-In message <20020403094739.GA5434@samba.org> you write:
-> On 3 Apr 2002, Rusty Russell <rusty@rustcorp.com.au> wrote:
-> > In message <20020403020455.GC18851@samba.org> you write:
-> > > Hi,
-> > >
-> > > I think you said the other day that you had a working --rsyncable
-> > > patch for gzip. Could I have it please?
->
-> Thanks for that. I'll put it in with our contrib patches and probably
-> merge it soon.
->
-> I was actually after the patch to gzip, rather than the fuzzy patch
-> for rsync. I'd like to try it myself and also distribute it along
-> with rsync.
+NOTE: this patch is for _gzip_!
-It's pending for actual gzip inclusion! The maintainer was dug up and
-everything: it'll even be on by default.
+This is pending for actual inclusion in gzip. It is currently being
+tried out in the default gzip for Debian Sarge, and may go into the
+upstream gzip at somepoint in the not-too-distant future.
-But here's the old patch,
-Rusty.
---
- Anyone who quotes me in their sig is an idiot. -- Rusty Russell.
-
-diff -ur gzip-1.2.4.orig/deflate.c gzip-1.2.4-rsync/deflate.c
---- gzip-1.2.4.orig/deflate.c Fri Aug 13 22:35:31 1993
-+++ gzip-1.2.4-rsync/deflate.c Sat Dec 30 15:33:25 2000
-@@ -121,6 +121,14 @@
+--- gzip-1.3.5/deflate.c 1999-10-07 23:46:28 -0700
++++ deflate.c 2005-02-05 09:40:33 -0800
+@@ -122,6 +122,14 @@
#endif
/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
/* ===========================================================================
* Local data used by the "longest match" routines.
*/
-@@ -202,6 +210,8 @@
+@@ -203,6 +211,8 @@
unsigned near good_match;
/* Use a faster search when the previous match is longer than this */
/* 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
-@@ -300,6 +310,10 @@
+@@ -301,6 +311,10 @@
#endif
/* prev will be initialized on the fly */
/* Set the default configuration parameters:
*/
max_lazy_match = configuration_table[pack_level].max_lazy;
-@@ -536,6 +550,8 @@
+@@ -537,6 +551,8 @@
memcpy((char*)window, (char*)window+WSIZE, (unsigned)WSIZE);
match_start -= WSIZE;
strstart -= WSIZE; /* we now have strstart >= MAX_DIST: */
block_start -= (long) WSIZE;
-@@ -563,13 +579,46 @@
+@@ -564,13 +580,46 @@
}
}
/* ===========================================================================
* Processes a new input file and return its compressed length. This
-@@ -580,7 +629,7 @@
- local ulg deflate_fast()
+@@ -581,7 +630,7 @@
+ local off_t deflate_fast()
{
IPos hash_head; /* head of the hash chain */
- int flush; /* set if current block must be flushed */
unsigned match_length = 0; /* length of best match */
prev_length = MIN_MATCH-1;
-@@ -609,6 +658,7 @@
+@@ -610,6 +659,7 @@
lookahead -= match_length;
/* Insert new strings in the hash table only if the match length
* is not too large. This saves time but degrades compression.
*/
-@@ -637,9 +687,14 @@
+@@ -638,9 +688,14 @@
/* No match, output a literal byte */
Tracevv((stderr,"%c",window[strstart]));
flush = ct_tally (0, window[strstart]);
if (flush) FLUSH_BLOCK(0), block_start = strstart;
/* Make sure that we always have enough lookahead, except
-@@ -715,6 +770,7 @@
+@@ -713,6 +768,7 @@
*/
lookahead -= prev_length-1;
prev_length -= 2;
do {
strstart++;
INSERT_STRING(strstart, hash_head);
-@@ -727,24 +783,39 @@
+@@ -725,24 +781,39 @@
match_available = 0;
match_length = MIN_MATCH-1;
strstart++;
strstart++;
lookahead--;
}
-diff -ur gzip-1.2.4.orig/gzip.c gzip-1.2.4-rsync/gzip.c
---- gzip-1.2.4.orig/gzip.c Thu Aug 19 23:39:43 1993
-+++ gzip-1.2.4-rsync/gzip.c Fri Dec 29 21:20:54 2000
-@@ -239,6 +239,7 @@
+--- gzip-1.3.5/gzip.c 2002-09-28 00:38:43 -0700
++++ gzip.c 2005-02-05 09:40:33 -0800
+@@ -256,6 +256,7 @@
unsigned insize; /* valid bytes in inbuf */
unsigned inptr; /* index of next byte to be processed in inbuf */
unsigned outcnt; /* bytes in output buffer */
struct option longopts[] =
{
-@@ -268,6 +269,7 @@
+@@ -285,6 +286,7 @@
{"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) */
{ 0, 0, 0, 0 }
};
-@@ -357,6 +359,7 @@
+@@ -365,6 +367,7 @@
" -Z --lzw produce output compatible with old compress",
" -b --bits maxbits max number of bits per code (implies -Z)",
#endif
+ " --rsyncable Make rsync-friendly archive",
" file... files to (de)compress. If none given, use standard input.",
+ "Report bugs to <bug-gzip@gnu.org>.",
0};
- char **p = help_msg;
-@@ -516,6 +519,9 @@
+@@ -543,6 +546,9 @@
#else
recursive = 1; break;
#endif
case 'S':
#ifdef NO_MULTIPLE_DOTS
if (*optarg == '.') optarg++;
-diff -ur gzip-1.2.4.orig/gzip.h gzip-1.2.4-rsync/gzip.h
---- gzip-1.2.4.orig/gzip.h Fri Aug 13 22:35:33 1993
-+++ gzip-1.2.4-rsync/gzip.h Sat Dec 30 15:26:56 2000
-@@ -131,6 +131,7 @@
+--- gzip-1.3.5/gzip.h 2001-09-30 23:53:41 -0700
++++ gzip.h 2005-02-05 09:40:33 -0800
+@@ -133,6 +133,7 @@
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 long bytes_in; /* number of input bytes */
- extern long bytes_out; /* number of output bytes */
-@@ -282,7 +283,7 @@
+ extern off_t bytes_in; /* number of input bytes */
+ extern off_t bytes_out; /* number of output bytes */
+@@ -281,7 +282,7 @@
/* in trees.c */
void ct_init OF((ush *attr, int *method));
int ct_tally OF((int dist, int lc));
--ulg flush_block OF((char *buf, ulg stored_len, int eof));
-+ulg flush_block OF((char *buf, ulg stored_len, int pad, int eof));
+-off_t flush_block OF((char *buf, ulg stored_len, int eof));
++off_t flush_block OF((char *buf, ulg stored_len, int pad, int eof));
/* in bits.c */
void bi_init OF((file_t zipfile));
-diff -ur gzip-1.2.4.orig/gzip.texi gzip-1.2.4-rsync/gzip.texi
---- gzip-1.2.4.orig/gzip.texi Thu Aug 19 06:42:50 1993
-+++ gzip-1.2.4-rsync/gzip.texi Fri Dec 29 21:20:54 2000
-@@ -316,6 +316,14 @@
+--- gzip-1.3.5/gzip.texi 2002-09-29 23:57:29 -0700
++++ gzip.texi 2005-02-05 09:40:33 -0800
+@@ -334,6 +334,14 @@
into the directory and compress all the files it finds there (or
decompress them in the case of @code{gunzip}).
+@item --rsyncable
+While compressing, synchronize the output occasionally based on the
-+input. This reduces compression by about 1 percent most cases, but
++input. This increases size by less than 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
@item --suffix @var{suf}
@itemx -S @var{suf}
Use suffix @samp{@var{suf}} instead of @samp{.gz}. Any suffix can be
-diff -ur gzip-1.2.4.orig/trees.c gzip-1.2.4-rsync/trees.c
---- gzip-1.2.4.orig/trees.c Wed Aug 18 03:36:32 1993
-+++ gzip-1.2.4-rsync/trees.c Sat Dec 30 15:37:00 2000
-@@ -850,9 +850,10 @@
+--- gzip-1.3.5/trees.c 1999-10-05 22:00:00 -0700
++++ trees.c 2005-02-05 09:40:33 -0800
+@@ -46,12 +46,13 @@
+ * void ct_tally (int dist, int lc);
+ * Save the match info and tally the frequency counts.
+ *
+- * off_t flush_block (char *buf, ulg stored_len, int eof)
++ * off_t flush_block (char *buf, ulg stored_len, int pad, int eof)
+ * Determine the best encoding for the current block: dynamic trees,
+ * static trees or store, and output the encoded block to the zip
+- * file. Returns the total compressed length for the file so far.
+- *
+- */
++ * file. If pad is set, pads the block to the next
++ * byte. Returns the total compressed length for the file so
++ * far.
++ * */
+
+ #include <config.h>
+ #include <ctype.h>
+@@ -847,9 +848,10 @@
* trees or store, and output the encoded block to the zip file. This function
* returns the total compressed length for the file so far.
*/
--ulg flush_block(buf, stored_len, eof)
-+ulg flush_block(buf, stored_len, pad, eof)
+-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 */
-@@ -944,6 +967,10 @@
- Assert (input_len == isize, "bad input size");
+@@ -941,6 +943,10 @@
+ Assert (input_len == bytes_in, "bad input size");
bi_windup();
compressed_len += 7; /* align on byte boundary */
+ } else if (pad && (compressed_len % 8) != 0) {
+ compressed_len = (compressed_len + 3 + 7) & ~7L;
+ copy_block(buf, 0, 1); /* with header */
}
- Tracev((stderr,"\ncomprlen %lu(%lu) ", compressed_len>>3,
- compressed_len-7*eof));
+
+ return compressed_len >> 3;