH(sum) ((sum) % RSYNC_WIN == 0)
15	+/* Whether window sum matches magic value */
16	+
17	/* ===========================================================================
18	* Local data used by the "longest match" routines.
19	*/
20	@@ -212,6 +220,8 @@ local int compr_level;
21	unsigned good_match;
22	/* Use a faster search when the previous match is longer than this */
23
24	+local ulg rsync_sum; /* rolling sum of rsync window */
25	+local ulg rsync_chunk_end; /* next rsync sequence point */
26
27	/* Values for max_lazy_match, good_match and max_chain_length, depending on
28	* the desired pack level (0..9). The values given below have been tuned to
29	@@ -314,6 +324,10 @@ void lm_init (pack_level, flags)
30	#endif
31	/* prev will be initialized on the fly */
32
33	+ /* rsync params */
34	+ rsync_chunk_end = 0xFFFFFFFFUL;
35	+ rsync_sum = 0;
36	+
37	/* Set the default configuration parameters:
38	*/
39	max_lazy_match = configuration_table[pack_level].max_lazy;
40	@@ -550,6 +564,8 @@ local void fill_window()
41	memcpy((char)window, (char)window+WSIZE, (unsigned)WSIZE);
42	match_start -= WSIZE;
43	strstart -= WSIZE; /* we now have strstart >= MAX_DIST: */
44	+ if (rsync_chunk_end != 0xFFFFFFFFUL)
45	+ rsync_chunk_end -= WSIZE;
46
47	block_start -= (long) WSIZE;
48
49	@@ -579,13 +595,46 @@ local void fill_window()
50	}
51	}
52
53	+local void rsync_roll(start, num)
54	+ unsigned start;
55	+ unsigned num;
56	+{
57	+ unsigned i;
58	+
59	+ if (start < RSYNC_WIN) {
60	+ /* before window fills. */
61	+ for (i = start; i < RSYNC_WIN; i++) {
62	+ if (i == start + num) return;
63	+ rsync_sum += (ulg)window[i];
64	+ }
65	+ num -= (RSYNC_WIN - start);
66	+ start = RSYNC_WIN;
67	+ }
68	+
69	+ /* buffer after window full */
70	+ for (i = start; i < start+num; i++) {
71	+ /* New character in */
72	+ rsync_sum += (ulg)window[i];
73	+ /* Old character out */
74	+ rsync_sum -= (ulg)window[i - RSYNC_WIN];
75	+ if (rsync_chunk_end == 0xFFFFFFFFUL && RSYNC_SUM_MATCH(rsync_sum))
76	+ rsync_chunk_end = i;
77	+ }
78	+}
79	+
80	+/* ===========================================================================
81	+ * Set rsync_chunk_end if window sum matches magic value.
82	+ */
83	+#define RSYNC_ROLL(s, n) \
84	+ do { if (rsync) rsync_roll((s), (n)); } while(0)
85	+
86	/* ===========================================================================
87	* Flush the current block, with given end-of-file flag.
88	* IN assertion: strstart is set to the end of the current match.
89	*/
90	#define FLUSH_BLOCK(eof) \
91	flush_block(block_start >= 0L ? (char*)&window[(unsigned)block_start] : \
92	- (char*)NULL, (long)strstart - block_start, (eof))
93	+ (char*)NULL, (long)strstart - block_start, flush-1, (eof))
94
95	/* ===========================================================================
96	* Processes a new input file and return its compressed length. This
97	@@ -596,7 +645,7 @@ local void fill_window()
98	local off_t deflate_fast()
99	{
100	IPos hash_head; /* head of the hash chain */
101	- int flush; /* set if current block must be flushed */
102	+ int flush; /* set if current block must be flushed, 2=>and padded */
103	unsigned match_length = 0; /* length of best match */
104
105	prev_length = MIN_MATCH-1;
106	@@ -625,7 +674,8 @@ local off_t deflate_fast()
107	flush = ct_tally(strstart-match_start, match_length - MIN_MATCH);
108
109	lookahead -= match_length;
110	-
111	+
112	+ RSYNC_ROLL(strstart, match_length);
113	/* Insert new strings in the hash table only if the match length
114	* is not too large. This saves time but degrades compression.
115	*/
116	@@ -654,9 +704,14 @@ local off_t deflate_fast()
117	/* No match, output a literal byte */
118	Tracevv((stderr,"%c",window[strstart]));
119	flush = ct_tally (0, window[strstart]);
120	+ RSYNC_ROLL(strstart, 1);
121	lookahead--;
122	strstart++;
123	}
124	+ if (rsync && strstart > rsync_chunk_end) {
125	+ rsync_chunk_end = 0xFFFFFFFFUL;
126	+ flush = 2;
127	+ }
128	if (flush) FLUSH_BLOCK(0), block_start = strstart;
129
130	/* Make sure that we always have enough lookahead, except
131	@@ -730,6 +785,7 @@ off_t deflate()
132	*/
133	lookahead -= prev_length-1;
134	prev_length -= 2;
135	+ RSYNC_ROLL(strstart, prev_length+1);
136	do {
137	strstart++;
138	INSERT_STRING(strstart, hash_head);
139	@@ -742,24 +798,40 @@ off_t deflate()
140	match_available = 0;
141	match_length = MIN_MATCH-1;
142	strstart++;
143	- if (flush) FLUSH_BLOCK(0), block_start = strstart;
144
145	+ if (rsync && strstart > rsync_chunk_end) {
146	+ rsync_chunk_end = 0xFFFFFFFFUL;
147	+ flush = 2;
148	+ }
149	+ if (flush) FLUSH_BLOCK(0), block_start = strstart;
150	} else if (match_available) {
151	/* If there was no match at the previous position, output a
152	* single literal. If there was a match but the current match
153	* is longer, truncate the previous match to a single literal.
154	*/
155	Tracevv((stderr,"%c",window[strstart-1]));
156	- if (ct_tally (0, window[strstart-1])) {
157	- FLUSH_BLOCK(0), block_start = strstart;
158	+ flush = ct_tally (0, window[strstart-1]);
159	+ if (rsync && strstart > rsync_chunk_end) {
160	+ rsync_chunk_end = 0xFFFFFFFFUL;
161	+ flush = 2;
162	}
163	+ if (flush) FLUSH_BLOCK(0), block_start = strstart;
164	+ RSYNC_ROLL(strstart, 1);
165	strstart++;
166	lookahead--;
167	} else {
168	/* There is no previous match to compare with, wait for
169	* the next step to decide.
170	*/
171	+ if (rsync && strstart > rsync_chunk_end) {
172	+ /* Reset huffman tree */
173	+ rsync_chunk_end = 0xFFFFFFFFUL;
174	+ flush = 2;
175	+ FLUSH_BLOCK(0), block_start = strstart;
176	+ }
177	+
178	match_available = 1;
179	+ RSYNC_ROLL(strstart, 1);
180	strstart++;
181	lookahead--;
182	}
183	Index: gzip-1.5/doc/gzip.texi
184	===================================================================
185	--- gzip-1.5.orig/doc/gzip.texi
186	+++ gzip-1.5/doc/gzip.texi
187	@@ -353,6 +353,14 @@ specified on the command line are direct
188	into the directory and compress all the files it finds there (or
189	decompress them in the case of @command{gunzip}).
190
191	+@item --rsyncable
192	+While compressing, synchronize the output occasionally based on the
193	+input. This reduces compression by about 1 percent most cases, but
194	+means that the @code{rsync} program can take advantage of similarities
195	+in the uncompressed input when syncronizing two files compressed with
196	+this flag. @code{gunzip} cannot tell the difference between a
197	+compressed file created with this option, and one created without it.
198	+
199	@item --suffix @var{suf}
200	@itemx -S @var{suf}
201	Use suffix @var{suf} instead of @samp{.gz}. Any suffix can be
202	Index: gzip-1.5/gzip.c
203	===================================================================
204	--- gzip-1.5.orig/gzip.c
205	+++ gzip-1.5/gzip.c
206	@@ -213,6 +213,7 @@ int ofd; /* output fil
207	unsigned insize; /* valid bytes in inbuf */
208	unsigned inptr; /* index of next byte to be processed in inbuf */
209	unsigned outcnt; /* bytes in output buffer */
210	+int rsync = 0; /* make ryncable chunks */
211
212	static int handled_sig[] =
213	{
214	@@ -270,7 +271,7 @@ static const struct option longopts[] =
215	{"best", 0, 0, '9'}, /* compress better */
216	{"lzw", 0, 0, 'Z'}, /* make output compatible with old compress */
217	{"bits", 1, 0, 'b'}, /* max number of bits per code (implies -Z) */
218	-
219	+ {"rsyncable", 0, 0, 'R'}, /* make rsync-friendly archive */
220	{ 0, 0, 0, 0 }
221	};
222
223	@@ -353,6 +354,7 @@ local void help()
224	" -Z, --lzw produce output compatible with old compress",
225	" -b, --bits=BITS max number of bits per code (implies -Z)",
226	#endif
227	+ " --rsyncable Make rsync-friendly archive",
228	"",
229	"With no FILE, or when FILE is -, read standard input.",
230	"",
231	@@ -482,6 +484,9 @@ int main (int argc, char **argv)
232	recursive = 1;
233	#endif
234	break;
235	+
236	+ case 'R':
237	+ rsync = 1; break;
238	case 'S':
239	#ifdef NO_MULTIPLE_DOTS
240	if (*optarg == '.') optarg++;
241	Index: gzip-1.5/gzip.h
242	===================================================================
243	--- gzip-1.5.orig/gzip.h
244	+++ gzip-1.5/gzip.h
245	@@ -140,6 +140,7 @@ EXTERN(uch, window); /* Sliding
246	extern unsigned insize; /* valid bytes in inbuf */
247	extern unsigned inptr; /* index of next byte to be processed in inbuf */
248	extern unsigned outcnt; /* bytes in output buffer */
249	+extern int rsync; /* deflate into rsyncable chunks */
250
251	extern off_t bytes_in; /* number of input bytes */
252	extern off_t bytes_out; /* number of output bytes */
253	@@ -287,7 +288,7 @@ extern off_t deflate (void);
254	/* in trees.c */
255	extern void ct_init (ush attr, int method);
256	extern int ct_tally (int dist, int lc);
257	-extern off_t flush_block (char *buf, ulg stored_len, int eof);
258	+extern off_t flush_block (char *buf, ulg stored_len, int pad, int eof);
259
260	/* in bits.c */
261	extern void bi_init (file_t zipfile);
262	Index: gzip-1.5/trees.c
263	===================================================================
264	--- gzip-1.5.orig/trees.c
265	+++ gzip-1.5/trees.c
266	@@ -856,9 +856,10 @@ local void send_all_trees(lcodes, dcodes
267	* trees or store, and output the encoded block to the zip file. This function
268	* returns the total compressed length for the file so far.
269	*/
270	-off_t flush_block(buf, stored_len, eof)
271	+off_t flush_block(buf, stored_len, pad, eof)
272	char buf; / input block, or NULL if too old */
273	ulg stored_len; /* length of input block */
274	+ int pad; /* pad output to byte boundary */
275	int eof; /* true if this is the last block for a file */
276	{
277	ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
278	@@ -951,6 +952,10 @@ off_t flush_block(buf, stored_len, eof)
279	Assert (input_len == bytes_in, "bad input size");
280	bi_windup();
281	compressed_len += 7; /* align on byte boundary */
282	+ } else if (pad && (compressed_len % 8) != 0) {
283	+ send_bits((STORED_BLOCK<<1)+eof, 3); /* send block type */
284	+ compressed_len = (compressed_len + 3 + 7) & ~7L;
285	+ copy_block(buf, 0, 1); /* with header */
286	}
287
288	return compressed_len >> 3;