30 files changed, 2691 insertions, 375 deletions

diff --git a/.gitignore b/.gitignore
index b5959d6311..7906909b30 100644
--- a/.gitignore
+++ b/.gitignore
@@ -123,6 +123,7 @@ git-write-tree
 git-core-*/?*
 test-date
 test-delta
+test-dump-cache-tree
 common-cmds.h
 *.tar.gz
 *.dsc
diff --git a/Documentation/git-update-index.txt b/Documentation/git-update-index.txt
index 57177c7fb6..d043e86a77 100644
--- a/Documentation/git-update-index.txt
+++ b/Documentation/git-update-index.txt
@@ -15,7 +15,7 @@ SYNOPSIS
 	     [--cacheinfo <mode> <object> <file>]\*
 	     [--chmod=(+|-)x]
 	     [--assume-unchanged | --no-assume-unchanged]
-	     [--really-refresh] [--unresolve]
+	     [--really-refresh] [--unresolve] [--again]
 	     [--info-only] [--index-info]
 	     [-z] [--stdin]
 	     [--verbose]
@@ -80,6 +80,10 @@ OPTIONS
 	filesystem that has very slow lstat(2) system call
 	(e.g. cifs).
 
+--again::
+	Runs `git-update-index` itself on the paths whose index
+	entries are different from those from the `HEAD` commit.
+
 --unresolve::
 	Restores the 'unmerged' or 'needs updating' state of a
 	file during a merge if it was cleared by accident.
diff --git a/Makefile b/Makefile
index 45484fcd82..070c478bf2 100644
--- a/Makefile
+++ b/Makefile
@@ -205,7 +205,7 @@ DIFF_OBJS = \
 	diffcore-delta.o log-tree.o
 
 LIB_OBJS = \
-	blob.o commit.o connect.o csum-file.o \
+	blob.o commit.o connect.o csum-file.o cache-tree.o base85.o \
 	date.o diff-delta.o entry.o exec_cmd.o ident.o index.o \
 	object.o pack-check.o patch-delta.o path.o pkt-line.o \
 	quote.o read-cache.o refs.o run-command.o \
@@ -215,7 +215,8 @@ LIB_OBJS = \
 	$(DIFF_OBJS)
 
 BUILTIN_OBJS = \
-	builtin-log.o builtin-help.o builtin-count.o builtin-diff.o builtin-push.o
+	builtin-log.o builtin-help.o builtin-count.o builtin-diff.o builtin-push.o \
+	builtin-grep.o
 
 GITLIBS = $(LIB_FILE) $(XDIFF_LIB)
 LIBS = $(GITLIBS) -lz
@@ -607,7 +608,10 @@ test-date$X: test-date.c date.o ctype.o
 	$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) test-date.c date.o ctype.o
 
 test-delta$X: test-delta.c diff-delta.o patch-delta.o
-	$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $^ -lz
+	$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $^
+
+test-dump-cache-tree$X: dump-cache-tree.o $(GITLIBS)
+	$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) $(LIBS)
 
 check:
 	for i in *.c; do sparse $(ALL_CFLAGS) $(SPARSE_FLAGS) $$i || exit; done
diff --git a/apply.c b/apply.c
index 269210a578..88d2a3206b 100644
--- a/apply.c
+++ b/apply.c
@@ -8,8 +8,10 @@
  */
 #include <fnmatch.h>
 #include "cache.h"
+#include "cache-tree.h"
 #include "quote.h"
 #include "blob.h"
+#include "delta.h"
 
 //  --check turns on checking that the working tree matches the
 //    files that are being modified, but doesn't apply the patch
@@ -113,6 +115,9 @@ struct patch {
 	char *new_name, *old_name, *def_name;
 	unsigned int old_mode, new_mode;
 	int is_rename, is_copy, is_new, is_delete, is_binary;
+#define BINARY_DELTA_DEFLATED 1
+#define BINARY_LITERAL_DEFLATED 2
+	unsigned long deflate_origlen;
 	int lines_added, lines_deleted;
 	int score;
 	struct fragment *fragments;
@@ -966,6 +971,88 @@ static inline int metadata_changes(struct patch *patch)
 		 patch->old_mode != patch->new_mode);
 }
 
+static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
+{
+	/* We have read "GIT binary patch\n"; what follows is a line
+	 * that says the patch method (currently, either "deflated
+	 * literal" or "deflated delta") and the length of data before
+	 * deflating; a sequence of 'length-byte' followed by base-85
+	 * encoded data follows.
+	 *
+	 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
+	 * and we would limit the patch line to 66 characters,
+	 * so one line can fit up to 13 groups that would decode
+	 * to 52 bytes max.  The length byte 'A'-'Z' corresponds
+	 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
+	 * The end of binary is signalled with an empty line.
+	 */
+	int llen, used;
+	struct fragment *fragment;
+	char *data = NULL;
+
+	patch->fragments = fragment = xcalloc(1, sizeof(*fragment));
+
+	/* Grab the type of patch */
+	llen = linelen(buffer, size);
+	used = llen;
+	linenr++;
+
+	if (!strncmp(buffer, "delta ", 6)) {
+		patch->is_binary = BINARY_DELTA_DEFLATED;
+		patch->deflate_origlen = strtoul(buffer + 6, NULL, 10);
+	}
+	else if (!strncmp(buffer, "literal ", 8)) {
+		patch->is_binary = BINARY_LITERAL_DEFLATED;
+		patch->deflate_origlen = strtoul(buffer + 8, NULL, 10);
+	}
+	else
+		return error("unrecognized binary patch at line %d: %.*s",
+			     linenr-1, llen-1, buffer);
+	buffer += llen;
+	while (1) {
+		int byte_length, max_byte_length, newsize;
+		llen = linelen(buffer, size);
+		used += llen;
+		linenr++;
+		if (llen == 1)
+			break;
+		/* Minimum line is "A00000\n" which is 7-byte long,
+		 * and the line length must be multiple of 5 plus 2.
+		 */
+		if ((llen < 7) || (llen-2) % 5)
+			goto corrupt;
+		max_byte_length = (llen - 2) / 5 * 4;
+		byte_length = *buffer;
+		if ('A' <= byte_length && byte_length <= 'Z')
+			byte_length = byte_length - 'A' + 1;
+		else if ('a' <= byte_length && byte_length <= 'z')
+			byte_length = byte_length - 'a' + 27;
+		else
+			goto corrupt;
+		/* if the input length was not multiple of 4, we would
+		 * have filler at the end but the filler should never
+		 * exceed 3 bytes
+		 */
+		if (max_byte_length < byte_length ||
+		    byte_length <= max_byte_length - 4)
+			goto corrupt;
+		newsize = fragment->size + byte_length;
+		data = xrealloc(data, newsize);
+		if (decode_85(data + fragment->size,
+			      buffer + 1,
+			      byte_length))
+			goto corrupt;
+		fragment->size = newsize;
+		buffer += llen;
+		size -= llen;
+	}
+	fragment->patch = data;
+	return used;
+ corrupt:
+	return error("corrupt binary patch at line %d: %.*s",
+		     linenr-1, llen-1, buffer);
+}
+
 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
 {
 	int hdrsize, patchsize;
@@ -982,19 +1069,34 @@ static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
 			"Files ",
 			NULL,
 		};
+		static const char git_binary[] = "GIT binary patch\n";
 		int i;
 		int hd = hdrsize + offset;
 		unsigned long llen = linelen(buffer + hd, size - hd);
 
-		if (!memcmp(" differ\n", buffer + hd + llen - 8, 8))
+		if (llen == sizeof(git_binary) - 1 &&
+		    !memcmp(git_binary, buffer + hd, llen)) {
+			int used;
+			linenr++;
+			used = parse_binary(buffer + hd + llen,
+					    size - hd - llen, patch);
+			if (used)
+				patchsize = used + llen;
+			else
+				patchsize = 0;
+		}
+		else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
 			for (i = 0; binhdr[i]; i++) {
 				int len = strlen(binhdr[i]);
 				if (len < size - hd &&
 				    !memcmp(binhdr[i], buffer + hd, len)) {
+					linenr++;
 					patch->is_binary = 1;
+					patchsize = llen;
 					break;
 				}
 			}
+		}
 
 		/* Empty patch cannot be applied if:
 		 * - it is a binary patch and we do not do binary_replace, or
@@ -1345,76 +1447,150 @@ static int apply_one_fragment(struct buffer_desc *desc, struct fragment *frag)
 	return offset;
 }
 
-static int apply_fragments(struct buffer_desc *desc, struct patch *patch)
+static char *inflate_it(const void *data, unsigned long size,
+			unsigned long inflated_size)
+{
+	z_stream stream;
+	void *out;
+	int st;
+
+	memset(&stream, 0, sizeof(stream));
+
+	stream.next_in = (unsigned char *)data;
+	stream.avail_in = size;
+	stream.next_out = out = xmalloc(inflated_size);
+	stream.avail_out = inflated_size;
+	inflateInit(&stream);
+	st = inflate(&stream, Z_FINISH);
+	if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
+		free(out);
+		return NULL;
+	}
+	return out;
+}
+
+static int apply_binary_fragment(struct buffer_desc *desc, struct patch *patch)
+{
+	unsigned long dst_size;
+	struct fragment *fragment = patch->fragments;
+	void *data;
+	void *result;
+
+	data = inflate_it(fragment->patch, fragment->size,
+			  patch->deflate_origlen);
+	if (!data)
+		return error("corrupt patch data");
+	switch (patch->is_binary) {
+	case BINARY_DELTA_DEFLATED:
+		result = patch_delta(desc->buffer, desc->size,
+				     data,
+				     patch->deflate_origlen,
+				     &dst_size);
+		free(desc->buffer);
+		desc->buffer = result;
+		free(data);
+		break;
+	case BINARY_LITERAL_DEFLATED:
+		free(desc->buffer);
+		desc->buffer = data;
+		dst_size = patch->deflate_origlen;
+		break;
+	}
+	if (!desc->buffer)
+		return -1;
+	desc->size = desc->alloc = dst_size;
+	return 0;
+}
+
+static int apply_binary(struct buffer_desc *desc, struct patch *patch)
 {
-	struct fragment *frag = patch->fragments;
 	const char *name = patch->old_name ? patch->old_name : patch->new_name;
+	unsigned char sha1[20];
+	unsigned char hdr[50];
+	int hdrlen;
 
-	if (patch->is_binary) {
-		unsigned char sha1[20];
+	if (!allow_binary_replacement)
+		return error("cannot apply binary patch to '%s' "
+			     "without --allow-binary-replacement",
+			     name);
 
-		if (!allow_binary_replacement)
-			return error("cannot apply binary patch to '%s' "
-				     "without --allow-binary-replacement",
-				     name);
+	/* For safety, we require patch index line to contain
+	 * full 40-byte textual SHA1 for old and new, at least for now.
+	 */
+	if (strlen(patch->old_sha1_prefix) != 40 ||
+	    strlen(patch->new_sha1_prefix) != 40 ||
+	    get_sha1_hex(patch->old_sha1_prefix, sha1) ||
+	    get_sha1_hex(patch->new_sha1_prefix, sha1))
+		return error("cannot apply binary patch to '%s' "
+			     "without full index line", name);
 
-		/* For safety, we require patch index line to contain
-		 * full 40-byte textual SHA1 for old and new, at least for now.
+	if (patch->old_name) {
+		/* See if the old one matches what the patch
+		 * applies to.
 		 */
-		if (strlen(patch->old_sha1_prefix) != 40 ||
-		    strlen(patch->new_sha1_prefix) != 40 ||
-		    get_sha1_hex(patch->old_sha1_prefix, sha1) ||
-		    get_sha1_hex(patch->new_sha1_prefix, sha1))
-			return error("cannot apply binary patch to '%s' "
-				     "without full index line", name);
-
-		if (patch->old_name) {
-			unsigned char hdr[50];
-			int hdrlen;
-
-			/* See if the old one matches what the patch
-			 * applies to.
-			 */
-			write_sha1_file_prepare(desc->buffer, desc->size,
-						blob_type, sha1, hdr, &hdrlen);
-			if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
-				return error("the patch applies to '%s' (%s), "
-					     "which does not match the "
-					     "current contents.",
-					     name, sha1_to_hex(sha1));
-		}
-		else {
-			/* Otherwise, the old one must be empty. */
-			if (desc->size)
-				return error("the patch applies to an empty "
-					     "'%s' but it is not empty", name);
-		}
+		write_sha1_file_prepare(desc->buffer, desc->size,
+					blob_type, sha1, hdr, &hdrlen);
+		if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
+			return error("the patch applies to '%s' (%s), "
+				     "which does not match the "
+				     "current contents.",
+				     name, sha1_to_hex(sha1));
+	}
+	else {
+		/* Otherwise, the old one must be empty. */
+		if (desc->size)
+			return error("the patch applies to an empty "
+				     "'%s' but it is not empty", name);
+	}
+
+	get_sha1_hex(patch->new_sha1_prefix, sha1);
+	if (!memcmp(sha1, null_sha1, 20)) {
+		free(desc->buffer);
+		desc->alloc = desc->size = 0;
+		desc->buffer = NULL;
+		return 0; /* deletion patch */
+	}
 
-		/* For now, we do not record post-image data in the patch,
-		 * and require the object already present in the recipient's
-		 * object database.
+	if (has_sha1_file(sha1)) {
+		/* We already have the postimage */
+		char type[10];
+		unsigned long size;
+
+		free(desc->buffer);
+		desc->buffer = read_sha1_file(sha1, type, &size);
+		if (!desc->buffer)
+			return error("the necessary postimage %s for "
+				     "'%s' cannot be read",
+				     patch->new_sha1_prefix, name);
+		desc->alloc = desc->size = size;
+	}
+	else {
+		/* We have verified desc matches the preimage;
+		 * apply the patch data to it, which is stored
+		 * in the patch->fragments->{patch,size}.
 		 */
-		if (desc->buffer) {
-			free(desc->buffer);
-			desc->alloc = desc->size = 0;
-		}
-		get_sha1_hex(patch->new_sha1_prefix, sha1);
-
-		if (memcmp(sha1, null_sha1, 20)) {
-			char type[10];
-			unsigned long size;
-
-			desc->buffer = read_sha1_file(sha1, type, &size);
-			if (!desc->buffer)
-				return error("the necessary postimage %s for "
-					     "'%s' does not exist",
-					     patch->new_sha1_prefix, name);
-			desc->alloc = desc->size = size;
-		}
+		if (apply_binary_fragment(desc, patch))
+			return error("binary patch does not apply to '%s'",
+				     name);
 
-		return 0;
+		/* verify that the result matches */
+		write_sha1_file_prepare(desc->buffer, desc->size, blob_type,
+					sha1, hdr, &hdrlen);
+		if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
+			return error("binary patch to '%s' creates incorrect result", name);
 	}
 
+	return 0;
+}
+
+static int apply_fragments(struct buffer_desc *desc, struct patch *patch)
+{
+	struct fragment *frag = patch->fragments;
+	const char *name = patch->old_name ? patch->old_name : patch->new_name;
+
+	if (patch->is_binary)
+		return apply_binary(desc, patch);
+
 	while (frag) {
 		if (apply_one_fragment(desc, frag) < 0)
 			return error("patch failed: %s:%ld",
@@ -1717,6 +1893,7 @@ static void remove_file(struct patch *patch)
 	if (write_index) {
 		if (remove_file_from_cache(patch->old_name) < 0)
 			die("unable to remove %s from index", patch->old_name);
+		cache_tree_invalidate_path(active_cache_tree, patch->old_name);
 	}
 	unlink(patch->old_name);
 }
@@ -1813,8 +1990,9 @@ static void create_file(struct patch *patch)
 
 	if (!mode)
 		mode = S_IFREG | 0644;
-	create_one_file(path, mode, buf, size);	
+	create_one_file(path, mode, buf, size);
 	add_index_file(path, mode, buf, size);
+	cache_tree_invalidate_path(active_cache_tree, path);
 }
 
 static void write_out_one_result(struct patch *patch)
@@ -1990,7 +2168,8 @@ int main(int argc, char **argv)
 			diffstat = 1;
 			continue;
 		}
-		if (!strcmp(arg, "--allow-binary-replacement")) {
+		if (!strcmp(arg, "--allow-binary-replacement") ||
+		    !strcmp(arg, "--binary")) {
 			allow_binary_replacement = 1;
 			continue;
 		}
diff --git a/base85.c b/base85.c
new file mode 100644
index 0000000000..b97f7f933a
--- /dev/null
+++ b/base85.c
@@ -0,0 +1,134 @@
+#include "cache.h"
+
+#undef DEBUG_85
+
+#ifdef DEBUG_85
+#define say(a) fprintf(stderr, a)
+#define say1(a,b) fprintf(stderr, a, b)
+#define say2(a,b,c) fprintf(stderr, a, b, c)
+#else
+#define say(a) do {} while(0)
+#define say1(a,b) do {} while(0)
+#define say2(a,b,c) do {} while(0)
+#endif
+
+static const char en85[] = {
+	'0', '1', '2', '3', '4', '5', '6', '7', '8', '9',
+	'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J',
+	'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T',
+	'U', 'V', 'W', 'X', 'Y', 'Z',
+	'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j',
+	'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't',
+	'u', 'v', 'w', 'x', 'y', 'z',
+	'!', '#', '$', '%', '&', '(', ')', '*', '+', '-',
+	';', '<', '=', '>', '?', '@', '^', '_',	'`', '{',
+	'|', '}', '~'
+};
+
+static char de85[256];
+static void prep_base85(void)
+{
+	int i;
+	if (de85['Z'])
+		return;
+	for (i = 0; i < ARRAY_SIZE(en85); i++) {
+		int ch = en85[i];
+		de85[ch] = i + 1;
+	}
+}
+
+int decode_85(char *dst, char *buffer, int len)
+{
+	prep_base85();
+
+	say2("decode 85 <%.*s>", len/4*5, buffer);
+	while (len) {
+		unsigned acc = 0;
+		int cnt;
+		for (cnt = 0; cnt < 5; cnt++, buffer++) {
+			int ch = *((unsigned char *)buffer);
+			int de = de85[ch];
+			if (!de)
+				return error("invalid base85 alphabet %c", ch);
+			de--;
+			if (cnt == 4) {
+				/*
+				 * Detect overflow.  The largest
+				 * 5-letter possible is "|NsC0" to
+				 * encode 0xffffffff, and "|NsC" gives
+				 * 0x03030303 at this point (i.e.
+				 * 0xffffffff = 0x03030303 * 85).
+				 */
+				if (0x03030303 < acc ||
+				    (0x03030303 == acc && de))
+					error("invalid base85 sequence %.5s",
+					      buffer-3);
+			}
+			acc = acc * 85 + de;
+			say1(" <%08x>", acc);
+		}
+		say1(" %08x", acc);
+		for (cnt = 0; cnt < 4 && len; cnt++, len--) {
+			*dst++ = (acc >> 24) & 0xff;
+			acc = acc << 8;
+		}
+	}
+	say("\n");
+
+	return 0;
+}
+
+void encode_85(char *buf, unsigned char *data, int bytes)
+{
+	prep_base85();
+
+	say("encode 85");
+	while (bytes) {
+		unsigned acc = 0;
+		int cnt;
+		for (cnt = 0; cnt < 4 && bytes; cnt++, bytes--) {
+			int ch = *data++;
+			acc |= ch << ((3-cnt)*8);
+		}
+		say1(" %08x", acc);
+		for (cnt = 0; cnt < 5; cnt++) {
+			int val = acc % 85;
+			acc /= 85;
+			buf[4-cnt] = en85[val];
+		}
+		buf += 5;
+	}
+	say("\n");
+
+	*buf = 0;
+}
+
+#ifdef DEBUG_85
+int main(int ac, char **av)
+{
+	char buf[1024];
+
+	if (!strcmp(av[1], "-e")) {
+		int len = strlen(av[2]);
+		encode_85(buf, av[2], len);
+		if (len <= 26) len = len + 'A' - 1;
+		else len = len + 'a' - 26 + 1;
+		printf("encoded: %c%s\n", len, buf);
+		return 0;
+	}
+	if (!strcmp(av[1], "-d")) {
+		int len = *av[2];
+		if ('A' <= len && len <= 'Z') len = len - 'A' + 1;
+		else len = len - 'a' + 26 + 1;
+		decode_85(buf, av[2]+1, len);
+		printf("decoded: %.*s\n", len, buf);
+		return 0;
+	}
+	if (!strcmp(av[1], "-t")) {
+		char t[4] = { -1,-1,-1,-1 };
+		encode_85(buf, t, 4);
+		printf("encoded: D%s\n", buf);
+		return 0;
+	}
+}
+#endif
diff --git a/builtin-grep.c b/builtin-grep.c
new file mode 100644
index 0000000000..c89ee33a1e
--- /dev/null
+++ b/builtin-grep.c
@@ -0,0 +1,682 @@
+/*
+ * Builtin "git grep"
+ *
+ * Copyright (c) 2006 Junio C Hamano
+ */
+#include "cache.h"
+#include "blob.h"
+#include "tree.h"
+#include "commit.h"
+#include "tag.h"
+#include "tree-walk.h"
+#include "builtin.h"
+#include <regex.h>
+#include <fnmatch.h>
+
+/*
+ * git grep pathspecs are somewhat different from diff-tree pathspecs;
+ * pathname wildcards are allowed.
+ */
+static int pathspec_matches(const char **paths, const char *name)
+{
+	int namelen, i;
+	if (!paths || !*paths)
+		return 1;
+	namelen = strlen(name);
+	for (i = 0; paths[i]; i++) {
+		const char *match = paths[i];
+		int matchlen = strlen(match);
+		const char *cp, *meta;
+
+		if ((matchlen <= namelen) &&
+		    !strncmp(name, match, matchlen) &&
+		    (match[matchlen-1] == '/' ||
+		     name[matchlen] == '\0' || name[matchlen] == '/'))
+			return 1;
+		if (!fnmatch(match, name, 0))
+			return 1;
+		if (name[namelen-1] != '/')
+			continue;
+
+		/* We are being asked if the directory ("name") is worth
+		 * descending into.
+		 *
+		 * Find the longest leading directory name that does
+		 * not have metacharacter in the pathspec; the name
+		 * we are looking at must overlap with that directory.
+		 */
+		for (cp = match, meta = NULL; cp - match < matchlen; cp++) {
+			char ch = *cp;
+			if (ch == '*' || ch == '[' || ch == '?') {
+				meta = cp;
+				break;
+			}
+		}
+		if (!meta)
+			meta = cp; /* fully literal */
+
+		if (namelen <= meta - match) {
+			/* Looking at "Documentation/" and
+			 * the pattern says "Documentation/howto/", or
+			 * "Documentation/diff*.txt".  The name we
+			 * have should match prefix.
+			 */
+			if (!memcmp(match, name, namelen))
+				return 1;
+			continue;
+		}
+
+		if (meta - match < namelen) {
+			/* Looking at "Documentation/howto/" and
+			 * the pattern says "Documentation/h*";
+			 * match up to "Do.../h"; this avoids descending
+			 * into "Documentation/technical/".
+			 */
+			if (!memcmp(match, name, meta - match))
+				return 1;
+			continue;
+		}
+	}
+	return 0;
+}
+
+struct grep_pat {
+	struct grep_pat *next;
+	const char *pattern;
+	regex_t regexp;
+};
+
+struct grep_opt {
+	struct grep_pat *pattern_list;
+	struct grep_pat **pattern_tail;
+	regex_t regexp;
+	unsigned linenum:1;
+	unsigned invert:1;
+	unsigned name_only:1;
+	unsigned unmatch_name_only:1;
+	unsigned count:1;
+	unsigned word_regexp:1;
+#define GREP_BINARY_DEFAULT	0
+#define GREP_BINARY_NOMATCH	1
+#define GREP_BINARY_TEXT	2
+	unsigned binary:2;
+	int regflags;
+	unsigned pre_context;
+	unsigned post_context;
+};
+
+static void add_pattern(struct grep_opt *opt, const char *pat)
+{
+	struct grep_pat *p = xcalloc(1, sizeof(*p));
+	p->pattern = pat;
+	*opt->pattern_tail = p;
+	opt->pattern_tail = &p->next;
+	p->next = NULL;
+}
+
+static void compile_patterns(struct grep_opt *opt)
+{
+	struct grep_pat *p;
+	for (p = opt->pattern_list; p; p = p->next) {
+		int err = regcomp(&p->regexp, p->pattern, opt->regflags);
+		if (err) {
+			char errbuf[1024];
+			regerror(err, &p->regexp, errbuf, 1024);
+			regfree(&p->regexp);
+			die("'%s': %s", p->pattern, errbuf);
+		}
+	}
+}
+
+static char *end_of_line(char *cp, unsigned long *left)
+{
+	unsigned long l = *left;
+	while (l && *cp != '\n') {
+		l--;
+		cp++;
+	}
+	*left = l;
+	return cp;
+}
+
+static int word_char(char ch)
+{
+	return isalnum(ch) || ch == '_';
+}
+
+static void show_line(struct grep_opt *opt, const char *bol, const char *eol,
+		      const char *name, unsigned lno, char sign)
+{
+	printf("%s%c", name, sign);
+	if (opt->linenum)
+		printf("%d%c", lno, sign);
+	printf("%.*s\n", (int)(eol-bol), bol);
+}
+
+/*
+ * NEEDSWORK: share code with diff.c
+ */
+#define FIRST_FEW_BYTES 8000
+static int buffer_is_binary(const char *ptr, unsigned long size)
+{
+	if (FIRST_FEW_BYTES < size)
+		size = FIRST_FEW_BYTES;
+	if (memchr(ptr, 0, size))
+		return 1;
+	return 0;
+}
+
+static int grep_buffer(struct grep_opt *opt, const char *name,
+		       char *buf, unsigned long size)
+{
+	char *bol = buf;
+	unsigned long left = size;
+	unsigned lno = 1;
+	struct pre_context_line {
+		char *bol;
+		char *eol;
+	} *prev = NULL, *pcl;
+	unsigned last_hit = 0;
+	unsigned last_shown = 0;
+	int binary_match_only = 0;
+	const char *hunk_mark = "";
+	unsigned count = 0;
+
+	if (buffer_is_binary(buf, size)) {
+		switch (opt->binary) {
+		case GREP_BINARY_DEFAULT:
+			binary_match_only = 1;
+			break;
+		case GREP_BINARY_NOMATCH:
+			return 0; /* Assume unmatch */
+			break;
+		default:
+			break;
+		}
+	}
+
+	if (opt->pre_context)
+		prev = xcalloc(opt->pre_context, sizeof(*prev));
+	if (opt->pre_context || opt->post_context)
+		hunk_mark = "--\n";
+
+	while (left) {
+		regmatch_t pmatch[10];
+		char *eol, ch;
+		int hit = 0;
+		struct grep_pat *p;
+
+		eol = end_of_line(bol, &left);
+		ch = *eol;
+		*eol = 0;
+
+		for (p = opt->pattern_list; p; p = p->next) {
+			regex_t *exp = &p->regexp;
+			hit = !regexec(exp, bol, ARRAY_SIZE(pmatch),
+				       pmatch, 0);
+
+			if (hit && opt->word_regexp) {
+				/* Match beginning must be either
+				 * beginning of the line, or at word
+				 * boundary (i.e. the last char must
+				 * not be alnum or underscore).
+				 */
+				if ((pmatch[0].rm_so < 0) ||
+				    (eol - bol) <= pmatch[0].rm_so ||
+				    (pmatch[0].rm_eo < 0) ||
+				    (eol - bol) < pmatch[0].rm_eo)
+					die("regexp returned nonsense");
+				if (pmatch[0].rm_so != 0 &&
+				    word_char(bol[pmatch[0].rm_so-1]))
+					continue; /* not a word boundary */
+				if ((eol-bol) < pmatch[0].rm_eo &&
+				    word_char(bol[pmatch[0].rm_eo]))
+					continue; /* not a word boundary */
+			}
+			if (hit)
+				break;
+		}
+		/* "grep -v -e foo -e bla" should list lines
+		 * that do not have either, so inversion should
+		 * be done outside.
+		 */
+		if (opt->invert)
+			hit = !hit;
+		if (opt->unmatch_name_only) {
+			if (hit)
+				return 0;
+			goto next_line;
+		}
+		if (hit) {
+			count++;
+			if (binary_match_only) {
+				printf("Binary file %s matches\n", name);
+				return 1;
+			}
+			if (opt->name_only) {
+				printf("%s\n", name);
+				return 1;
+			}
+			/* Hit at this line.  If we haven't shown the
+			 * pre-context lines, we would need to show them.
+			 * When asked to do "count", this still show
+			 * the context which is nonsense, but the user
+			 * deserves to get that ;-).
+			 */
+			if (opt->pre_context) {
+				unsigned from;
+				if (opt->pre_context < lno)
+					from = lno - opt->pre_context;
+				else
+					from = 1;
+				if (from <= last_shown)
+					from = last_shown + 1;
+				if (last_shown && from != last_shown + 1)
+					printf(hunk_mark);
+				while (from < lno) {
+					pcl = &prev[lno-from-1];
+					show_line(opt, pcl->bol, pcl->eol,
+						  name, from, '-');
+					from++;
+				}
+				last_shown = lno-1;
+			}
+			if (last_shown && lno != last_shown + 1)
+				printf(hunk_mark);
+			if (!opt->count)
+				show_line(opt, bol, eol, name, lno, ':');
+			last_shown = last_hit = lno;
+		}
+		else if (last_hit &&
+			 lno <= last_hit + opt->post_context) {
+			/* If the last hit is within the post context,
+			 * we need to show this line.
+			 */
+			if (last_shown && lno != last_shown + 1)
+				printf(hunk_mark);
+			show_line(opt, bol, eol, name, lno, '-');
+			last_shown = lno;
+		}
+		if (opt->pre_context) {
+			memmove(prev+1, prev,
+				(opt->pre_context-1) * sizeof(*prev));
+			prev->bol = bol;
+			prev->eol = eol;
+		}
+
+	next_line:
+		*eol = ch;
+		bol = eol + 1;
+		if (!left)
+			break;
+		left--;
+		lno++;
+	}
+
+	if (opt->unmatch_name_only) {
+		/* We did not see any hit, so we want to show this */
+		printf("%s\n", name);
+		return 1;
+	}
+
+	/* NEEDSWORK:
+	 * The real "grep -c foo *.c" gives many "bar.c:0" lines,
+	 * which feels mostly useless but sometimes useful.  Maybe
+	 * make it another option?  For now suppress them.
+	 */
+	if (opt->count && count)
+		printf("%s:%u\n", name, count);
+	return !!last_hit;
+}
+
+static int grep_sha1(struct grep_opt *opt, const unsigned char *sha1, const char *name)
+{
+	unsigned long size;
+	char *data;
+	char type[20];
+	int hit;
+	data = read_sha1_file(sha1, type, &size);
+	if (!data) {
+		error("'%s': unable to read %s", name, sha1_to_hex(sha1));
+		return 0;
+	}
+	hit = grep_buffer(opt, name, data, size);
+	free(data);
+	return hit;
+}
+
+static int grep_file(struct grep_opt *opt, const char *filename)
+{
+	struct stat st;
+	int i;
+	char *data;
+	if (lstat(filename, &st) < 0) {
+	err_ret:
+		if (errno != ENOENT)
+			error("'%s': %s", filename, strerror(errno));
+		return 0;
+	}
+	if (!st.st_size)
+		return 0; /* empty file -- no grep hit */
+	if (!S_ISREG(st.st_mode))
+		return 0;
+	i = open(filename, O_RDONLY);
+	if (i < 0)
+		goto err_ret;
+	data = xmalloc(st.st_size + 1);
+	if (st.st_size != xread(i, data, st.st_size)) {
+		error("'%s': short read %s", filename, strerror(errno));
+		close(i);
+		free(data);
+		return 0;
+	}
+	close(i);
+	i = grep_buffer(opt, filename, data, st.st_size);
+	free(data);
+	return i;
+}
+
+static int grep_cache(struct grep_opt *opt, const char **paths, int cached)
+{
+	int hit = 0;
+	int nr;
+	read_cache();
+
+	for (nr = 0; nr < active_nr; nr++) {
+		struct cache_entry *ce = active_cache[nr];
+		if (ce_stage(ce) || !S_ISREG(ntohl(ce->ce_mode)))
+			continue;
+		if (!pathspec_matches(paths, ce->name))
+			continue;
+		if (cached)
+			hit |= grep_sha1(opt, ce->sha1, ce->name);
+		else
+			hit |= grep_file(opt, ce->name);
+	}
+	return hit;
+}
+
+static int grep_tree(struct grep_opt *opt, const char **paths,
+		     struct tree_desc *tree,
+		     const char *tree_name, const char *base)
+{
+	unsigned mode;
+	int len;
+	int hit = 0;
+	const char *path;
+	const unsigned char *sha1;
+	char *down;
+	char *path_buf = xmalloc(PATH_MAX + strlen(tree_name) + 100);
+
+	if (tree_name[0]) {
+		int offset = sprintf(path_buf, "%s:", tree_name);
+		down = path_buf + offset;
+		strcat(down, base);
+	}
+	else {
+		down = path_buf;
+		strcpy(down, base);
+	}
+	len = strlen(path_buf);
+
+	while (tree->size) {
+		int pathlen;
+		sha1 = tree_entry_extract(tree, &path, &mode);
+		pathlen = strlen(path);
+		strcpy(path_buf + len, path);
+
+		if (S_ISDIR(mode))
+			/* Match "abc/" against pathspec to
+			 * decide if we want to descend into "abc"
+			 * directory.
+			 */
+			strcpy(path_buf + len + pathlen, "/");
+
+		if (!pathspec_matches(paths, down))
+			;
+		else if (S_ISREG(mode))
+			hit |= grep_sha1(opt, sha1, path_buf);
+		else if (S_ISDIR(mode)) {
+			char type[20];
+			struct tree_desc sub;
+			void *data;
+			data = read_sha1_file(sha1, type, &sub.size);
+			if (!data)
+				die("unable to read tree (%s)",
+				    sha1_to_hex(sha1));
+			sub.buf = data;
+			hit |= grep_tree(opt, paths, &sub, tree_name, down);
+			free(data);
+		}
+		update_tree_entry(tree);
+	}
+	return hit;
+}
+
+static int grep_object(struct grep_opt *opt, const char **paths,
+		       struct object *obj, const char *name)
+{
+	if (!strcmp(obj->type, blob_type))
+		return grep_sha1(opt, obj->sha1, name);
+	if (!strcmp(obj->type, commit_type) ||
+	    !strcmp(obj->type, tree_type)) {
+		struct tree_desc tree;
+		void *data;
+		int hit;
+		data = read_object_with_reference(obj->sha1, tree_type,
+						  &tree.size, NULL);
+		if (!data)
+			die("unable to read tree (%s)", sha1_to_hex(obj->sha1));
+		tree.buf = data;
+		hit = grep_tree(opt, paths, &tree, name, "");
+		free(data);
+		return hit;
+	}
+	die("unable to grep from object of type %s", obj->type);
+}
+
+static const char builtin_grep_usage[] =
+"git-grep <option>* <rev>* [-e] <pattern> [<path>...]";
+
+int cmd_grep(int argc, const char **argv, char **envp)
+{
+	int hit = 0;
+	int no_more_flags = 0;
+	int seen_noncommit = 0;
+	int cached = 0;
+	struct grep_opt opt;
+	struct object_list *list, **tail, *object_list = NULL;
+	const char *prefix = setup_git_directory();
+	const char **paths = NULL;
+
+	memset(&opt, 0, sizeof(opt));
+	opt.pattern_tail = &opt.pattern_list;
+	opt.regflags = REG_NEWLINE;
+
+	/*
+	 * No point using rev_info, really.
+	 */
+	while (1 < argc) {
+		const char *arg = argv[1];
+		argc--; argv++;
+		if (!strcmp("--cached", arg)) {
+			cached = 1;
+			continue;
+		}
+		if (!strcmp("-a", arg) ||
+		    !strcmp("--text", arg)) {
+			opt.binary = GREP_BINARY_TEXT;
+			continue;
+		}
+		if (!strcmp("-i", arg) ||
+		    !strcmp("--ignore-case", arg)) {
+			opt.regflags |= REG_ICASE;
+			continue;
+		}
+		if (!strcmp("-I", arg)) {
+			opt.binary = GREP_BINARY_NOMATCH;
+			continue;
+		}
+		if (!strcmp("-v", arg) ||
+		    !strcmp("--invert-match", arg)) {
+			opt.invert = 1;
+			continue;
+		}
+		if (!strcmp("-E", arg) ||
+		    !strcmp("--extended-regexp", arg)) {
+			opt.regflags |= REG_EXTENDED;
+			continue;
+		}
+		if (!strcmp("-G", arg) ||
+		    !strcmp("--basic-regexp", arg)) {
+			opt.regflags &= ~REG_EXTENDED;
+			continue;
+		}
+		if (!strcmp("-n", arg)) {
+			opt.linenum = 1;
+			continue;
+		}
+		if (!strcmp("-H", arg)) {
+			/* We always show the pathname, so this
+			 * is a noop.
+			 */
+			continue;
+		}
+		if (!strcmp("-l", arg) ||
+		    !strcmp("--files-with-matches", arg)) {
+			opt.name_only = 1;
+			continue;
+		}
+		if (!strcmp("-L", arg) ||
+		    !strcmp("--files-without-match", arg)) {
+			opt.unmatch_name_only = 1;
+			continue;
+		}
+		if (!strcmp("-c", arg) ||
+		    !strcmp("--count", arg)) {
+			opt.count = 1;
+			continue;
+		}
+		if (!strcmp("-w", arg) ||
+		    !strcmp("--word-regexp", arg)) {
+			opt.word_regexp = 1;
+			continue;
+		}
+		if (!strncmp("-A", arg, 2) ||
+		    !strncmp("-B", arg, 2) ||
+		    !strncmp("-C", arg, 2) ||
+		    (arg[0] == '-' && '1' <= arg[1] && arg[1] <= '9')) {
+			unsigned num;
+			const char *scan;
+			switch (arg[1]) {
+			case 'A': case 'B': case 'C':
+				if (!arg[2]) {
+					if (argc <= 1)
+						usage(builtin_grep_usage);
+					scan = *++argv;
+					argc--;
+				}
+				else
+					scan = arg + 2;
+				break;
+			default:
+				scan = arg + 1;
+				break;
+			}
+			if (sscanf(scan, "%u", &num) != 1)
+				usage(builtin_grep_usage);
+			switch (arg[1]) {
+			case 'A':
+				opt.post_context = num;
+				break;
+			default:
+			case 'C':
+				opt.post_context = num;
+			case 'B':
+				opt.pre_context = num;
+				break;
+			}
+			continue;
+		}
+		if (!strcmp("-e", arg)) {
+			if (1 < argc) {
+				add_pattern(&opt, argv[1]);
+				argv++;
+				argc--;
+				continue;
+			}
+			usage(builtin_grep_usage);
+		}
+		if (!strcmp("--", arg)) {
+			no_more_flags = 1;
+			continue;
+		}
+		/* Either unrecognized option or a single pattern */
+		if (!no_more_flags && *arg == '-')
+			usage(builtin_grep_usage);
+		if (!opt.pattern_list) {
+			add_pattern(&opt, arg);
+			break;
+		}
+		else {
+			/* We are looking at the first path or rev;
+			 * it is found at argv[0] after leaving the
+			 * loop.
+			 */
+			argc++; argv--;
+			break;
+		}
+	}
+	if (!opt.pattern_list)
+		die("no pattern given.");
+	compile_patterns(&opt);
+	tail = &object_list;
+	while (1 < argc) {
+		struct object *object;
+		struct object_list *elem;
+		const char *arg = argv[1];
+		unsigned char sha1[20];
+		if (get_sha1(arg, sha1) < 0)
+			break;
+		object = parse_object(sha1);
+		if (!object)
+			die("bad object %s", arg);
+		elem = object_list_insert(object, tail);
+		elem->name = arg;
+		tail = &elem->next;
+		argc--; argv++;
+	}
+	if (1 < argc)
+		paths = get_pathspec(prefix, argv + 1);
+	else if (prefix) {
+		paths = xcalloc(2, sizeof(const char *));
+		paths[0] = prefix;
+		paths[1] = NULL;
+	}
+
+	if (!object_list)
+		return !grep_cache(&opt, paths, cached);
+	/*
+	 * Do not walk "grep -e foo master next pu -- Documentation/"
+	 * but do walk "grep -e foo master..next -- Documentation/".
+	 * Ranged request mixed with a blob or tree object, like
+	 * "grep -e foo v1.0.0:Documentation/ master..next"
+	 * so detect that and complain.
+	 */
+	for (list = object_list; list; list = list->next) {
+		struct object *real_obj;
+		real_obj = deref_tag(list->item, NULL, 0);
+		if (strcmp(real_obj->type, commit_type))
+			seen_noncommit = 1;
+	}
+	if (cached)
+		die("both --cached and revisions given.");
+
+	for (list = object_list; list; list = list->next) {
+		struct object *real_obj;
+		real_obj = deref_tag(list->item, NULL, 0);
+		if (grep_object(&opt, paths, real_obj, list->name))
+			hit = 1;
+	}
+	return !hit;
+}
diff --git a/builtin.h b/builtin.h
index 635a00bf7a..97e2464468 100644
--- a/builtin.h
+++ b/builtin.h
@@ -24,5 +24,6 @@ extern int cmd_format_patch(int argc, const char **argv, char **envp);
 extern int cmd_count_objects(int argc, const char **argv, char **envp);
 
 extern int cmd_push(int argc, const char **argv, char **envp);
+extern int cmd_grep(int argc, const char **argv, char **envp);
 
 #endif
diff --git a/cache-tree.c b/cache-tree.c
new file mode 100644
index 0000000000..a880c97b38
--- /dev/null
+++ b/cache-tree.c
@@ -0,0 +1,531 @@
+#include "cache.h"
+#include "tree.h"
+#include "cache-tree.h"
+
+#define DEBUG 0
+
+struct cache_tree *cache_tree(void)
+{
+	struct cache_tree *it = xcalloc(1, sizeof(struct cache_tree));
+	it->entry_count = -1;
+	return it;
+}
+
+void cache_tree_free(struct cache_tree **it_p)
+{
+	int i;
+	struct cache_tree *it = *it_p;
+
+	if (!it)
+		return;
+	for (i = 0; i < it->subtree_nr; i++)
+		if (it->down[i])
+			cache_tree_free(&it->down[i]->cache_tree);
+	free(it->down);
+	free(it);
+	*it_p = NULL;
+}
+
+static int subtree_name_cmp(const char *one, int onelen,
+			    const char *two, int twolen)
+{
+	if (onelen < twolen)
+		return -1;
+	if (twolen < onelen)
+		return 1;
+	return memcmp(one, two, onelen);
+}
+
+static int subtree_pos(struct cache_tree *it, const char *path, int pathlen)
+{
+	struct cache_tree_sub **down = it->down;
+	int lo, hi;
+	lo = 0;
+	hi = it->subtree_nr;
+	while (lo < hi) {
+		int mi = (lo + hi) / 2;
+		struct cache_tree_sub *mdl = down[mi];
+		int cmp = subtree_name_cmp(path, pathlen,
+					   mdl->name, mdl->namelen);
+		if (!cmp)
+			return mi;
+		if (cmp < 0)
+			hi = mi;
+		else
+			lo = mi + 1;
+	}
+	return -lo-1;
+}
+
+static struct cache_tree_sub *find_subtree(struct cache_tree *it,
+					   const char *path,
+					   int pathlen,
+					   int create)
+{
+	struct cache_tree_sub *down;
+	int pos = subtree_pos(it, path, pathlen);
+	if (0 <= pos)
+		return it->down[pos];
+	if (!create)
+		return NULL;
+
+	pos = -pos-1;
+	if (it->subtree_alloc <= it->subtree_nr) {
+		it->subtree_alloc = alloc_nr(it->subtree_alloc);
+		it->down = xrealloc(it->down, it->subtree_alloc *
+				    sizeof(*it->down));
+	}
+	it->subtree_nr++;
+
+	down = xmalloc(sizeof(*down) + pathlen + 1);
+	down->cache_tree = NULL;
+	down->namelen = pathlen;
+	memcpy(down->name, path, pathlen);
+	down->name[pathlen] = 0;
+
+	if (pos < it->subtree_nr)
+		memmove(it->down + pos + 1,
+			it->down + pos,
+			sizeof(down) * (it->subtree_nr - pos - 1));
+	it->down[pos] = down;
+	return down;
+}
+
+struct cache_tree_sub *cache_tree_sub(struct cache_tree *it, const char *path)
+{
+	int pathlen = strlen(path);
+	return find_subtree(it, path, pathlen, 1);
+}
+
+void cache_tree_invalidate_path(struct cache_tree *it, const char *path)
+{
+	/* a/b/c
+	 * ==> invalidate self
+	 * ==> find "a", have it invalidate "b/c"
+	 * a
+	 * ==> invalidate self
+	 * ==> if "a" exists as a subtree, remove it.
+	 */
+	const char *slash;
+	int namelen;
+	struct cache_tree_sub *down;
+
+#if DEBUG
+	fprintf(stderr, "cache-tree invalidate <%s>\n", path);
+#endif
+
+	if (!it)
+		return;
+	slash = strchr(path, '/');
+	it->entry_count = -1;
+	if (!slash) {
+		int pos;
+		namelen = strlen(path);
+		pos = subtree_pos(it, path, namelen);
+		if (0 <= pos) {
+			cache_tree_free(&it->down[pos]->cache_tree);
+			free(it->down[pos]);
+			/* 0 1 2 3 4 5
+			 *       ^     ^subtree_nr = 6
+			 *       pos
+			 * move 4 and 5 up one place (2 entries)
+			 * 2 = 6 - 3 - 1 = subtree_nr - pos - 1
+			 */
+			memmove(it->down+pos, it->down+pos+1,
+				sizeof(struct cache_tree_sub *) *
+				(it->subtree_nr - pos - 1));
+			it->subtree_nr--;
+		}
+		return;
+	}
+	namelen = slash - path;
+	down = find_subtree(it, path, namelen, 0);
+	if (down)
+		cache_tree_invalidate_path(down->cache_tree, slash + 1);
+}
+
+static int verify_cache(struct cache_entry **cache,
+			int entries)
+{
+	int i, funny;
+
+	/* Verify that the tree is merged */
+	funny = 0;
+	for (i = 0; i < entries; i++) {
+		struct cache_entry *ce = cache[i];
+		if (ce_stage(ce)) {
+			if (10 < ++funny) {
+				fprintf(stderr, "...\n");
+				break;
+			}
+			fprintf(stderr, "%s: unmerged (%s)\n",
+				ce->name, sha1_to_hex(ce->sha1));
+		}
+	}
+	if (funny)
+		return -1;
+
+	/* Also verify that the cache does not have path and path/file
+	 * at the same time.  At this point we know the cache has only
+	 * stage 0 entries.
+	 */
+	funny = 0;
+	for (i = 0; i < entries - 1; i++) {
+		/* path/file always comes after path because of the way
+		 * the cache is sorted.  Also path can appear only once,
+		 * which means conflicting one would immediately follow.
+		 */
+		const char *this_name = cache[i]->name;
+		const char *next_name = cache[i+1]->name;
+		int this_len = strlen(this_name);
+		if (this_len < strlen(next_name) &&
+		    strncmp(this_name, next_name, this_len) == 0 &&
+		    next_name[this_len] == '/') {
+			if (10 < ++funny) {
+				fprintf(stderr, "...\n");
+				break;
+			}
+			fprintf(stderr, "You have both %s and %s\n",
+				this_name, next_name);
+		}
+	}
+	if (funny)
+		return -1;
+	return 0;
+}
+
+static void discard_unused_subtrees(struct cache_tree *it)
+{
+	struct cache_tree_sub **down = it->down;
+	int nr = it->subtree_nr;
+	int dst, src;
+	for (dst = src = 0; src < nr; src++) {
+		struct cache_tree_sub *s = down[src];
+		if (s->used)
+			down[dst++] = s;
+		else {
+			cache_tree_free(&s->cache_tree);
+			free(s);
+			it->subtree_nr--;
+		}
+	}
+}
+
+int cache_tree_fully_valid(struct cache_tree *it)
+{
+	int i;
+	if (!it)
+		return 0;
+	if (it->entry_count < 0 || !has_sha1_file(it->sha1))
+		return 0;
+	for (i = 0; i < it->subtree_nr; i++) {
+		if (!cache_tree_fully_valid(it->down[i]->cache_tree))
+			return 0;
+	}
+	return 1;
+}
+
+static int update_one(struct cache_tree *it,
+		      struct cache_entry **cache,
+		      int entries,
+		      const char *base,
+		      int baselen,
+		      int missing_ok,
+		      int dryrun)
+{
+	unsigned long size, offset;
+	char *buffer;
+	int i;
+
+	if (0 <= it->entry_count && has_sha1_file(it->sha1))
+		return it->entry_count;
+
+	/*
+	 * We first scan for subtrees and update them; we start by
+	 * marking existing subtrees -- the ones that are unmarked
+	 * should not be in the result.
+	 */
+	for (i = 0; i < it->subtree_nr; i++)
+		it->down[i]->used = 0;
+
+	/*
+	 * Find the subtrees and update them.
+	 */
+	for (i = 0; i < entries; i++) {
+		struct cache_entry *ce = cache[i];
+		struct cache_tree_sub *sub;
+		const char *path, *slash;
+		int pathlen, sublen, subcnt;
+
+		path = ce->name;
+		pathlen = ce_namelen(ce);
+		if (pathlen <= baselen || memcmp(base, path, baselen))
+			break; /* at the end of this level */
+
+		slash = strchr(path + baselen, '/');
+		if (!slash)
+			continue;
+		/*
+		 * a/bbb/c (base = a/, slash = /c)
+		 * ==>
+		 * path+baselen = bbb/c, sublen = 3
+		 */
+		sublen = slash - (path + baselen);
+		sub = find_subtree(it, path + baselen, sublen, 1);
+		if (!sub->cache_tree)
+			sub->cache_tree = cache_tree();
+		subcnt = update_one(sub->cache_tree,
+				    cache + i, entries - i,
+				    path,
+				    baselen + sublen + 1,
+				    missing_ok,
+				    dryrun);
+		i += subcnt - 1;
+		sub->used = 1;
+	}
+
+	discard_unused_subtrees(it);
+
+	/*
+	 * Then write out the tree object for this level.
+	 */
+	size = 8192;
+	buffer = xmalloc(size);
+	offset = 0;
+
+	for (i = 0; i < entries; i++) {
+		struct cache_entry *ce = cache[i];
+		struct cache_tree_sub *sub;
+		const char *path, *slash;
+		int pathlen, entlen;
+		const unsigned char *sha1;
+		unsigned mode;
+
+		path = ce->name;
+		pathlen = ce_namelen(ce);
+		if (pathlen <= baselen || memcmp(base, path, baselen))
+			break; /* at the end of this level */
+
+		slash = strchr(path + baselen, '/');
+		if (slash) {
+			entlen = slash - (path + baselen);
+			sub = find_subtree(it, path + baselen, entlen, 0);
+			if (!sub)
+				die("cache-tree.c: '%.*s' in '%s' not found",
+				    entlen, path + baselen, path);
+			i += sub->cache_tree->entry_count - 1;
+			sha1 = sub->cache_tree->sha1;
+			mode = S_IFDIR;
+		}
+		else {
+			sha1 = ce->sha1;
+			mode = ntohl(ce->ce_mode);
+			entlen = pathlen - baselen;
+		}
+		if (!missing_ok && !has_sha1_file(sha1))
+			return error("invalid object %s", sha1_to_hex(sha1));
+
+		if (!ce->ce_mode)
+			continue; /* entry being removed */
+
+		if (size < offset + entlen + 100) {
+			size = alloc_nr(offset + entlen + 100);
+			buffer = xrealloc(buffer, size);
+		}
+		offset += sprintf(buffer + offset,
+				  "%o %.*s", mode, entlen, path + baselen);
+		buffer[offset++] = 0;
+		memcpy(buffer + offset, sha1, 20);
+		offset += 20;
+
+#if DEBUG
+		fprintf(stderr, "cache-tree update-one %o %.*s\n",
+			mode, entlen, path + baselen);
+#endif
+	}
+
+	if (dryrun) {
+		unsigned char hdr[200];
+		int hdrlen;
+		write_sha1_file_prepare(buffer, offset, tree_type, it->sha1,
+					hdr, &hdrlen);
+	}
+	else
+		write_sha1_file(buffer, offset, tree_type, it->sha1);
+	free(buffer);
+	it->entry_count = i;
+#if DEBUG
+	fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
+		it->entry_count, it->subtree_nr,
+		sha1_to_hex(it->sha1));
+#endif
+	return i;
+}
+
+int cache_tree_update(struct cache_tree *it,
+		      struct cache_entry **cache,
+		      int entries,
+		      int missing_ok,
+		      int dryrun)
+{
+	int i;
+	i = verify_cache(cache, entries);
+	if (i)
+		return i;
+	i = update_one(it, cache, entries, "", 0, missing_ok, dryrun);
+	if (i < 0)
+		return i;
+	return 0;
+}
+
+static void *write_one(struct cache_tree *it,
+		       char *path,
+		       int pathlen,
+		       char *buffer,
+		       unsigned long *size,
+		       unsigned long *offset)
+{
+	int i;
+
+	/* One "cache-tree" entry consists of the following:
+	 * path (NUL terminated)
+	 * entry_count, subtree_nr ("%d %d\n")
+	 * tree-sha1 (missing if invalid)
+	 * subtree_nr "cache-tree" entries for subtrees.
+	 */
+	if (*size < *offset + pathlen + 100) {
+		*size = alloc_nr(*offset + pathlen + 100);
+		buffer = xrealloc(buffer, *size);
+	}
+	*offset += sprintf(buffer + *offset, "%.*s%c%d %d\n",
+			   pathlen, path, 0,
+			   it->entry_count, it->subtree_nr);
+
+#if DEBUG
+	if (0 <= it->entry_count)
+		fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
+			pathlen, path, it->entry_count, it->subtree_nr,
+			sha1_to_hex(it->sha1));
+	else
+		fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
+			pathlen, path, it->subtree_nr);
+#endif
+
+	if (0 <= it->entry_count) {
+		memcpy(buffer + *offset, it->sha1, 20);
+		*offset += 20;
+	}
+	for (i = 0; i < it->subtree_nr; i++) {
+		struct cache_tree_sub *down = it->down[i];
+		if (i) {
+			struct cache_tree_sub *prev = it->down[i-1];
+			if (subtree_name_cmp(down->name, down->namelen,
+					     prev->name, prev->namelen) <= 0)
+				die("fatal - unsorted cache subtree");
+		}
+		buffer = write_one(down->cache_tree, down->name, down->namelen,
+				   buffer, size, offset);
+	}
+	return buffer;
+}
+
+void *cache_tree_write(struct cache_tree *root, unsigned long *size_p)
+{
+	char path[PATH_MAX];
+	unsigned long size = 8192;
+	char *buffer = xmalloc(size);
+
+	*size_p = 0;
+	path[0] = 0;
+	return write_one(root, path, 0, buffer, &size, size_p);
+}
+
+static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
+{
+	const char *buf = *buffer;
+	unsigned long size = *size_p;
+	const char *cp;
+	char *ep;
+	struct cache_tree *it;
+	int i, subtree_nr;
+
+	it = NULL;
+	/* skip name, but make sure name exists */
+	while (size && *buf) {
+		size--;
+		buf++;
+	}
+	if (!size)
+		goto free_return;
+	buf++; size--;
+	it = cache_tree();
+
+	cp = buf;
+	it->entry_count = strtol(cp, &ep, 10);
+	if (cp == ep)
+		goto free_return;
+	cp = ep;
+	subtree_nr = strtol(cp, &ep, 10);
+	if (cp == ep)
+		goto free_return;
+	while (size && *buf && *buf != '\n') {
+		size--;
+		buf++;
+	}
+	if (!size)
+		goto free_return;
+	buf++; size--;
+	if (0 <= it->entry_count) {
+		if (size < 20)
+			goto free_return;
+		memcpy(it->sha1, buf, 20);
+		buf += 20;
+		size -= 20;
+	}
+
+#if DEBUG
+	if (0 <= it->entry_count)
+		fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
+			*buffer, it->entry_count, subtree_nr,
+			sha1_to_hex(it->sha1));
+	else
+		fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
+			*buffer, subtree_nr);
+#endif
+
+	/*
+	 * Just a heuristic -- we do not add directories that often but
+	 * we do not want to have to extend it immediately when we do,
+	 * hence +2.
+	 */
+	it->subtree_alloc = subtree_nr + 2;
+	it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
+	for (i = 0; i < subtree_nr; i++) {
+		/* read each subtree */
+		struct cache_tree *sub;
+		struct cache_tree_sub *subtree;
+		const char *name = buf;
+
+		sub = read_one(&buf, &size);
+		if (!sub)
+			goto free_return;
+		subtree = cache_tree_sub(it, name);
+		subtree->cache_tree = sub;
+	}
+	if (subtree_nr != it->subtree_nr)
+		die("cache-tree: internal error");
+	*buffer = buf;
+	*size_p = size;
+	return it;
+
+ free_return:
+	cache_tree_free(&it);
+	return NULL;
+}
+
+struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
+{
+	if (buffer[0])
+		return NULL; /* not the whole tree */
+	return read_one(&buffer, &size);
+}
diff --git a/cache-tree.h b/cache-tree.h
new file mode 100644
index 0000000000..72c64801f5
--- /dev/null
+++ b/cache-tree.h
@@ -0,0 +1,31 @@
+#ifndef CACHE_TREE_H
+#define CACHE_TREE_H
+
+struct cache_tree;
+struct cache_tree_sub {
+	struct cache_tree *cache_tree;
+	int namelen;
+	int used;
+	char name[FLEX_ARRAY];
+};
+
+struct cache_tree {
+	int entry_count; /* negative means "invalid" */
+	unsigned char sha1[20];
+	int subtree_nr;
+	int subtree_alloc;
+	struct cache_tree_sub **down;
+};
+
+struct cache_tree *cache_tree(void);
+void cache_tree_free(struct cache_tree **);
+void cache_tree_invalidate_path(struct cache_tree *, const char *);
+struct cache_tree_sub *cache_tree_sub(struct cache_tree *, const char *);
+
+void *cache_tree_write(struct cache_tree *root, unsigned long *size_p);
+struct cache_tree *cache_tree_read(const char *buffer, unsigned long size);
+
+int cache_tree_fully_valid(struct cache_tree *);
+int cache_tree_update(struct cache_tree *, struct cache_entry **, int, int, int);
+
+#endif
diff --git a/cache.h b/cache.h
index 605b80e268..b1300cd989 100644
--- a/cache.h
+++ b/cache.h
@@ -114,6 +114,7 @@ static inline unsigned int create_ce_mode(unsigned int mode)
 
 extern struct cache_entry **active_cache;
 extern unsigned int active_nr, active_alloc, active_cache_changed;
+extern struct cache_tree *active_cache_tree;
 
 #define GIT_DIR_ENVIRONMENT "GIT_DIR"
 #define DEFAULT_GIT_DIR_ENVIRONMENT ".git"
@@ -364,4 +365,8 @@ extern int receive_keep_pack(int fd[2], const char *me, int quiet);
 /* pager.c */
 extern void setup_pager(void);
 
+/* base85 */
+int decode_85(char *dst, char *line, int linelen);
+void encode_85(char *buf, unsigned char *data, int bytes);
+
 #endif /* CACHE_H */
diff --git a/checkout-index.c b/checkout-index.c
index 0b9cabc61c..9876af6fd6 100644
--- a/checkout-index.c
+++ b/checkout-index.c
@@ -39,6 +39,7 @@
 #include "cache.h"
 #include "strbuf.h"
 #include "quote.h"
+#include "cache-tree.h"
 
 #define CHECKOUT_ALL 4
 static const char *prefix;
@@ -277,7 +278,7 @@ int main(int argc, char **argv)
 			die("git-checkout-index: don't mix '--stdin' and explicit filenames");
 		p = prefix_path(prefix, prefix_length, arg);
 		checkout_file(p);
-		if (p != arg)
+		if (p < arg || p > arg + strlen(arg))
 			free((char*)p);
 	}
 
@@ -299,7 +300,7 @@ int main(int argc, char **argv)
 				path_name = buf.buf;
 			p = prefix_path(prefix, prefix_length, path_name);
 			checkout_file(p);
-			if (p != path_name)
+			if (p < path_name || p > path_name + strlen(path_name))
 				free((char *)p);
 			if (path_name != buf.buf)
 				free(path_name);
diff --git a/contrib/remotes2config.sh b/contrib/remotes2config.sh
new file mode 100644
index 0000000000..25901e2b3b
--- /dev/null
+++ b/contrib/remotes2config.sh
@@ -0,0 +1,35 @@
+#!/bin/sh
+
+# Use this tool to rewrite your .git/remotes/ files into the config.
+
+. git-sh-setup
+
+if [ -d "$GIT_DIR"/remotes ]; then
+	echo "Rewriting $GIT_DIR/remotes" >&2
+	error=0
+	# rewrite into config
+	{
+		cd "$GIT_DIR"/remotes
+		ls | while read f; do
+			name=$(echo -n "$f" | tr -c "A-Za-z0-9" ".")
+			sed -n \
+			-e "s/^URL: \(.*\)$/remote.$name.url \1 ./p" \
+			-e "s/^Pull: \(.*\)$/remote.$name.fetch \1 ^$ /p" \
+			-e "s/^Push: \(.*\)$/remote.$name.push \1 ^$ /p" \
+			< "$f"
+		done
+		echo done
+	} | while read key value regex; do
+		case $key in
+		done)
+			if [ $error = 0 ]; then
+				mv "$GIT_DIR"/remotes "$GIT_DIR"/remotes.old
+			fi ;;
+		*)
+			echo "git-repo-config $key "$value" $regex"
+			git-repo-config $key "$value" $regex || error=1 ;;
+		esac
+	done
+fi
+
+
diff --git a/delta.h b/delta.h
index 09b2e1eed5..727ae30e9e 100644
--- a/delta.h
+++ b/delta.h
@@ -1,11 +1,72 @@
 #ifndef DELTA_H
 #define DELTA_H
 
-/* handling of delta buffers */
-extern void *diff_delta(const void *from_buf, unsigned long from_size,
-			const void *to_buf, unsigned long to_size,
-		        unsigned long *delta_size, unsigned long max_size);
-extern void *patch_delta(void *src_buf, unsigned long src_size,
+/* opaque object for delta index */
+struct delta_index;
+
+/*
+ * create_delta_index: compute index data from given buffer
+ *
+ * This returns a pointer to a struct delta_index that should be passed to
+ * subsequent create_delta() calls, or to free_delta_index().  A NULL pointer
+ * is returned on failure.  The given buffer must not be freed nor altered
+ * before free_delta_index() is called.  The returned pointer must be freed
+ * using free_delta_index().
+ */
+extern struct delta_index *
+create_delta_index(const void *buf, unsigned long bufsize);
+
+/*
+ * free_delta_index: free the index created by create_delta_index()
+ */
+extern void free_delta_index(struct delta_index *index);
+
+/*
+ * create_delta: create a delta from given index for the given buffer
+ *
+ * This function may be called multiple times with different buffers using
+ * the same delta_index pointer.  If max_delta_size is non-zero and the
+ * resulting delta is to be larger than max_delta_size then NULL is returned.
+ * On success, a non-NULL pointer to the buffer with the delta data is
+ * returned and *delta_size is updated with its size.  The returned buffer
+ * must be freed by the caller.
+ */
+extern void *
+create_delta(const struct delta_index *index,
+	     const void *buf, unsigned long bufsize,
+	     unsigned long *delta_size, unsigned long max_delta_size);
+
+/*
+ * diff_delta: create a delta from source buffer to target buffer
+ *
+ * If max_delta_size is non-zero and the resulting delta is to be larger
+ * than max_delta_size then NULL is returned.  On success, a non-NULL
+ * pointer to the buffer with the delta data is returned and *delta_size is
+ * updated with its size.  The returned buffer must be freed by the caller.
+ */
+static inline void *
+diff_delta(const void *src_buf, unsigned long src_bufsize,
+	   const void *trg_buf, unsigned long trg_bufsize,
+	   unsigned long *delta_size, unsigned long max_delta_size)
+{
+	struct delta_index *index = create_delta_index(src_buf, src_bufsize);
+	if (index) {
+		void *delta = create_delta(index, trg_buf, trg_bufsize,
+					   delta_size, max_delta_size);
+		free_delta_index(index);
+		return delta;
+	}
+	return NULL;
+}
+
+/*
+ * patch_delta: recreate target buffer given source buffer and delta data
+ *
+ * On success, a non-NULL pointer to the target buffer is returned and
+ * *trg_bufsize is updated with its size.  On failure a NULL pointer is
+ * returned.  The returned buffer must be freed by the caller.
+ */
+extern void *patch_delta(const void *src_buf, unsigned long src_size,
 			 const void *delta_buf, unsigned long delta_size,
 			 unsigned long *dst_size);
 
@@ -14,7 +75,7 @@ extern void *patch_delta(void *src_buf, unsigned long src_size,
 
 /*
  * This must be called twice on the delta data buffer, first to get the
- * expected reference buffer size, and again to get the result buffer size.
+ * expected source buffer size, and again to get the target buffer size.
  */
 static inline unsigned long get_delta_hdr_size(const unsigned char **datap,
 					       const unsigned char *top)
diff --git a/diff-delta.c b/diff-delta.c
index 5bc22047f8..c618875188 100644
--- a/diff-delta.c
+++ b/diff-delta.c
@@ -20,69 +20,188 @@
 
 #include <stdlib.h>
 #include <string.h>
-#include <zlib.h>
 #include "delta.h"
 
 
-/* block size: min = 16, max = 64k, power of 2 */
-#define BLK_SIZE 16
-
-#define MIN(a, b) ((a) < (b) ? (a) : (b))
+/* maximum hash entry list for the same hash bucket */
+#define HASH_LIMIT 64
+
+#define RABIN_SHIFT 23
+#define RABIN_WINDOW 16
+
+static const unsigned int T[256] = {
+	0x00000000, 0xab59b4d1, 0x56b369a2, 0xfdeadd73, 0x063f6795, 0xad66d344,
+	0x508c0e37, 0xfbd5bae6, 0x0c7ecf2a, 0xa7277bfb, 0x5acda688, 0xf1941259,
+	0x0a41a8bf, 0xa1181c6e, 0x5cf2c11d, 0xf7ab75cc, 0x18fd9e54, 0xb3a42a85,
+	0x4e4ef7f6, 0xe5174327, 0x1ec2f9c1, 0xb59b4d10, 0x48719063, 0xe32824b2,
+	0x1483517e, 0xbfdae5af, 0x423038dc, 0xe9698c0d, 0x12bc36eb, 0xb9e5823a,
+	0x440f5f49, 0xef56eb98, 0x31fb3ca8, 0x9aa28879, 0x6748550a, 0xcc11e1db,
+	0x37c45b3d, 0x9c9defec, 0x6177329f, 0xca2e864e, 0x3d85f382, 0x96dc4753,
+	0x6b369a20, 0xc06f2ef1, 0x3bba9417, 0x90e320c6, 0x6d09fdb5, 0xc6504964,
+	0x2906a2fc, 0x825f162d, 0x7fb5cb5e, 0xd4ec7f8f, 0x2f39c569, 0x846071b8,
+	0x798aaccb, 0xd2d3181a, 0x25786dd6, 0x8e21d907, 0x73cb0474, 0xd892b0a5,
+	0x23470a43, 0x881ebe92, 0x75f463e1, 0xdeadd730, 0x63f67950, 0xc8afcd81,
+	0x354510f2, 0x9e1ca423, 0x65c91ec5, 0xce90aa14, 0x337a7767, 0x9823c3b6,
+	0x6f88b67a, 0xc4d102ab, 0x393bdfd8, 0x92626b09, 0x69b7d1ef, 0xc2ee653e,
+	0x3f04b84d, 0x945d0c9c, 0x7b0be704, 0xd05253d5, 0x2db88ea6, 0x86e13a77,
+	0x7d348091, 0xd66d3440, 0x2b87e933, 0x80de5de2, 0x7775282e, 0xdc2c9cff,
+	0x21c6418c, 0x8a9ff55d, 0x714a4fbb, 0xda13fb6a, 0x27f92619, 0x8ca092c8,
+	0x520d45f8, 0xf954f129, 0x04be2c5a, 0xafe7988b, 0x5432226d, 0xff6b96bc,
+	0x02814bcf, 0xa9d8ff1e, 0x5e738ad2, 0xf52a3e03, 0x08c0e370, 0xa39957a1,
+	0x584ced47, 0xf3155996, 0x0eff84e5, 0xa5a63034, 0x4af0dbac, 0xe1a96f7d,
+	0x1c43b20e, 0xb71a06df, 0x4ccfbc39, 0xe79608e8, 0x1a7cd59b, 0xb125614a,
+	0x468e1486, 0xedd7a057, 0x103d7d24, 0xbb64c9f5, 0x40b17313, 0xebe8c7c2,
+	0x16021ab1, 0xbd5bae60, 0x6cb54671, 0xc7ecf2a0, 0x3a062fd3, 0x915f9b02,
+	0x6a8a21e4, 0xc1d39535, 0x3c394846, 0x9760fc97, 0x60cb895b, 0xcb923d8a,
+	0x3678e0f9, 0x9d215428, 0x66f4eece, 0xcdad5a1f, 0x3047876c, 0x9b1e33bd,
+	0x7448d825, 0xdf116cf4, 0x22fbb187, 0x89a20556, 0x7277bfb0, 0xd92e0b61,
+	0x24c4d612, 0x8f9d62c3, 0x7836170f, 0xd36fa3de, 0x2e857ead, 0x85dcca7c,
+	0x7e09709a, 0xd550c44b, 0x28ba1938, 0x83e3ade9, 0x5d4e7ad9, 0xf617ce08,
+	0x0bfd137b, 0xa0a4a7aa, 0x5b711d4c, 0xf028a99d, 0x0dc274ee, 0xa69bc03f,
+	0x5130b5f3, 0xfa690122, 0x0783dc51, 0xacda6880, 0x570fd266, 0xfc5666b7,
+	0x01bcbbc4, 0xaae50f15, 0x45b3e48d, 0xeeea505c, 0x13008d2f, 0xb85939fe,
+	0x438c8318, 0xe8d537c9, 0x153feaba, 0xbe665e6b, 0x49cd2ba7, 0xe2949f76,
+	0x1f7e4205, 0xb427f6d4, 0x4ff24c32, 0xe4abf8e3, 0x19412590, 0xb2189141,
+	0x0f433f21, 0xa41a8bf0, 0x59f05683, 0xf2a9e252, 0x097c58b4, 0xa225ec65,
+	0x5fcf3116, 0xf49685c7, 0x033df00b, 0xa86444da, 0x558e99a9, 0xfed72d78,
+	0x0502979e, 0xae5b234f, 0x53b1fe3c, 0xf8e84aed, 0x17bea175, 0xbce715a4,
+	0x410dc8d7, 0xea547c06, 0x1181c6e0, 0xbad87231, 0x4732af42, 0xec6b1b93,
+	0x1bc06e5f, 0xb099da8e, 0x4d7307fd, 0xe62ab32c, 0x1dff09ca, 0xb6a6bd1b,
+	0x4b4c6068, 0xe015d4b9, 0x3eb80389, 0x95e1b758, 0x680b6a2b, 0xc352defa,
+	0x3887641c, 0x93ded0cd, 0x6e340dbe, 0xc56db96f, 0x32c6cca3, 0x999f7872,
+	0x6475a501, 0xcf2c11d0, 0x34f9ab36, 0x9fa01fe7, 0x624ac294, 0xc9137645,
+	0x26459ddd, 0x8d1c290c, 0x70f6f47f, 0xdbaf40ae, 0x207afa48, 0x8b234e99,
+	0x76c993ea, 0xdd90273b, 0x2a3b52f7, 0x8162e626, 0x7c883b55, 0xd7d18f84,
+	0x2c043562, 0x875d81b3, 0x7ab75cc0, 0xd1eee811
+};
 
-#define GR_PRIME 0x9e370001
-#define HASH(v, shift) (((unsigned int)(v) * GR_PRIME) >> (shift))
+static const unsigned int U[256] = {
+	0x00000000, 0x7eb5200d, 0x5633f4cb, 0x2886d4c6, 0x073e5d47, 0x798b7d4a,
+	0x510da98c, 0x2fb88981, 0x0e7cba8e, 0x70c99a83, 0x584f4e45, 0x26fa6e48,
+	0x0942e7c9, 0x77f7c7c4, 0x5f711302, 0x21c4330f, 0x1cf9751c, 0x624c5511,
+	0x4aca81d7, 0x347fa1da, 0x1bc7285b, 0x65720856, 0x4df4dc90, 0x3341fc9d,
+	0x1285cf92, 0x6c30ef9f, 0x44b63b59, 0x3a031b54, 0x15bb92d5, 0x6b0eb2d8,
+	0x4388661e, 0x3d3d4613, 0x39f2ea38, 0x4747ca35, 0x6fc11ef3, 0x11743efe,
+	0x3eccb77f, 0x40799772, 0x68ff43b4, 0x164a63b9, 0x378e50b6, 0x493b70bb,
+	0x61bda47d, 0x1f088470, 0x30b00df1, 0x4e052dfc, 0x6683f93a, 0x1836d937,
+	0x250b9f24, 0x5bbebf29, 0x73386bef, 0x0d8d4be2, 0x2235c263, 0x5c80e26e,
+	0x740636a8, 0x0ab316a5, 0x2b7725aa, 0x55c205a7, 0x7d44d161, 0x03f1f16c,
+	0x2c4978ed, 0x52fc58e0, 0x7a7a8c26, 0x04cfac2b, 0x73e5d470, 0x0d50f47d,
+	0x25d620bb, 0x5b6300b6, 0x74db8937, 0x0a6ea93a, 0x22e87dfc, 0x5c5d5df1,
+	0x7d996efe, 0x032c4ef3, 0x2baa9a35, 0x551fba38, 0x7aa733b9, 0x041213b4,
+	0x2c94c772, 0x5221e77f, 0x6f1ca16c, 0x11a98161, 0x392f55a7, 0x479a75aa,
+	0x6822fc2b, 0x1697dc26, 0x3e1108e0, 0x40a428ed, 0x61601be2, 0x1fd53bef,
+	0x3753ef29, 0x49e6cf24, 0x665e46a5, 0x18eb66a8, 0x306db26e, 0x4ed89263,
+	0x4a173e48, 0x34a21e45, 0x1c24ca83, 0x6291ea8e, 0x4d29630f, 0x339c4302,
+	0x1b1a97c4, 0x65afb7c9, 0x446b84c6, 0x3adea4cb, 0x1258700d, 0x6ced5000,
+	0x4355d981, 0x3de0f98c, 0x15662d4a, 0x6bd30d47, 0x56ee4b54, 0x285b6b59,
+	0x00ddbf9f, 0x7e689f92, 0x51d01613, 0x2f65361e, 0x07e3e2d8, 0x7956c2d5,
+	0x5892f1da, 0x2627d1d7, 0x0ea10511, 0x7014251c, 0x5facac9d, 0x21198c90,
+	0x099f5856, 0x772a785b, 0x4c921c31, 0x32273c3c, 0x1aa1e8fa, 0x6414c8f7,
+	0x4bac4176, 0x3519617b, 0x1d9fb5bd, 0x632a95b0, 0x42eea6bf, 0x3c5b86b2,
+	0x14dd5274, 0x6a687279, 0x45d0fbf8, 0x3b65dbf5, 0x13e30f33, 0x6d562f3e,
+	0x506b692d, 0x2ede4920, 0x06589de6, 0x78edbdeb, 0x5755346a, 0x29e01467,
+	0x0166c0a1, 0x7fd3e0ac, 0x5e17d3a3, 0x20a2f3ae, 0x08242768, 0x76910765,
+	0x59298ee4, 0x279caee9, 0x0f1a7a2f, 0x71af5a22, 0x7560f609, 0x0bd5d604,
+	0x235302c2, 0x5de622cf, 0x725eab4e, 0x0ceb8b43, 0x246d5f85, 0x5ad87f88,
+	0x7b1c4c87, 0x05a96c8a, 0x2d2fb84c, 0x539a9841, 0x7c2211c0, 0x029731cd,
+	0x2a11e50b, 0x54a4c506, 0x69998315, 0x172ca318, 0x3faa77de, 0x411f57d3,
+	0x6ea7de52, 0x1012fe5f, 0x38942a99, 0x46210a94, 0x67e5399b, 0x19501996,
+	0x31d6cd50, 0x4f63ed5d, 0x60db64dc, 0x1e6e44d1, 0x36e89017, 0x485db01a,
+	0x3f77c841, 0x41c2e84c, 0x69443c8a, 0x17f11c87, 0x38499506, 0x46fcb50b,
+	0x6e7a61cd, 0x10cf41c0, 0x310b72cf, 0x4fbe52c2, 0x67388604, 0x198da609,
+	0x36352f88, 0x48800f85, 0x6006db43, 0x1eb3fb4e, 0x238ebd5d, 0x5d3b9d50,
+	0x75bd4996, 0x0b08699b, 0x24b0e01a, 0x5a05c017, 0x728314d1, 0x0c3634dc,
+	0x2df207d3, 0x534727de, 0x7bc1f318, 0x0574d315, 0x2acc5a94, 0x54797a99,
+	0x7cffae5f, 0x024a8e52, 0x06852279, 0x78300274, 0x50b6d6b2, 0x2e03f6bf,
+	0x01bb7f3e, 0x7f0e5f33, 0x57888bf5, 0x293dabf8, 0x08f998f7, 0x764cb8fa,
+	0x5eca6c3c, 0x207f4c31, 0x0fc7c5b0, 0x7172e5bd, 0x59f4317b, 0x27411176,
+	0x1a7c5765, 0x64c97768, 0x4c4fa3ae, 0x32fa83a3, 0x1d420a22, 0x63f72a2f,
+	0x4b71fee9, 0x35c4dee4, 0x1400edeb, 0x6ab5cde6, 0x42331920, 0x3c86392d,
+	0x133eb0ac, 0x6d8b90a1, 0x450d4467, 0x3bb8646a
+};
 
-struct index {
+struct index_entry {
 	const unsigned char *ptr;
 	unsigned int val;
-	struct index *next;
+	struct index_entry *next;
+};
+
+struct delta_index {
+	const void *src_buf;
+	unsigned long src_size;
+	unsigned int hash_mask;
+	struct index_entry *hash[0];
 };
 
-static struct index ** delta_index(const unsigned char *buf,
-				   unsigned long bufsize,
-				   unsigned long trg_bufsize,
-				   unsigned int *hash_shift)
+struct delta_index * create_delta_index(const void *buf, unsigned long bufsize)
 {
-	unsigned int i, hsize, hshift, hlimit, entries, *hash_count;
-	const unsigned char *data;
-	struct index *entry, **hash;
+	unsigned int i, hsize, hmask, entries, prev_val, *hash_count;
+	const unsigned char *data, *buffer = buf;
+	struct delta_index *index;
+	struct index_entry *entry, **hash;
 	void *mem;
+	unsigned long memsize;
+
+	if (!buf || !bufsize)
+		return NULL;
 
-	/* determine index hash size */
-	entries = bufsize  / BLK_SIZE;
+	/* Determine index hash size.  Note that indexing skips the
+	   first byte to allow for optimizing the rabin polynomial
+	   initialization in create_delta(). */
+	entries = (bufsize - 1)  / RABIN_WINDOW;
 	hsize = entries / 4;
 	for (i = 4; (1 << i) < hsize && i < 31; i++);
 	hsize = 1 << i;
-	hshift = 32 - i;
-	*hash_shift = hshift;
+	hmask = hsize - 1;
 
 	/* allocate lookup index */
-	mem = malloc(hsize * sizeof(*hash) + entries * sizeof(*entry));
+	memsize = sizeof(*index) +
+		  sizeof(*hash) * hsize +
+		  sizeof(*entry) * entries;
+	mem = malloc(memsize);
 	if (!mem)
 		return NULL;
+	index = mem;
+	mem = index + 1;
 	hash = mem;
-	entry = mem + hsize * sizeof(*hash);
+	mem = hash + hsize;
+	entry = mem;
+
+	index->src_buf = buf;
+	index->src_size = bufsize;
+	index->hash_mask = hmask;
 	memset(hash, 0, hsize * sizeof(*hash));
 
 	/* allocate an array to count hash entries */
 	hash_count = calloc(hsize, sizeof(*hash_count));
 	if (!hash_count) {
-		free(hash);
+		free(index);
 		return NULL;
 	}
 
 	/* then populate the index */
-	data = buf + entries * BLK_SIZE - BLK_SIZE;
-	while (data >= buf) {
-		unsigned int val = adler32(0, data, BLK_SIZE);
-		i = HASH(val, hshift);
-		entry->ptr = data;
-		entry->val = val;
-		entry->next = hash[i];
-		hash[i] = entry++;
-		hash_count[i]++;
-		data -= BLK_SIZE;
- 	}
+	prev_val = ~0;
+	for (data = buffer + entries * RABIN_WINDOW - RABIN_WINDOW;
+	     data >= buffer;
+	     data -= RABIN_WINDOW) {
+		unsigned int val = 0;
+		for (i = 1; i <= RABIN_WINDOW; i++)
+			val = ((val << 8) | data[i]) ^ T[val >> RABIN_SHIFT];
+		if (val == prev_val) {
+			/* keep the lowest of consecutive identical blocks */
+			entry[-1].ptr = data + RABIN_WINDOW;
+		} else {
+			prev_val = val;
+			i = val & hmask;
+			entry->ptr = data + RABIN_WINDOW;
+			entry->val = val;
+			entry->next = hash[i];
+			hash[i] = entry++;
+			hash_count[i]++;
+			entries--;
+		}
+	}
 
 	/*
 	 * Determine a limit on the number of entries in the same hash
@@ -91,27 +210,18 @@ static struct index ** delta_index(const unsigned char *buf,
 	 * bucket that would bring us to O(m*n) computing costs (m and n
 	 * corresponding to reference and target buffer sizes).
 	 *
-	 * The more the target buffer is large, the more it is important to
-	 * have small entry lists for each hash buckets.  With such a limit
-	 * the cost is bounded to something more like O(m+n).
-	 */
-	hlimit = (1 << 26) / trg_bufsize;
-	if (hlimit < 4*BLK_SIZE)
-		hlimit = 4*BLK_SIZE;
-
-	/*
-	 * Now make sure none of the hash buckets has more entries than
+	 * Make sure none of the hash buckets has more entries than
 	 * we're willing to test.  Otherwise we cull the entry list
 	 * uniformly to still preserve a good repartition across
 	 * the reference buffer.
 	 */
 	for (i = 0; i < hsize; i++) {
-		if (hash_count[i] < hlimit)
+		if (hash_count[i] < HASH_LIMIT)
 			continue;
 		entry = hash[i];
 		do {
-			struct index *keep = entry;
-			int skip = hash_count[i] / hlimit / 2;
+			struct index_entry *keep = entry;
+			int skip = hash_count[i] / HASH_LIMIT / 2;
 			do {
 				entry = entry->next;
 			} while(--skip && entry);
@@ -120,32 +230,35 @@ static struct index ** delta_index(const unsigned char *buf,
 	}
 	free(hash_count);
 
-	return hash;
+	/* If we didn't use all hash entries, free the unused memory. */
+	if (entries)
+		index = realloc(index, memsize - entries * sizeof(*entry));
+
+	return index;
 }
 
-/* provide the size of the copy opcode given the block offset and size */
-#define COPYOP_SIZE(o, s) \
-    (!!(o & 0xff) + !!(o & 0xff00) + !!(o & 0xff0000) + !!(o & 0xff000000) + \
-     !!(s & 0xff) + !!(s & 0xff00) + 1)
+void free_delta_index(struct delta_index *index)
+{
+	free(index);
+}
 
-/* the maximum size for any opcode */
-#define MAX_OP_SIZE COPYOP_SIZE(0xffffffff, 0xffffffff)
+/*
+ * The maximum size for any opcode sequence, including the initial header
+ * plus rabin window plus biggest copy.
+ */
+#define MAX_OP_SIZE	(5 + 5 + 1 + RABIN_WINDOW + 7)
 
-void *diff_delta(const void *from_buf, unsigned long from_size,
-		 const void *to_buf, unsigned long to_size,
-		 unsigned long *delta_size,
-		 unsigned long max_size)
+void *
+create_delta(const struct delta_index *index,
+	     const void *trg_buf, unsigned long trg_size,
+	     unsigned long *delta_size, unsigned long max_size)
 {
-	unsigned int i, outpos, outsize, hash_shift;
+	unsigned int i, outpos, outsize, val;
 	int inscnt;
 	const unsigned char *ref_data, *ref_top, *data, *top;
 	unsigned char *out;
-	struct index *entry, **hash;
 
-	if (!from_size || !to_size)
-		return NULL;
-	hash = delta_index(from_buf, from_size, to_size, &hash_shift);
-	if (!hash)
+	if (!trg_buf || !trg_size)
 		return NULL;
 
 	outpos = 0;
@@ -153,64 +266,66 @@ void *diff_delta(const void *from_buf, unsigned long from_size,
 	if (max_size && outsize >= max_size)
 		outsize = max_size + MAX_OP_SIZE + 1;
 	out = malloc(outsize);
-	if (!out) {
-		free(hash);
+	if (!out)
 		return NULL;
-	}
-
-	ref_data = from_buf;
-	ref_top = from_buf + from_size;
-	data = to_buf;
-	top = to_buf + to_size;
 
 	/* store reference buffer size */
-	out[outpos++] = from_size;
-	from_size >>= 7;
-	while (from_size) {
-		out[outpos - 1] |= 0x80;
-		out[outpos++] = from_size;
-		from_size >>= 7;
+	i = index->src_size;
+	while (i >= 0x80) {
+		out[outpos++] = i | 0x80;
+		i >>= 7;
 	}
+	out[outpos++] = i;
 
 	/* store target buffer size */
-	out[outpos++] = to_size;
-	to_size >>= 7;
-	while (to_size) {
-		out[outpos - 1] |= 0x80;
-		out[outpos++] = to_size;
-		to_size >>= 7;
+	i = trg_size;
+	while (i >= 0x80) {
+		out[outpos++] = i | 0x80;
+		i >>= 7;
 	}
-
-	inscnt = 0;
+	out[outpos++] = i;
+
+	ref_data = index->src_buf;
+	ref_top = ref_data + index->src_size;
+	data = trg_buf;
+	top = trg_buf + trg_size;
+
+	outpos++;
+	val = 0;
+	for (i = 0; i < RABIN_WINDOW && data < top; i++, data++) {
+		out[outpos++] = *data;
+		val = ((val << 8) | *data) ^ T[val >> RABIN_SHIFT];
+	}
+	inscnt = i;
 
 	while (data < top) {
 		unsigned int moff = 0, msize = 0;
-		if (data + BLK_SIZE <= top) {
-			unsigned int val = adler32(0, data, BLK_SIZE);
-			i = HASH(val, hash_shift);
-			for (entry = hash[i]; entry; entry = entry->next) {
-				const unsigned char *ref = entry->ptr;
-				const unsigned char *src = data;
-				unsigned int ref_size = ref_top - ref;
-				if (entry->val != val)
-					continue;
-				if (ref_size > top - src)
-					ref_size = top - src;
-				if (ref_size > 0x10000)
-					ref_size = 0x10000;
-				if (ref_size <= msize)
-					break;
-				while (ref_size-- && *src++ == *ref)
-					ref++;
-				if (msize < ref - entry->ptr) {
-					/* this is our best match so far */
-					msize = ref - entry->ptr;
-					moff = entry->ptr - ref_data;
-				}
+		struct index_entry *entry;
+		val ^= U[data[-RABIN_WINDOW]];
+		val = ((val << 8) | *data) ^ T[val >> RABIN_SHIFT];
+		i = val & index->hash_mask;
+		for (entry = index->hash[i]; entry; entry = entry->next) {
+			const unsigned char *ref = entry->ptr;
+			const unsigned char *src = data;
+			unsigned int ref_size = ref_top - ref;
+			if (entry->val != val)
+				continue;
+			if (ref_size > top - src)
+				ref_size = top - src;
+			if (ref_size > 0x10000)
+				ref_size = 0x10000;
+			if (ref_size <= msize)
+				break;
+			while (ref_size-- && *src++ == *ref)
+				ref++;
+			if (msize < ref - entry->ptr) {
+				/* this is our best match so far */
+				msize = ref - entry->ptr;
+				moff = entry->ptr - ref_data;
 			}
 		}
 
-		if (!msize || msize < COPYOP_SIZE(moff, msize)) {
+		if (msize < 4) {
 			if (!inscnt)
 				outpos++;
 			out[outpos++] = *data++;
@@ -222,6 +337,20 @@ void *diff_delta(const void *from_buf, unsigned long from_size,
 		} else {
 			unsigned char *op;
 
+			if (msize >= RABIN_WINDOW) {
+				const unsigned char *sk;
+				sk = data + msize - RABIN_WINDOW;
+				val = 0;
+				for (i = 0; i < RABIN_WINDOW; i++)
+					val = ((val << 8) | *sk++) ^ T[val >> RABIN_SHIFT];
+			} else {
+				const unsigned char *sk = data + 1;
+				for (i = 1; i < msize; i++) {
+					val ^= U[sk[-RABIN_WINDOW]];
+					val = ((val << 8) | *sk++) ^ T[val >> RABIN_SHIFT];
+				}
+			}
+
 			if (inscnt) {
 				while (moff && ref_data[moff-1] == data[-1]) {
 					if (msize == 0x10000)
@@ -266,12 +395,10 @@ void *diff_delta(const void *from_buf, unsigned long from_size,
 			if (max_size && outsize >= max_size)
 				outsize = max_size + MAX_OP_SIZE + 1;
 			if (max_size && outpos > max_size)
-				out = NULL;
-			else
-				out = realloc(out, outsize);
+				break;
+			out = realloc(out, outsize);
 			if (!out) {
 				free(tmp);
-				free(hash);
 				return NULL;
 			}
 		}
@@ -280,7 +407,11 @@ void *diff_delta(const void *from_buf, unsigned long from_size,
 	if (inscnt)
 		out[outpos - inscnt - 1] = inscnt;
 
-	free(hash);
+	if (max_size && outpos > max_size) {
+		free(out);
+		return NULL;
+	}
+
 	*delta_size = outpos;
 	return out;
 }
diff --git a/diff.c b/diff.c
index c845c87113..bfe54c3e09 100644
--- a/diff.c
+++ b/diff.c
@@ -8,6 +8,7 @@
 #include "quote.h"
 #include "diff.h"
 #include "diffcore.h"
+#include "delta.h"
 #include "xdiff-interface.h"
 
 static int use_size_cache;
@@ -391,6 +392,90 @@ static void show_stats(struct diffstat_t* data)
 			total_files, adds, dels);
 }
 
+static unsigned char *deflate_it(char *data,
+				 unsigned long size,
+				 unsigned long *result_size)
+{
+	int bound;
+	unsigned char *deflated;
+	z_stream stream;
+
+	memset(&stream, 0, sizeof(stream));
+	deflateInit(&stream, Z_BEST_COMPRESSION);
+	bound = deflateBound(&stream, size);
+	deflated = xmalloc(bound);
+	stream.next_out = deflated;
+	stream.avail_out = bound;
+
+	stream.next_in = (unsigned char *)data;
+	stream.avail_in = size;
+	while (deflate(&stream, Z_FINISH) == Z_OK)
+		; /* nothing */
+	deflateEnd(&stream);
+	*result_size = stream.total_out;
+	return deflated;
+}
+
+static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
+{
+	void *cp;
+	void *delta;
+	void *deflated;
+	void *data;
+	unsigned long orig_size;
+	unsigned long delta_size;
+	unsigned long deflate_size;
+	unsigned long data_size;
+
+	printf("GIT binary patch\n");
+	/* We could do deflated delta, or we could do just deflated two,
+	 * whichever is smaller.
+	 */
+	delta = NULL;
+	deflated = deflate_it(two->ptr, two->size, &deflate_size);
+	if (one->size && two->size) {
+		delta = diff_delta(one->ptr, one->size,
+				   two->ptr, two->size,
+				   &delta_size, deflate_size);
+		if (delta) {
+			void *to_free = delta;
+			orig_size = delta_size;
+			delta = deflate_it(delta, delta_size, &delta_size);
+			free(to_free);
+		}
+	}
+
+	if (delta && delta_size < deflate_size) {
+		printf("delta %lu\n", orig_size);
+		free(deflated);
+		data = delta;
+		data_size = delta_size;
+	}
+	else {
+		printf("literal %lu\n", two->size);
+		free(delta);
+		data = deflated;
+		data_size = deflate_size;
+	}
+
+	/* emit data encoded in base85 */
+	cp = data;
+	while (data_size) {
+		int bytes = (52 < data_size) ? 52 : data_size;
+		char line[70];
+		data_size -= bytes;
+		if (bytes <= 26)
+			line[0] = bytes + 'A' - 1;
+		else
+			line[0] = bytes - 26 + 'a' - 1;
+		encode_85(line + 1, cp, bytes);
+		cp += bytes;
+		puts(line);
+	}
+	printf("\n");
+	free(data);
+}
+
 #define FIRST_FEW_BYTES 8000
 static int mmfile_is_binary(mmfile_t *mf)
 {
@@ -407,6 +492,7 @@ static void builtin_diff(const char *name_a,
 			 struct diff_filespec *one,
 			 struct diff_filespec *two,
 			 const char *xfrm_msg,
+			 struct diff_options *o,
 			 int complete_rewrite)
 {
 	mmfile_t mf1, mf2;
@@ -451,8 +537,17 @@ static void builtin_diff(const char *name_a,
 	if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
 		die("unable to read files to diff");
 
-	if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2))
-		printf("Binary files %s and %s differ\n", lbl[0], lbl[1]);
+	if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2)) {
+		/* Quite common confusing case */
+		if (mf1.size == mf2.size &&
+		    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
+			goto free_ab_and_return;
+		if (o->binary)
+			emit_binary_diff(&mf1, &mf2);
+		else
+			printf("Binary files %s and %s differ\n",
+			       lbl[0], lbl[1]);
+	}
 	else {
 		/* Crazy xdl interfaces.. */
 		const char *diffopts = getenv("GIT_DIFF_OPTS");
@@ -928,6 +1023,7 @@ static void run_diff_cmd(const char *pgm,
 			 struct diff_filespec *one,
 			 struct diff_filespec *two,
 			 const char *xfrm_msg,
+			 struct diff_options *o,
 			 int complete_rewrite)
 {
 	if (pgm) {
@@ -937,7 +1033,7 @@ static void run_diff_cmd(const char *pgm,
 	}
 	if (one && two)
 		builtin_diff(name, other ? other : name,
-			     one, two, xfrm_msg, complete_rewrite);
+			     one, two, xfrm_msg, o, complete_rewrite);
 	else
 		printf("* Unmerged path %s\n", name);
 }
@@ -971,7 +1067,7 @@ static void run_diff(struct diff_filepair *p, struct diff_options *o)
 
 	if (DIFF_PAIR_UNMERGED(p)) {
 		/* unmerged */
-		run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, 0);
+		run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
 		return;
 	}
 
@@ -1041,14 +1137,14 @@ static void run_diff(struct diff_filepair *p, struct diff_options *o)
 		 * needs to be split into deletion and creation.
 		 */
 		struct diff_filespec *null = alloc_filespec(two->path);
-		run_diff_cmd(NULL, name, other, one, null, xfrm_msg, 0);
+		run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
 		free(null);
 		null = alloc_filespec(one->path);
-		run_diff_cmd(NULL, name, other, null, two, xfrm_msg, 0);
+		run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
 		free(null);
 	}
 	else
-		run_diff_cmd(pgm, name, other, one, two, xfrm_msg,
+		run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
 			     complete_rewrite);
 
 	free(name_munged);
@@ -1147,6 +1243,10 @@ int diff_opt_parse(struct diff_options *options, const char **av, int ac)
 		options->rename_limit = strtoul(arg+2, NULL, 10);
 	else if (!strcmp(arg, "--full-index"))
 		options->full_index = 1;
+	else if (!strcmp(arg, "--binary")) {
+		options->output_format = DIFF_FORMAT_PATCH;
+		options->full_index = options->binary = 1;
+	}
 	else if (!strcmp(arg, "--name-only"))
 		options->output_format = DIFF_FORMAT_NAME;
 	else if (!strcmp(arg, "--name-status"))
diff --git a/diff.h b/diff.h
index b3b2c4dd28..d052608404 100644
--- a/diff.h
+++ b/diff.h
@@ -28,6 +28,7 @@ struct diff_options {
 		 with_raw:1,
 		 with_stat:1,
 		 tree_in_recursive:1,
+		 binary:1,
 		 full_index:1,
 		 silent_on_remove:1,
 		 find_copies_harder:1;
diff --git a/dump-cache-tree.c b/dump-cache-tree.c
new file mode 100644
index 0000000000..1ccaf51773
--- /dev/null
+++ b/dump-cache-tree.c
@@ -0,0 +1,64 @@
+#include "cache.h"
+#include "tree.h"
+#include "cache-tree.h"
+
+
+static void dump_one(struct cache_tree *it, const char *pfx, const char *x)
+{
+	if (it->entry_count < 0)
+		printf("%-40s %s%s (%d subtrees)\n",
+		       "invalid", x, pfx, it->subtree_nr);
+	else
+		printf("%s %s%s (%d entries, %d subtrees)\n",
+		       sha1_to_hex(it->sha1), x, pfx,
+		       it->entry_count, it->subtree_nr);
+}
+
+static int dump_cache_tree(struct cache_tree *it,
+			   struct cache_tree *ref,
+			   const char *pfx)
+{
+	int i;
+	int errs = 0;
+
+	if (!it || !ref)
+		/* missing in either */
+		return 0;
+
+	if (it->entry_count < 0) {
+		dump_one(it, pfx, "");
+		dump_one(ref, pfx, "#(ref) ");
+		if (it->subtree_nr != ref->subtree_nr)
+			errs = 1;
+	}
+	else {
+		dump_one(it, pfx, "");
+		if (memcmp(it->sha1, ref->sha1, 20) ||
+		    ref->entry_count != it->entry_count ||
+		    ref->subtree_nr != it->subtree_nr) {
+			dump_one(ref, pfx, "#(ref) ");
+			errs = 1;
+		}
+	}
+
+	for (i = 0; i < it->subtree_nr; i++) {
+		char path[PATH_MAX];
+		struct cache_tree_sub *down = it->down[i];
+		struct cache_tree_sub *rdwn;
+
+		rdwn = cache_tree_sub(ref, down->name);
+		sprintf(path, "%s%.*s/", pfx, down->namelen, down->name);
+		if (dump_cache_tree(down->cache_tree, rdwn->cache_tree, path))
+			errs = 1;
+	}
+	return errs;
+}
+
+int main(int ac, char **av)
+{
+	struct cache_tree *another = cache_tree();
+	if (read_cache() < 0)
+		die("unable to read index file");
+	cache_tree_update(another, active_cache, active_nr, 0, 1);
+	return dump_cache_tree(active_cache_tree, another, "");
+}
diff --git a/fsck-objects.c b/fsck-objects.c
index 59b25904cb..1922b6d84c 100644
--- a/fsck-objects.c
+++ b/fsck-objects.c
@@ -8,6 +8,7 @@
 #include "tag.h"
 #include "refs.h"
 #include "pack.h"
+#include "cache-tree.h"
 
 #define REACHABLE 0x0001
 
@@ -438,6 +439,28 @@ static int fsck_head_link(void)
 	return 0;
 }
 
+static int fsck_cache_tree(struct cache_tree *it)
+{
+	int i;
+	int err = 0;
+
+	if (0 <= it->entry_count) {
+		struct object *obj = parse_object(it->sha1);
+		if (!obj) {
+			error("%s: invalid sha1 pointer in cache-tree",
+			      sha1_to_hex(it->sha1));
+			return 1;
+		}
+		mark_reachable(obj, REACHABLE);
+		obj->used = 1;
+		if (obj->type != tree_type)
+			err |= objerror(obj, "non-tree in cache-tree");
+	}
+	for (i = 0; i < it->subtree_nr; i++)
+		err |= fsck_cache_tree(it->down[i]->cache_tree);
+	return err;
+}
+
 int main(int argc, char **argv)
 {
 	int i, heads;
@@ -547,6 +570,8 @@ int main(int argc, char **argv)
 			obj->used = 1;
 			mark_reachable(obj, REACHABLE);
 		}
+		if (active_cache_tree)
+			fsck_cache_tree(active_cache_tree);
 	}
 
 	check_connectivity();
diff --git a/git-parse-remote.sh b/git-parse-remote.sh
index c9b899e3d7..187f0883c9 100755
--- a/git-parse-remote.sh
+++ b/git-parse-remote.sh
@@ -10,7 +10,10 @@ get_data_source () {
 		# Not so fast.	This could be the partial URL shorthand...
 		token=$(expr "z$1" : 'z\([^/]*\)/')
 		remainder=$(expr "z$1" : 'z[^/]*/\(.*\)')
-		if test -f "$GIT_DIR/branches/$token"
+		if test "$(git-repo-config --get "remote.$token.url")"
+		then
+			echo config-partial
+		elif test -f "$GIT_DIR/branches/$token"
 		then
 			echo branches-partial
 		else
@@ -18,7 +21,10 @@ get_data_source () {
 		fi
 		;;
 	*)
-		if test -f "$GIT_DIR/remotes/$1"
+		if test "$(git-repo-config --get "remote.$1.url")"
+		then
+			echo config
+		elif test -f "$GIT_DIR/remotes/$1"
 		then
 			echo remotes
 		elif test -f "$GIT_DIR/branches/$1"
@@ -35,6 +41,15 @@ get_remote_url () {
 	case "$data_source" in
 	'')
 		echo "$1" ;;
+	config-partial)
+		token=$(expr "z$1" : 'z\([^/]*\)/')
+		remainder=$(expr "z$1" : 'z[^/]*/\(.*\)')
+		url=$(git-repo-config --get "remote.$token.url")
+		echo "$url/$remainder"
+		;;
+	config)
+		git-repo-config --get "remote.$1.url"
+		;;
 	remotes)
 		sed -ne '/^URL: */{
 			s///p
@@ -56,8 +71,10 @@ get_remote_url () {
 get_remote_default_refs_for_push () {
 	data_source=$(get_data_source "$1")
 	case "$data_source" in
-	'' | branches | branches-partial)
+	'' | config-partial | branches | branches-partial)
 		;; # no default push mapping, just send matching refs.
+	config)
+		git-repo-config --get-all "remote.$1.push" ;;
 	remotes)
 		sed -ne '/^Push: */{
 			s///p
@@ -111,8 +128,11 @@ canon_refs_list_for_fetch () {
 get_remote_default_refs_for_fetch () {
 	data_source=$(get_data_source "$1")
 	case "$data_source" in
-	'' | branches-partial)
+	'' | config-partial | branches-partial)
 		echo "HEAD:" ;;
+	config)
+		canon_refs_list_for_fetch \
+			$(git-repo-config --get-all "remote.$1.fetch") ;;
 	branches)
 		remote_branch=$(sed -ne '/#/s/.*#//p' "$GIT_DIR/branches/$1")
 		case "$remote_branch" in '') remote_branch=master ;; esac
diff --git a/git.c b/git.c
index a8f7926b48..84803a62e6 100644
--- a/git.c
+++ b/git.c
@@ -50,6 +50,7 @@ static void handle_internal_command(int argc, const char **argv, char **envp)
 		{ "fmt-patch", cmd_format_patch },
 		{ "count-objects", cmd_count_objects },
 		{ "diff", cmd_diff },
+		{ "grep", cmd_grep },
 	};
 	int i;
 
diff --git a/pack-objects.c b/pack-objects.c
index 53caed42dd..523a1c7da8 100644
--- a/pack-objects.c
+++ b/pack-objects.c
@@ -994,6 +994,7 @@ static int type_size_sort(const struct object_entry *a, const struct object_entr
 struct unpacked {
 	struct object_entry *entry;
 	void *data;
+	struct delta_index *index;
 };
 
 /*
@@ -1004,61 +1005,56 @@ struct unpacked {
  * more importantly, the bigger file is likely the more recent
  * one.
  */
-static int try_delta(struct unpacked *cur, struct unpacked *old, unsigned max_depth)
+static int try_delta(struct unpacked *trg, struct unpacked *src,
+		     struct delta_index *src_index, unsigned max_depth)
 {
-	struct object_entry *cur_entry = cur->entry;
-	struct object_entry *old_entry = old->entry;
-	unsigned long size, oldsize, delta_size, sizediff;
-	long max_size;
+	struct object_entry *trg_entry = trg->entry;
+	struct object_entry *src_entry = src->entry;
+	unsigned long size, src_size, delta_size, sizediff, max_size;
 	void *delta_buf;
 
 	/* Don't bother doing diffs between different types */
-	if (cur_entry->type != old_entry->type)
+	if (trg_entry->type != src_entry->type)
 		return -1;
 
 	/* We do not compute delta to *create* objects we are not
 	 * going to pack.
 	 */
-	if (cur_entry->preferred_base)
+	if (trg_entry->preferred_base)
 		return -1;
 
-	/* If the current object is at pack edge, take the depth the
+	/*
+	 * If the current object is at pack edge, take the depth the
 	 * objects that depend on the current object into account --
 	 * otherwise they would become too deep.
 	 */
-	if (cur_entry->delta_child) {
-		if (max_depth <= cur_entry->delta_limit)
+	if (trg_entry->delta_child) {
+		if (max_depth <= trg_entry->delta_limit)
 			return 0;
-		max_depth -= cur_entry->delta_limit;
+		max_depth -= trg_entry->delta_limit;
 	}
-
-	if (old_entry->depth >= max_depth)
+	if (src_entry->depth >= max_depth)
 		return 0;
 
-	/*
-	 * NOTE!
-	 *
-	 * We always delta from the bigger to the smaller, since that's
-	 * more space-efficient (deletes don't have to say _what_ they
-	 * delete).
-	 */
-	size = cur_entry->size;
+	/* Now some size filtering euristics. */
+	size = trg_entry->size;
 	max_size = size / 2 - 20;
-	if (cur_entry->delta)
-		max_size = cur_entry->delta_size-1;
-	oldsize = old_entry->size;
-	sizediff = oldsize < size ? size - oldsize : 0;
+	if (trg_entry->delta)
+		max_size = trg_entry->delta_size-1;
+	src_size = src_entry->size;
+	sizediff = src_size < size ? size - src_size : 0;
 	if (sizediff >= max_size)
 		return 0;
-	delta_buf = diff_delta(old->data, oldsize,
-			       cur->data, size, &delta_size, max_size);
+
+	delta_buf = create_delta(src_index, trg->data, size, &delta_size, max_size);
 	if (!delta_buf)
 		return 0;
-	cur_entry->delta = old_entry;
-	cur_entry->delta_size = delta_size;
-	cur_entry->depth = old_entry->depth + 1;
+
+	trg_entry->delta = src_entry;
+	trg_entry->delta_size = delta_size;
+	trg_entry->depth = src_entry->depth + 1;
 	free(delta_buf);
-	return 0;
+	return 1;
 }
 
 static void progress_interval(int signum)
@@ -1108,12 +1104,17 @@ static void find_deltas(struct object_entry **list, int window, int depth)
 
 		if (entry->size < 50)
 			continue;
-
+		if (n->index)
+			free_delta_index(n->index);
 		free(n->data);
 		n->entry = entry;
 		n->data = read_sha1_file(entry->sha1, type, &size);
 		if (size != entry->size)
-			die("object %s inconsistent object length (%lu vs %lu)", sha1_to_hex(entry->sha1), size, entry->size);
+			die("object %s inconsistent object length (%lu vs %lu)",
+			    sha1_to_hex(entry->sha1), size, entry->size);
+		n->index = create_delta_index(n->data, size);
+		if (!n->index)
+			die("out of memory");
 
 		j = window;
 		while (--j > 0) {
@@ -1124,7 +1125,7 @@ static void find_deltas(struct object_entry **list, int window, int depth)
 			m = array + other_idx;
 			if (!m->entry)
 				break;
-			if (try_delta(n, m, depth) < 0)
+			if (try_delta(n, m, m->index, depth) < 0)
 				break;
 		}
 #if 0
@@ -1144,8 +1145,11 @@ static void find_deltas(struct object_entry **list, int window, int depth)
 	if (progress)
 		fputc('\n', stderr);
 
-	for (i = 0; i < window; ++i)
+	for (i = 0; i < window; ++i) {
+		if (array[i].index)
+			free_delta_index(array[i].index);
 		free(array[i].data);
+	}
 	free(array);
 }
 
diff --git a/patch-delta.c b/patch-delta.c
index affd4f673d..8f318ed8aa 100644
--- a/patch-delta.c
+++ b/patch-delta.c
@@ -13,7 +13,7 @@
 #include <string.h>
 #include "delta.h"
 
-void *patch_delta(void *src_buf, unsigned long src_size,
+void *patch_delta(const void *src_buf, unsigned long src_size,
 		  const void *delta_buf, unsigned long delta_size,
 		  unsigned long *dst_size)
 {
diff --git a/read-cache.c b/read-cache.c
index f97f92d90a..1f71d12578 100644
--- a/read-cache.c
+++ b/read-cache.c
@@ -4,11 +4,26 @@
  * Copyright (C) Linus Torvalds, 2005
  */
 #include "cache.h"
+#include "cache-tree.h"
+
+/* Index extensions.
+ *
+ * The first letter should be 'A'..'Z' for extensions that are not
+ * necessary for a correct operation (i.e. optimization data).
+ * When new extensions are added that _needs_ to be understood in
+ * order to correctly interpret the index file, pick character that
+ * is outside the range, to cause the reader to abort.
+ */
+
+#define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) )
+#define CACHE_EXT_TREE 0x54524545	/* "TREE" */
 
 struct cache_entry **active_cache = NULL;
 static time_t index_file_timestamp;
 unsigned int active_nr = 0, active_alloc = 0, active_cache_changed = 0;
 
+struct cache_tree *active_cache_tree = NULL;
+
 /*
  * This only updates the "non-critical" parts of the directory
  * cache, ie the parts that aren't tracked by GIT, and only used
@@ -513,6 +528,22 @@ static int verify_hdr(struct cache_header *hdr, unsigned long size)
 	return 0;
 }
 
+static int read_index_extension(const char *ext, void *data, unsigned long sz)
+{
+	switch (CACHE_EXT(ext)) {
+	case CACHE_EXT_TREE:
+		active_cache_tree = cache_tree_read(data, sz);
+		break;
+	default:
+		if (*ext < 'A' || 'Z' < *ext)
+			return error("index uses %.4s extension, which we do not understand",
+				     ext);
+		fprintf(stderr, "ignoring %.4s extension\n", ext);
+		break;
+	}
+	return 0;
+}
+
 int read_cache(void)
 {
 	int fd, i;
@@ -561,6 +592,22 @@ int read_cache(void)
 		active_cache[i] = ce;
 	}
 	index_file_timestamp = st.st_mtime;
+	while (offset <= size - 20 - 8) {
+		/* After an array of active_nr index entries,
+		 * there can be arbitrary number of extended
+		 * sections, each of which is prefixed with
+		 * extension name (4-byte) and section length
+		 * in 4-byte network byte order.
+		 */
+		unsigned long extsize;
+		memcpy(&extsize, map + offset + 4, 4);
+		extsize = ntohl(extsize);
+		if (read_index_extension(map + offset,
+					 map + offset + 8, extsize) < 0)
+			goto unmap;
+		offset += 8;
+		offset += extsize;
+	}
 	return active_nr;
 
 unmap:
@@ -595,6 +642,17 @@ static int ce_write(SHA_CTX *context, int fd, void *data, unsigned int len)
  	return 0;
 }
 
+static int write_index_ext_header(SHA_CTX *context, int fd,
+				  unsigned long ext, unsigned long sz)
+{
+	ext = htonl(ext);
+	sz = htonl(sz);
+	if ((ce_write(context, fd, &ext, 4) < 0) ||
+	    (ce_write(context, fd, &sz, 4) < 0))
+		return -1;
+	return 0;
+}
+
 static int ce_flush(SHA_CTX *context, int fd)
 {
 	unsigned int left = write_buffer_len;
@@ -691,5 +749,19 @@ int write_cache(int newfd, struct cache_entry **cache, int entries)
 		if (ce_write(&c, newfd, ce, ce_size(ce)) < 0)
 			return -1;
 	}
+
+	/* Write extension data here */
+	if (active_cache_tree) {
+		unsigned long sz;
+		void *data = cache_tree_write(active_cache_tree, &sz);
+		if (data &&
+		    !write_index_ext_header(&c, newfd, CACHE_EXT_TREE, sz) &&
+		    !ce_write(&c, newfd, data, sz))
+			;
+		else {
+			free(data);
+			return -1;
+		}
+	}
 	return ce_flush(&c, newfd);
 }
diff --git a/read-tree.c b/read-tree.c
index 26f4f7e323..067fb95e9d 100644
--- a/read-tree.c
+++ b/read-tree.c
@@ -9,6 +9,7 @@
 
 #include "object.h"
 #include "tree.h"
+#include "cache-tree.h"
 #include <sys/time.h>
 #include <signal.h>
 
@@ -421,6 +422,12 @@ static void verify_uptodate(struct cache_entry *ce)
 	die("Entry '%s' not uptodate. Cannot merge.", ce->name);
 }
 
+static void invalidate_ce_path(struct cache_entry *ce)
+{
+	if (ce)
+		cache_tree_invalidate_path(active_cache_tree, ce->name);
+}
+
 static int merged_entry(struct cache_entry *merge, struct cache_entry *old)
 {
 	merge->ce_flags |= htons(CE_UPDATE);
@@ -436,8 +443,11 @@ static int merged_entry(struct cache_entry *merge, struct cache_entry *old)
 			*merge = *old;
 		} else {
 			verify_uptodate(old);
+			invalidate_ce_path(old);
 		}
 	}
+	else
+		invalidate_ce_path(merge);
 	merge->ce_flags &= ~htons(CE_STAGEMASK);
 	add_cache_entry(merge, ADD_CACHE_OK_TO_ADD);
 	return 1;
@@ -449,6 +459,7 @@ static int deleted_entry(struct cache_entry *ce, struct cache_entry *old)
 		verify_uptodate(old);
 	ce->ce_mode = 0;
 	add_cache_entry(ce, ADD_CACHE_OK_TO_ADD);
+	invalidate_ce_path(ce);
 	return 1;
 }
 
@@ -683,8 +694,10 @@ static int oneway_merge(struct cache_entry **src)
 		return error("Cannot do a oneway merge of %d trees",
 			     merge_size);
 
-	if (!a)
+	if (!a) {
+		invalidate_ce_path(old);
 		return 0;
+	}
 	if (old && same(old, a)) {
 		return keep_entry(old);
 	}
@@ -703,6 +716,7 @@ static int read_cache_unmerged(void)
 		struct cache_entry *ce = active_cache[i];
 		if (ce_stage(ce)) {
 			deleted++;
+			invalidate_ce_path(ce);
 			continue;
 		}
 		if (deleted)
@@ -713,6 +727,39 @@ static int read_cache_unmerged(void)
 	return deleted;
 }
 
+static void prime_cache_tree_rec(struct cache_tree *it, struct tree *tree)
+{
+	struct tree_entry_list *ent;
+	int cnt;
+	
+	memcpy(it->sha1, tree->object.sha1, 20);
+	for (cnt = 0, ent = tree->entries; ent; ent = ent->next) {
+		if (!ent->directory)
+			cnt++;
+		else {
+			struct cache_tree_sub *sub;
+			struct tree *subtree = (struct tree *)ent->item.tree;
+			if (!subtree->object.parsed)
+				parse_tree(subtree);
+			sub = cache_tree_sub(it, ent->name);
+			sub->cache_tree = cache_tree();
+			prime_cache_tree_rec(sub->cache_tree, subtree);
+			cnt += sub->cache_tree->entry_count;
+		}
+	}
+	it->entry_count = cnt;
+}
+
+static void prime_cache_tree(void)
+{
+	struct tree *tree = (struct tree *)trees->item;
+	if (!tree)
+		return;
+	active_cache_tree = cache_tree();
+	prime_cache_tree_rec(active_cache_tree, tree);
+
+}
+
 static const char read_tree_usage[] = "git-read-tree (<sha> | -m [--aggressive] [-u | -i] <sha1> [<sha2> [<sha3>]])";
 
 static struct cache_file cache_file;
@@ -814,10 +861,9 @@ int main(int argc, char **argv)
 			fn = twoway_merge;
 			break;
 		case 3:
-			fn = threeway_merge;
-			break;
 		default:
 			fn = threeway_merge;
+			cache_tree_free(&active_cache_tree);
 			break;
 		}
 
@@ -828,6 +874,18 @@ int main(int argc, char **argv)
 	}
 
 	unpack_trees(fn);
+
+	/*
+	 * When reading only one tree (either the most basic form,
+	 * "-m ent" or "--reset ent" form), we can obtain a fully
+	 * valid cache-tree because the index must match exactly
+	 * what came from the tree.
+	 */
+	if (trees->item && (!merge || (stage == 2))) {
+		cache_tree_free(&active_cache_tree);		
+		prime_cache_tree();
+	}
+
 	if (write_cache(newfd, active_cache, active_nr) ||
 	    commit_index_file(&cache_file))
 		die("unable to write new index file");
diff --git a/sha1_name.c b/sha1_name.c
index 345935bb2b..ec5cd2c9ea 100644
--- a/sha1_name.c
+++ b/sha1_name.c
@@ -458,17 +458,55 @@ int get_sha1(const char *name, unsigned char *sha1)
 {
 	int ret;
 	unsigned unused;
+	int namelen = strlen(name);
+	const char *cp;
 
 	prepare_alt_odb();
-	ret = get_sha1_1(name, strlen(name), sha1);
-	if (ret < 0) {
-		const char *cp = strchr(name, ':');
-		if (cp) {
-			unsigned char tree_sha1[20];
-			if (!get_sha1_1(name, cp-name, tree_sha1))
-				return get_tree_entry(tree_sha1, cp+1, sha1,
-						      &unused);
+	ret = get_sha1_1(name, namelen, sha1);
+	if (!ret)
+		return ret;
+	/* sha1:path --> object name of path in ent sha1
+	 * :path -> object name of path in index
+	 * :[0-3]:path -> object name of path in index at stage
+	 */
+	if (name[0] == ':') {
+		int stage = 0;
+		struct cache_entry *ce;
+		int pos;
+		if (namelen < 3 ||
+		    name[2] != ':' ||
+		    name[1] < '0' || '3' < name[1])
+			cp = name + 1;
+		else {
+			stage = name[1] - '0';
+			cp = name + 3;
 		}
+		namelen = namelen - (cp - name);
+		if (!active_cache)
+			read_cache();
+		if (active_nr < 0)
+			return -1;
+		pos = cache_name_pos(cp, namelen);
+		if (pos < 0)
+			pos = -pos - 1;
+		while (pos < active_nr) {
+			ce = active_cache[pos];
+			if (ce_namelen(ce) != namelen ||
+			    memcmp(ce->name, cp, namelen))
+				break;
+			if (ce_stage(ce) == stage) {
+				memcpy(sha1, ce->sha1, 20);
+				return 0;
+			}
+		}
+		return -1;
+	}
+	cp = strchr(name, ':');
+	if (cp) {
+		unsigned char tree_sha1[20];
+		if (!get_sha1_1(name, cp-name, tree_sha1))
+			return get_tree_entry(tree_sha1, cp+1, sha1,
+					      &unused);
 	}
 	return ret;
 }
diff --git a/t/t2101-update-index-reupdate.sh b/t/t2101-update-index-reupdate.sh
new file mode 100755
index 0000000000..77aed8d800
--- /dev/null
+++ b/t/t2101-update-index-reupdate.sh
@@ -0,0 +1,82 @@
+#!/bin/sh
+#
+# Copyright (c) 2006 Junio C Hamano
+#
+
+test_description='git-update-index --again test.
+'
+
+. ./test-lib.sh
+
+test_expect_success 'update-index --add' \
+	'echo hello world >file1 &&
+	 echo goodbye people >file2 &&
+	 git-update-index --add file1 file2 &&
+	 git-ls-files -s >current &&
+	 cmp current - <<\EOF
+100644 3b18e512dba79e4c8300dd08aeb37f8e728b8dad 0	file1
+100644 9db8893856a8a02eaa73470054b7c1c5a7c82e47 0	file2
+EOF'
+
+test_expect_success 'update-index --again' \
+	'rm -f file1 &&
+	echo hello everybody >file2 &&
+	if git-update-index --again
+	then
+		echo should have refused to remove file1
+		exit 1
+	else
+		echo happy - failed as expected
+	fi &&
+	 git-ls-files -s >current &&
+	 cmp current - <<\EOF
+100644 3b18e512dba79e4c8300dd08aeb37f8e728b8dad 0	file1
+100644 9db8893856a8a02eaa73470054b7c1c5a7c82e47 0	file2
+EOF'
+
+test_expect_success 'update-index --remove --again' \
+	'git-update-index --remove --again &&
+	 git-ls-files -s >current &&
+	 cmp current - <<\EOF
+100644 0f1ae1422c2bf43f117d3dbd715c988a9ed2103f 0	file2
+EOF'
+
+test_expect_success 'first commit' 'git-commit -m initial'
+
+test_expect_success 'update-index again' \
+	'mkdir -p dir1 &&
+	echo hello world >dir1/file3 &&
+	echo goodbye people >file2 &&
+	git-update-index --add file2 dir1/file3 &&
+	echo hello everybody >file2
+	echo happy >dir1/file3 &&
+	git-update-index --again &&
+	git-ls-files -s >current &&
+	cmp current - <<\EOF
+100644 53ab446c3f4e42ce9bb728a0ccb283a101be4979 0	dir1/file3
+100644 0f1ae1422c2bf43f117d3dbd715c988a9ed2103f 0	file2
+EOF'
+
+test_expect_success 'update-index --update from subdir' \
+	'echo not so happy >file2 &&
+	cd dir1 &&
+	cat ../file2 >file3 &&
+	git-update-index --again &&
+	cd .. &&
+	git-ls-files -s >current &&
+	cmp current - <<\EOF
+100644 d7fb3f695f06c759dbf3ab00046e7cc2da22d10f 0	dir1/file3
+100644 0f1ae1422c2bf43f117d3dbd715c988a9ed2103f 0	file2
+EOF'
+
+test_expect_success 'update-index --update with pathspec' \
+	'echo very happy >file2 &&
+	cat file2 >dir1/file3 &&
+	git-update-index --again dir1/ &&
+	git-ls-files -s >current &&
+	cmp current - <<\EOF
+100644 594fb5bb1759d90998e2bf2a38261ae8e243c760 0	dir1/file3
+100644 0f1ae1422c2bf43f117d3dbd715c988a9ed2103f 0	file2
+EOF'
+
+test_done
diff --git a/t/t4012-diff-binary.sh b/t/t4012-diff-binary.sh
new file mode 100755
index 0000000000..bdd95c0d3d
--- /dev/null
+++ b/t/t4012-diff-binary.sh
@@ -0,0 +1,85 @@
+#!/bin/sh
+#
+# Copyright (c) 2006 Junio C Hamano
+#
+
+test_description='Binary diff and apply
+'
+
+. ./test-lib.sh
+
+test_expect_success 'prepare repository' \
+	'echo AIT >a && echo BIT >b && echo CIT >c && echo DIT >d &&
+	 git-update-index --add a b c d &&
+	 echo git >a &&
+	 cat ../test4012.png >b &&
+	 echo git >c &&
+	 cat b b >d'
+
+test_expect_success 'diff without --binary' \
+	'git-diff | git-apply --stat --summary >current &&
+	 cmp current - <<\EOF
+ a |    2 +-
+ b |  Bin
+ c |    2 +-
+ d |  Bin
+ 4 files changed, 2 insertions(+), 2 deletions(-)
+EOF'
+
+test_expect_success 'diff with --binary' \
+	'git-diff --binary | git-apply --stat --summary >current &&
+	 cmp current - <<\EOF
+ a |    2 +-
+ b |  Bin
+ c |    2 +-
+ d |  Bin
+ 4 files changed, 2 insertions(+), 2 deletions(-)
+EOF'
+
+# apply needs to be able to skip the binary material correctly
+# in order to report the line number of a corrupt patch.
+test_expect_success 'apply detecting corrupt patch correctly' \
+	'git-diff | sed -e 's/-CIT/xCIT/' >broken &&
+	 if git-apply --stat --summary broken 2>detected
+	 then
+		echo unhappy - should have detected an error
+		(exit 1)
+	 else
+		echo happy
+	 fi &&
+	 detected=`cat detected` &&
+	 detected=`expr "$detected" : "fatal.*at line \\([0-9]*\\)\$"` &&
+	 detected=`sed -ne "${detected}p" broken` &&
+	 test "$detected" = xCIT'
+
+test_expect_success 'apply detecting corrupt patch correctly' \
+	'git-diff --binary | sed -e 's/-CIT/xCIT/' >broken &&
+	 if git-apply --stat --summary broken 2>detected
+	 then
+		echo unhappy - should have detected an error
+		(exit 1)
+	 else
+		echo happy
+	 fi &&
+	 detected=`cat detected` &&
+	 detected=`expr "$detected" : "fatal.*at line \\([0-9]*\\)\$"` &&
+	 detected=`sed -ne "${detected}p" broken` &&
+	 test "$detected" = xCIT'
+
+test_expect_success 'initial commit' 'git-commit -a -m initial'
+
+# Try removal (b), modification (d), and creation (e).
+test_expect_success 'diff-index with --binary' \
+	'echo AIT >a && mv b e && echo CIT >c && cat e >d &&
+	 git-update-index --add --remove a b c d e &&
+	 tree0=`git-write-tree` &&
+	 git-diff --cached --binary >current &&
+	 git-apply --stat --summary current'
+
+test_expect_success 'apply binary patch' \
+	'git-reset --hard &&
+	 git-apply --binary --index <current &&
+	 tree1=`git-write-tree` &&
+	 test "$tree1" = "$tree0"'
+
+test_done
diff --git a/t/test4012.png b/t/test4012.png
new file mode 100644
index 0000000000..7b181d15ce
--- /dev/null
+++ b/t/test4012.png
diff --git a/update-index.c b/update-index.c
index d63f8ac998..f6b09a4800 100644
--- a/update-index.c
+++ b/update-index.c
@@ -6,6 +6,7 @@
 #include "cache.h"
 #include "strbuf.h"
 #include "quote.h"
+#include "cache-tree.h"
 #include "tree-walk.h"
 
 /*
@@ -71,6 +72,7 @@ static int mark_valid(const char *path)
 			active_cache[pos]->ce_flags &= ~htons(CE_VALID);
 			break;
 		}
+		cache_tree_invalidate_path(active_cache_tree, path);
 		active_cache_changed = 1;
 		return 0;
 	}
@@ -84,6 +86,12 @@ static int add_file_to_cache(const char *path)
 	struct stat st;
 
 	status = lstat(path, &st);
+
+	/* We probably want to do this in remove_file_from_cache() and
+	 * add_cache_entry() instead...
+	 */
+	cache_tree_invalidate_path(active_cache_tree, path);
+
 	if (status < 0 || S_ISDIR(st.st_mode)) {
 		/* When we used to have "path" and now we want to add
 		 * "path/file", we need a way to remove "path" before
@@ -326,6 +334,7 @@ static int add_cacheinfo(unsigned int mode, const unsigned char *sha1,
 		return error("%s: cannot add to the index - missing --add option?",
 			     path);
 	report("add '%s'", path);
+	cache_tree_invalidate_path(active_cache_tree, path);
 	return 0;
 }
 
@@ -350,6 +359,7 @@ static void chmod_path(int flip, const char *path)
 	default:
 		goto fail;
 	}
+	cache_tree_invalidate_path(active_cache_tree, path);
 	active_cache_changed = 1;
 	report("chmod %cx '%s'", flip, path);
 	return;
@@ -371,6 +381,7 @@ static void update_one(const char *path, const char *prefix, int prefix_length)
 			die("Unable to mark file %s", path);
 		goto free_return;
 	}
+	cache_tree_invalidate_path(active_cache_tree, path);
 
 	if (force_remove) {
 		if (remove_file_from_cache(p))
@@ -382,7 +393,7 @@ static void update_one(const char *path, const char *prefix, int prefix_length)
 		die("Unable to process file %s", path);
 	report("add '%s'", path);
  free_return:
-	if (p != path)
+	if (p < path || p > path + strlen(path))
 		free((char*)p);
 }
 
@@ -449,6 +460,7 @@ static void read_index_info(int line_termination)
 				free(path_name);
 			continue;
 		}
+		cache_tree_invalidate_path(active_cache_tree, path_name);
 
 		if (!mode) {
 			/* mode == 0 means there is no such path -- remove */
@@ -476,7 +488,7 @@ static void read_index_info(int line_termination)
 }
 
 static const char update_index_usage[] =
-"git-update-index [-q] [--add] [--replace] [--remove] [--unmerged] [--refresh] [--really-refresh] [--cacheinfo] [--chmod=(+|-)x] [--assume-unchanged] [--info-only] [--force-remove] [--stdin] [--index-info] [--unresolve] [--ignore-missing] [-z] [--verbose] [--] <file>...";
+"git-update-index [-q] [--add] [--replace] [--remove] [--unmerged] [--refresh] [--really-refresh] [--cacheinfo] [--chmod=(+|-)x] [--assume-unchanged] [--info-only] [--force-remove] [--stdin] [--index-info] [--unresolve] [--again] [--ignore-missing] [-z] [--verbose] [--] <file>...";
 
 static unsigned char head_sha1[20];
 static unsigned char merge_head_sha1[20];
@@ -491,11 +503,13 @@ static struct cache_entry *read_one_ent(const char *which,
 	struct cache_entry *ce;
 
 	if (get_tree_entry(ent, path, sha1, &mode)) {
-		error("%s: not in %s branch.", path, which);
+		if (which)
+			error("%s: not in %s branch.", path, which);
 		return NULL;
 	}
 	if (mode == S_IFDIR) {
-		error("%s: not a blob in %s branch.", path, which);
+		if (which)
+			error("%s: not a blob in %s branch.", path, which);
 		return NULL;
 	}
 	size = cache_entry_size(namelen);
@@ -553,6 +567,7 @@ static int unresolve_one(const char *path)
 		goto free_return;
 	}
 
+	cache_tree_invalidate_path(active_cache_tree, path);
 	remove_file_from_cache(path);
 	if (add_cache_entry(ce_2, ADD_CACHE_OK_TO_ADD)) {
 		error("%s: cannot add our version to the index.", path);
@@ -594,12 +609,55 @@ static int do_unresolve(int ac, const char **av,
 		const char *arg = av[i];
 		const char *p = prefix_path(prefix, prefix_length, arg);
 		err |= unresolve_one(p);
-		if (p != arg)
+		if (p < arg || p > arg + strlen(arg))
 			free((char*)p);
 	}
 	return err;
 }
 
+static int do_reupdate(int ac, const char **av,
+		       const char *prefix, int prefix_length)
+{
+	/* Read HEAD and run update-index on paths that are
+	 * merged and already different between index and HEAD.
+	 */
+	int pos;
+	int has_head = 1;
+	const char **pathspec = get_pathspec(prefix, av + 1);
+
+	if (read_ref(git_path("HEAD"), head_sha1))
+		/* If there is no HEAD, that means it is an initial
+		 * commit.  Update everything in the index.
+		 */
+		has_head = 0;
+ redo:
+	for (pos = 0; pos < active_nr; pos++) {
+		struct cache_entry *ce = active_cache[pos];
+		struct cache_entry *old = NULL;
+		int save_nr;
+
+		if (ce_stage(ce) || !ce_path_match(ce, pathspec))
+			continue;
+		if (has_head)
+			old = read_one_ent(NULL, head_sha1,
+					   ce->name, ce_namelen(ce), 0);
+		if (old && ce->ce_mode == old->ce_mode &&
+		    !memcmp(ce->sha1, old->sha1, 20)) {
+			free(old);
+			continue; /* unchanged */
+		}
+		/* Be careful.  The working tree may not have the
+		 * path anymore, in which case, under 'allow_remove',
+		 * or worse yet 'allow_replace', active_nr may decrease.
+		 */
+		save_nr = active_nr;
+		update_one(ce->name + prefix_length, prefix, prefix_length);
+		if (save_nr != active_nr)
+			goto redo;
+	}
+	return 0;
+}
+
 int main(int argc, const char **argv)
 {
 	int i, newfd, entries, has_errors = 0, line_termination = '\n';
@@ -717,6 +775,13 @@ int main(int argc, const char **argv)
 					active_cache_changed = 0;
 				goto finish;
 			}
+			if (!strcmp(path, "--again")) {
+				has_errors = do_reupdate(argc - i, argv + i,
+							 prefix, prefix_length);
+				if (has_errors)
+					active_cache_changed = 0;
+				goto finish;
+			}
 			if (!strcmp(path, "--ignore-missing")) {
 				not_new = 1;
 				continue;
@@ -750,7 +815,7 @@ int main(int argc, const char **argv)
 			update_one(p, NULL, 0);
 			if (set_executable_bit)
 				chmod_path(set_executable_bit, p);
-			if (p != path_name)
+			if (p < path_name || p > path_name + strlen(path_name))
 				free((char*) p);
 			if (path_name != buf.buf)
 				free(path_name);
diff --git a/write-tree.c b/write-tree.c
index dcad6e6670..7a4f691d8a 100644
--- a/write-tree.c
+++ b/write-tree.c
@@ -5,95 +5,21 @@
  */
 #include "cache.h"
 #include "tree.h"
+#include "cache-tree.h"
 
 static int missing_ok = 0;
 
-static int check_valid_sha1(unsigned char *sha1)
-{
-	int ret;
-
-	/* If we were anal, we'd check that the sha1 of the contents actually matches */
-	ret = has_sha1_file(sha1);
-	if (ret == 0)
-		perror(sha1_file_name(sha1));
-	return ret ? 0 : -1;
-}
-
-static int write_tree(struct cache_entry **cachep, int maxentries, const char *base, int baselen, unsigned char *returnsha1)
-{
-	unsigned char subdir_sha1[20];
-	unsigned long size, offset;
-	char *buffer;
-	int nr;
-
-	/* Guess at some random initial size */
-	size = 8192;
-	buffer = xmalloc(size);
-	offset = 0;
-
-	nr = 0;
-	while (nr < maxentries) {
-		struct cache_entry *ce = cachep[nr];
-		const char *pathname = ce->name, *filename, *dirname;
-		int pathlen = ce_namelen(ce), entrylen;
-		unsigned char *sha1;
-		unsigned int mode;
-
-		/* Did we hit the end of the directory? Return how many we wrote */
-		if (baselen >= pathlen || memcmp(base, pathname, baselen))
-			break;
-
-		sha1 = ce->sha1;
-		mode = ntohl(ce->ce_mode);
-
-		/* Do we have _further_ subdirectories? */
-		filename = pathname + baselen;
-		dirname = strchr(filename, '/');
-		if (dirname) {
-			int subdir_written;
-
-			subdir_written = write_tree(cachep + nr, maxentries - nr, pathname, dirname-pathname+1, subdir_sha1);
-			nr += subdir_written;
-
-			/* Now we need to write out the directory entry into this tree.. */
-			mode = S_IFDIR;
-			pathlen = dirname - pathname;
-
-			/* ..but the directory entry doesn't count towards the total count */
-			nr--;
-			sha1 = subdir_sha1;
-		}
-
-		if (!missing_ok && check_valid_sha1(sha1) < 0)
-			exit(1);
-
-		entrylen = pathlen - baselen;
-		if (offset + entrylen + 100 > size) {
-			size = alloc_nr(offset + entrylen + 100);
-			buffer = xrealloc(buffer, size);
-		}
-		offset += sprintf(buffer + offset, "%o %.*s", mode, entrylen, filename);
-		buffer[offset++] = 0;
-		memcpy(buffer + offset, sha1, 20);
-		offset += 20;
-		nr++;
-	}
-
-	write_sha1_file(buffer, offset, tree_type, returnsha1);
-	free(buffer);
-	return nr;
-}
-
 static const char write_tree_usage[] = "git-write-tree [--missing-ok]";
 
+static struct cache_file cache_file;
+
 int main(int argc, char **argv)
 {
-	int i, funny;
-	int entries;
-	unsigned char sha1[20];
-	
+	int entries, was_valid, newfd;
+
 	setup_git_directory();
 
+	newfd = hold_index_file_for_update(&cache_file, get_index_file());
 	entries = read_cache();
 	if (argc == 2) {
 		if (!strcmp(argv[1], "--missing-ok"))
@@ -108,51 +34,26 @@ int main(int argc, char **argv)
 	if (entries < 0)
 		die("git-write-tree: error reading cache");
 
-	/* Verify that the tree is merged */
-	funny = 0;
-	for (i = 0; i < entries; i++) {
-		struct cache_entry *ce = active_cache[i];
-		if (ce_stage(ce)) {
-			if (10 < ++funny) {
-				fprintf(stderr, "...\n");
-				break;
-			}
-			fprintf(stderr, "%s: unmerged (%s)\n", ce->name, sha1_to_hex(ce->sha1));
+	if (!active_cache_tree)
+		active_cache_tree = cache_tree();
+
+	was_valid = cache_tree_fully_valid(active_cache_tree);
+	if (!was_valid) {
+		if (cache_tree_update(active_cache_tree,
+				      active_cache, active_nr,
+				      missing_ok, 0) < 0)
+			die("git-write-tree: error building trees");
+		if (0 <= newfd) {
+			if (!write_cache(newfd, active_cache, active_nr))
+				commit_index_file(&cache_file);
 		}
-	}
-	if (funny)
-		die("git-write-tree: not able to write tree");
-
-	/* Also verify that the cache does not have path and path/file
-	 * at the same time.  At this point we know the cache has only
-	 * stage 0 entries.
-	 */
-	funny = 0;
-	for (i = 0; i < entries - 1; i++) {
-		/* path/file always comes after path because of the way
-		 * the cache is sorted.  Also path can appear only once,
-		 * which means conflicting one would immediately follow.
+		/* Not being able to write is fine -- we are only interested
+		 * in updating the cache-tree part, and if the next caller
+		 * ends up using the old index with unupdated cache-tree part
+		 * it misses the work we did here, but that is just a
+		 * performance penalty and not a big deal.
 		 */
-		const char *this_name = active_cache[i]->name;
-		const char *next_name = active_cache[i+1]->name;
-		int this_len = strlen(this_name);
-		if (this_len < strlen(next_name) &&
-		    strncmp(this_name, next_name, this_len) == 0 &&
-		    next_name[this_len] == '/') {
-			if (10 < ++funny) {
-				fprintf(stderr, "...\n");
-				break;
-			}
-			fprintf(stderr, "You have both %s and %s\n",
-				this_name, next_name);
-		}
 	}
-	if (funny)
-		die("git-write-tree: not able to write tree");
-
-	/* Ok, write it out */
-	if (write_tree(active_cache, entries, "", 0, sha1) != entries)
-		die("git-write-tree: internal error");
-	printf("%s\n", sha1_to_hex(sha1));
+	printf("%s\n", sha1_to_hex(active_cache_tree->sha1));
 	return 0;
 }