[PATCH 0/4] Support multiple ancestors in read-tree

@@ -1,5 +1,7 @@
 #include "tree.h"
 #include "blob.h"
+#include "commit.h"
+#include "tag.h"
 #include "cache.h"
 #include <stdlib.h>
 

@@ -212,3 +214,22 @@ int parse_tree(struct tree *item)
 	free(buffer);
 	return ret;
 }
+
+struct tree *parse_tree_indirect(const unsigned char *sha1)
+{
+	struct object *obj = parse_object(sha1);
+	do {
+		if (!obj)
+			return NULL;
+		if (obj->type == tree_type)
+			return (struct tree *) obj;
+		else if (obj->type == commit_type)
+			obj = &(((struct commit *) obj)->tree->object);
+		else if (obj->type == tag_type)
+			obj = ((struct tag *) obj)->tagged;
+		else
+			return NULL;
+		if (!obj->parsed)
+			parse_object(obj->sha1);
+	} while (1);
+}

@@ -32,4 +32,7 @@ int parse_tree_buffer(struct tree *item,
 
 int parse_tree(struct tree *tree);
 
+/* Parses and returns the tree in the given ent, chasing tags and commits. */
+struct tree *parse_tree_indirect(const unsigned char *sha1);
+
 #endif /* TREE_H */

@@ -184,6 +184,17 @@ struct object_list *object_list_insert(s
         return new_list;
 }
 
+void object_list_append(struct object *item,
+			struct object_list **list_p)
+{
+	while (*list_p) {
+		list_p = &((*list_p)->next);
+	}
+	*list_p = xmalloc(sizeof(struct object_list));
+	(*list_p)->next = NULL;
+	(*list_p)->item = item;
+}
+
 unsigned object_list_length(struct object_list *list)
 {
 	unsigned ret = 0;

@@ -41,6 +41,9 @@ void mark_reachable(struct object *obj, 
 struct object_list *object_list_insert(struct object *item, 
 				       struct object_list **list_p);
 
+void object_list_append(struct object *item,
+			struct object_list **list_p);
+
 unsigned object_list_length(struct object_list *list);
 
 int object_list_contains(struct object_list *list, struct object *obj);

@@ -5,73 +5,291 @@
  */
 #include "cache.h"
 
-static int stage = 0;
+#include "object.h"
+#include "tree.h"
+
+static int merge = 0;
 static int update = 0;
 
-static int unpack_tree(unsigned char *sha1)
-{
-	void *buffer;
-	unsigned long size;
-	int ret;
+static int head_idx = -1;
+static int merge_size = 0;
 
-	buffer = read_object_with_reference(sha1, "tree", &size, NULL);
-	if (!buffer)
-		return -1;
-	ret = read_tree(buffer, size, stage, NULL);
-	free(buffer);
+static struct object_list *trees = NULL;
+
+static struct cache_entry df_conflict_entry = { 
+};
+
+static struct tree_entry_list df_conflict_list = {
+	.name = NULL,
+	.next = &df_conflict_list
+};
+
+typedef int (*merge_fn_t)(struct cache_entry **src);
+
+static int entcmp(char *name1, int dir1, char *name2, int dir2)
+{
+	int len1 = strlen(name1);
+	int len2 = strlen(name2);
+	int len = len1 < len2 ? len1 : len2;
+	int ret = memcmp(name1, name2, len);
+	unsigned char c1, c2;
+	if (ret)
+		return ret;
+	c1 = name1[len];
+	c2 = name2[len];
+	if (!c1 && dir1)
+		c1 = '/';
+	if (!c2 && dir2)
+		c2 = '/';
+	ret = (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0;
+	if (c1 && c2 && !ret)
+		ret = len1 - len2;
 	return ret;
 }
 
-static int path_matches(struct cache_entry *a, struct cache_entry *b)
+static int unpack_trees_rec(struct tree_entry_list **posns, int len,
+			    const char *base, merge_fn_t fn, int *indpos)
 {
-	int len = ce_namelen(a);
-	return ce_namelen(b) == len &&
-		!memcmp(a->name, b->name, len);
+	int baselen = strlen(base);
+	int src_size = len + 1;
+	do {
+		int i;
+		char *first;
+		int firstdir = 0;
+		int pathlen;
+		unsigned ce_size;
+		struct tree_entry_list **subposns;
+		struct cache_entry **src;
+		int any_files = 0;
+		int any_dirs = 0;
+		char *cache_name;
+		int ce_stage;
+
+		/* Find the first name in the input. */
+
+		first = NULL;
+		cache_name = NULL;
+
+		/* Check the cache */
+		if (merge && *indpos < active_nr) {
+			/* This is a bit tricky: */
+			/* If the index has a subdirectory (with
+			 * contents) as the first name, it'll get a
+			 * filename like "foo/bar". But that's after
+			 * "foo", so the entry in trees will get
+			 * handled first, at which point we'll go into
+			 * "foo", and deal with "bar" from the index,
+			 * because the base will be "foo/". The only
+			 * way we can actually have "foo/bar" first of
+			 * all the things is if the trees don't
+			 * contain "foo" at all, in which case we'll
+			 * handle "foo/bar" without going into the
+			 * directory, but that's fine (and will return
+			 * an error anyway, with the added unknown
+			 * file case.
+			 */
+
+			cache_name = active_cache[*indpos]->name;
+			if (strlen(cache_name) > baselen &&
+			    !memcmp(cache_name, base, baselen)) {
+				cache_name += baselen;
+				first = cache_name;
+			} else {
+				cache_name = NULL;
+			}
+		}
+
+		if (first)
+			printf("index %s\n", first);
+
+		for (i = 0; i < len; i++) {
+			if (!posns[i] || posns[i] == &df_conflict_list)
+				continue;
+			printf("%d %s\n", i + 1, posns[i]->name);
+			if (!first || entcmp(first, firstdir,
+					     posns[i]->name, 
+					     posns[i]->directory) > 0) {
+				first = posns[i]->name;
+				firstdir = posns[i]->directory;
+			}
+		}
+		/* No name means we're done */
+		if (!first)
+			return 0;
+
+		pathlen = strlen(first);
+		ce_size = cache_entry_size(baselen + pathlen);
+
+		src = xmalloc(sizeof(struct cache_entry *) * src_size);
+		memset(src, 0, sizeof(struct cache_entry *) * src_size);
+
+		subposns = xmalloc(sizeof(struct tree_list_entry *) * len);
+		memset(subposns, 0, sizeof(struct tree_list_entry *) * len);
+
+		if (cache_name && !strcmp(cache_name, first)) {
+			any_files = 1;
+			src[0] = active_cache[*indpos];
+			remove_cache_entry_at(*indpos);
+		}
+
+		for (i = 0; i < len; i++) {
+			struct cache_entry *ce;
+
+			if (!posns[i] ||
+			    (posns[i] != &df_conflict_list &&
+			     strcmp(first, posns[i]->name))) {
+				continue;
+			}
+
+			if (posns[i] == &df_conflict_list) {
+				src[i + merge] = &df_conflict_entry;
+				continue;
+			}
+
+			if (posns[i]->directory) {
+				any_dirs = 1;
+				parse_tree(posns[i]->item.tree);
+				subposns[i] = posns[i]->item.tree->entries;
+				posns[i] = posns[i]->next;
+				src[i + merge] = &df_conflict_entry;
+				continue;
+			}
+
+			if (!merge)
+				ce_stage = 0;
+			else if (i + 1 < head_idx)
+				ce_stage = 1;
+			else if (i + 1 > head_idx)
+				ce_stage = 3;
+			else
+				ce_stage = 2;
+
+			ce = xmalloc(ce_size);
+			memset(ce, 0, ce_size);
+			ce->ce_mode = create_ce_mode(posns[i]->mode);
+			ce->ce_flags = create_ce_flags(baselen + pathlen,
+						       ce_stage);
+			memcpy(ce->name, base, baselen);
+			memcpy(ce->name + baselen, first, pathlen + 1);
+
+			any_files = 1;
+
+			memcpy(ce->sha1, posns[i]->item.any->sha1, 20);
+			src[i + merge] = ce;
+			subposns[i] = &df_conflict_list;
+			posns[i] = posns[i]->next;
+		}
+		if (any_files) {
+			if (merge) {
+				int ret;
+
+				printf("%s:\n", first);
+				for (i = 0; i < src_size; i++) {
+					printf(" %d ", i);
+					if (src[i])
+						printf("%s\n", sha1_to_hex(src[i]->sha1));
+					else
+						printf("\n");
+				}
+
+				ret = fn(src);
+				
+				printf("Added %d entries\n", ret);
+
+				*indpos += ret;
+			} else {
+				for (i = 0; i < src_size; i++) {
+					if (src[i]) {
+						add_cache_entry(src[i], ADD_CACHE_OK_TO_ADD|ADD_CACHE_SKIP_DFCHECK);
+					}
+				}
+			}
+		}
+		if (any_dirs) {
+			char *newbase = xmalloc(baselen + 2 + pathlen);
+			memcpy(newbase, base, baselen);
+			memcpy(newbase + baselen, first, pathlen);
+			newbase[baselen + pathlen] = '/';
+			newbase[baselen + pathlen + 1] = '\0';
+			if (unpack_trees_rec(subposns, len, newbase, fn,
+					     indpos))
+				return -1;
+		}
+		free(subposns);
+		free(src);
+	} while (1);
 }
 
-static int same(struct cache_entry *a, struct cache_entry *b)
+static void reject_merge(struct cache_entry *ce)
 {
-	return a->ce_mode == b->ce_mode && 
-		!memcmp(a->sha1, b->sha1, 20);
+	die("Entry '%s' would be overwritten by merge. Cannot merge.", 
+	    ce->name);
 }
 
+static void check_updates(struct cache_entry **src, int nr)
+{
+	static struct checkout state = {
+		.base_dir = "",
+		.force = 1,
+		.quiet = 1,
+		.refresh_cache = 1,
+	};
+	unsigned short mask = htons(CE_UPDATE);
+	while (nr--) {
+		struct cache_entry *ce = *src++;
+		if (!ce->ce_mode) {
+			if (update)
+				unlink(ce->name);
+			continue;
+		}
+		if (ce->ce_flags & mask) {
+			ce->ce_flags &= ~mask;
+			if (update)
+				checkout_entry(ce, &state);
+		}
+	}
+}
 
-/*
- * This removes all trivial merges that don't change the tree
- * and collapses them to state 0.
- */
-static struct cache_entry *merge_entries(struct cache_entry *a,
-					 struct cache_entry *b,
-					 struct cache_entry *c)
+static int unpack_trees(merge_fn_t fn)
 {
-	/*
-	 * Ok, all three entries describe the same
-	 * filename, but maybe the contents or file
-	 * mode have changed?
-	 *
-	 * The trivial cases end up being the ones where two
-	 * out of three files are the same:
-	 *  - both destinations the same, trivially take either
-	 *  - one of the destination versions hasn't changed,
-	 *    take the other.
-	 *
-	 * The "all entries exactly the same" case falls out as
-	 * a special case of any of the "two same" cases.
-	 *
-	 * Here "a" is "original", and "b" and "c" are the two
-	 * trees we are merging.
-	 */
-	if (a && b && c) {
-		if (same(b,c))
-			return c;
-		if (same(a,b))
-			return c;
-		if (same(a,c))
-			return b;
+	int indpos = 0;
+	unsigned len = object_list_length(trees);
+	struct tree_entry_list **posns = 
+		xmalloc(len * sizeof(struct tree_entry_list *));
+	int i;
+	struct object_list *posn = trees;
+	merge_size = len;
+	for (i = 0; i < len; i++) {
+		posns[i] = ((struct tree *) posn->item)->entries;
+		posn = posn->next;
 	}
-	return NULL;
+	if (unpack_trees_rec(posns, len, "", fn, &indpos))
+		return -1;
+
+	check_updates(active_cache, active_nr);
+	return 0;
 }
 
+static int list_tree(unsigned char *sha1)
+{
+	struct tree *tree = parse_tree_indirect(sha1);
+	if (!tree)
+		return -1;
+	object_list_append(&tree->object, &trees);
+	return 0;
+}
+
+static int same(struct cache_entry *a, struct cache_entry *b)
+{
+	if (!!a != !!b)
+		return 0;
+	if (!a && !b)
+		return 1;
+	return a->ce_mode == b->ce_mode && 
+		!memcmp(a->sha1, b->sha1, 20);
+}
+
+
 /*
  * When a CE gets turned into an unmerged entry, we
  * want it to be up-to-date

@@ -91,18 +309,7 @@ static void verify_uptodate(struct cache
 	die("Entry '%s' not uptodate. Cannot merge.", ce->name);
 }
 
-/*
- * If the old tree contained a CE that isn't even in the
- * result, that's always a problem, regardless of whether
- * it's up-to-date or not (ie it can be a file that we
- * have updated but not committed yet).
- */
-static void reject_merge(struct cache_entry *ce)
-{
-	die("Entry '%s' would be overwritten by merge. Cannot merge.", ce->name);
-}
-
-static int merged_entry_internal(struct cache_entry *merge, struct cache_entry *old, struct cache_entry **dst, int allow_dirty)
+static int merged_entry(struct cache_entry *merge, struct cache_entry *old)
 {
 	merge->ce_flags |= htons(CE_UPDATE);
 	if (old) {

@@ -115,195 +322,129 @@ static int merged_entry_internal(struct 
 		 */
 		if (same(old, merge)) {
 			*merge = *old;
-		} else if (!allow_dirty) {
+		} else {
 			verify_uptodate(old);
 		}
 	}
 	merge->ce_flags &= ~htons(CE_STAGEMASK);
-	*dst++ = merge;
+	add_cache_entry(merge, ADD_CACHE_OK_TO_ADD);
 	return 1;
 }
 
-static int merged_entry_allow_dirty(struct cache_entry *merge, struct cache_entry *old, struct cache_entry **dst)
-{
-	return merged_entry_internal(merge, old, dst, 1);
-}
-
-static int merged_entry(struct cache_entry *merge, struct cache_entry *old, struct cache_entry **dst)
-{
-	return merged_entry_internal(merge, old, dst, 0);
-}
-
-static int deleted_entry(struct cache_entry *ce, struct cache_entry *old, struct cache_entry **dst)
+static int deleted_entry(struct cache_entry *ce, struct cache_entry *old)
 {
 	if (old)
 		verify_uptodate(old);
 	ce->ce_mode = 0;
-	*dst++ = ce;
+	add_cache_entry(ce, ADD_CACHE_OK_TO_ADD);
 	return 1;
 }
 
-static int causes_df_conflict(struct cache_entry *ce, int stage,
-			      struct cache_entry **dst_,
-			      struct cache_entry **next_,
-			      int tail)
-{
-	/* This is called during the merge operation and walking
-	 * the active_cache[] array is messy, because it is in the
-	 * middle of overlapping copy operation.  The invariants
-	 * are:
-	 * (1) active_cache points at the first (zeroth) entry.
-	 * (2) up to dst pointer are resolved entries.
-	 * (3) from the next pointer (head-inclusive) to the tail
-	 *     of the active_cache array have the remaining paths
-	 *     to be processed.  There can be a gap between dst
-	 *     and next.  Note that next is called "src" in the
-	 *     merge_cache() function, and tail is the original
-	 *     end of active_cache array when merge_cache() started.
-	 * (4) the path corresponding to *ce is not found in (2)
-	 *     or (3).  It is in the gap.
-	 *
-	 *  active_cache -----......+++++++++++++.
-	 *                    ^dst  ^next        ^tail
-	 */
-	int i, next, dst;
-	const char *path = ce->name;
-	int namelen = ce_namelen(ce);
-
-	next = next_ - active_cache;
-	dst = dst_ - active_cache;
-
-	for (i = 0; i < tail; i++) {
-		int entlen, len;
-		const char *one, *two;
-		if (dst <= i && i < next)
-			continue;
-		ce = active_cache[i];
-		if (ce_stage(ce) != stage)
-			continue;
-		/* If ce->name is a prefix of path, then path is a file
-		 * that hangs underneath ce->name, which is bad.
-		 * If path is a prefix of ce->name, then it is the
-		 * other way around which also is bad.
-		 */
-		entlen = ce_namelen(ce);
-		if (namelen == entlen)
-			continue;
-		if (namelen < entlen) {
-			len = namelen;
-			one = path;
-			two = ce->name;
-		} else {
-			len = entlen;
-			one = ce->name;
-			two = path;
-		}
-		if (memcmp(one, two, len))
-			continue;
-		if (two[len] == '/')
-			return 1;
-	}
-	return 0;
+static int keep_entry(struct cache_entry *ce)
+{
+	add_cache_entry(ce, ADD_CACHE_OK_TO_ADD);
+	return 1;
 }
 
-static int threeway_merge(struct cache_entry *stages[4],
-			  struct cache_entry **dst,
-			  struct cache_entry **next, int tail)
-{
-	struct cache_entry *old = stages[0];
-	struct cache_entry *a = stages[1], *b = stages[2], *c = stages[3];
-	struct cache_entry *merge;
+static int threeway_merge(struct cache_entry **stages)
+{
+	struct cache_entry *index;
+	struct cache_entry *head; 
+	struct cache_entry *remote = stages[head_idx + 1];
 	int count;
+	int head_match = 0;
+	int remote_match = 0;
 
-	/* #5ALT */
-	if (!a && b && c && same(b, c)) {
-		if (old && !same(b, old))
-			return -1;
-		return merged_entry_allow_dirty(b, old, dst);
+	int df_conflict_head = 0;
+	int df_conflict_remote = 0;
+
+	int any_anc_missing = 0;
+	int i;
+
+	for (i = 1; i < head_idx; i++) {
+		if (!stages[i])
+			any_anc_missing = 1;
 	}
-	/* #2ALT and #3ALT */
-	if (!a && (!!b != !!c)) {
-		/*
-		 * The reason we need to worry about directory/file
-		 * conflicts only in #2ALT and #3ALT case is this:
-		 *
-		 * (1) For all other cases that read-tree internally
-		 *     resolves a path, we always have such a path in
-		 *     *both* stage2 and stage3 when we begin.
-		 *     Traditionally, the behaviour has been even
-		 *     stricter and we did not resolve a path without
-		 *     initially being in all of stage1, 2, and 3.
-		 *
-		 * (2) When read-tree finishes, all resolved paths (i.e.
-		 *     the paths that are in stage0) must have come from
-		 *     either stage2 or stage3.  It is not possible to
-		 *     have a stage0 path as a result of a merge if
-		 *     neither stage2 nor stage3 had that path.
-		 *
-		 * (3) It is guaranteed that just after reading the
-		 *     stages, each stage cannot have directory/file
-		 *     conflicts on its own, because they are populated
-		 *     by reading hierarchy of a tree.  Combined with
-		 *     (1) and (2) above, this means that no matter what
-		 *     combination of paths we take from stage2 and
-		 *     stage3 as a result of a merge, they cannot cause
-		 *     a directory/file conflict situation (otherwise
-		 *     the "guilty" path would have already had such a
-		 *     conflict in the original stage, either stage2
-		 *     or stage3).  Although its stage2 is synthesized
-		 *     by overlaying the current index on top of "our
-		 *     head" tree, --emu23 case also has this guarantee,
-		 *     by calling add_cache_entry() to create such stage2
-		 *     entries.
-		 *
-		 * (4) Only #2ALT and #3ALT lack the guarantee (1).
-		 *     They resolve paths that exist only in stage2
-		 *     or stage3.  The stage2 tree may have a file DF
-		 *     while stage3 tree may have a file DF/DF.  If
-		 *     #2ALT and #3ALT rules happen to apply to both
-		 *     of them, we would end up having DF (coming from
-		 *     stage2) and DF/DF (from stage3) in the result.
-		 *     When we attempt to resolve a path that exists
-		 *     only in stage2, we need to make sure there is
-		 *     no path that would conflict with it in stage3
-		 *     and vice versa.
-		 */
-		if (c) { /* #2ALT */
-			if (!causes_df_conflict(c, 2, dst, next, tail) &&
-			    (!old || same(c, old)))
-				return merged_entry_allow_dirty(c, old, dst);
-		}
-		else { /* #3ALT */
-			if (!causes_df_conflict(b, 3, dst, next, tail) &&
-			    (!old || same(b, old)))
-				return merged_entry_allow_dirty(b, old, dst);
-		}
-		/* otherwise we will apply the original rule */
-	}
-	/* #14ALT */
-	if (a && b && c && same(a, b) && !same(a, c)) {
-		if (old && same(old, c))
-			return merged_entry_allow_dirty(c, old, dst);
-		/* otherwise the regular rule applies */
+
+	index = stages[0];
+	head = stages[head_idx];
+
+	if (head == &df_conflict_entry) {
+		df_conflict_head = 1;
+		head = NULL;
+	}
+
+	if (remote == &df_conflict_entry) {
+		df_conflict_remote = 1;
+		remote = NULL;
+	}
+
+	/* First, if there's a #16 situation, note that to prevent #13
+	 * and #14. 
+	 */
+	if (!same(remote, head)) {
+		for (i = 1; i < head_idx; i++) {
+			if (same(stages[i], head)) {
+				head_match = 1;
+			}
+			if (same(stages[i], remote)) {
+				remote_match = 1;
+			}
+		}
+	}
+
+	/* We start with cases where the index is allowed to match
+	 * something other than the head: #14(ALT) and #2ALT, where it
+	 * is permitted to match the result instead.
+	 */
+	/* #14, #14ALT, #2ALT */
+	if (remote && !df_conflict_head && head_match && !remote_match) {
+		if (index && !same(index, remote) && !same(index, head))
+			reject_merge(index);
+		return merged_entry(remote, index);
 	}
 	/*
-	 * If we have an entry in the index cache ("old"), then we want
-	 * to make sure that it matches any entries in stage 2 ("first
-	 * branch", aka "b").
+	 * If we have an entry in the index cache, then we want to
+	 * make sure that it matches head.
 	 */
-	if (old) {
-		if (!b || !same(old, b))
-			return -1;
+	if (index && !same(index, head)) {
+		reject_merge(index);
 	}
-	merge = merge_entries(a, b, c);
-	if (merge)
-		return merged_entry(merge, old, dst);
-	if (old)
-		verify_uptodate(old);
+
+	if (head) {
+		/* #5ALT, #15 */
+		if (same(head, remote))
+			return merged_entry(head, index);
+		/* #13, #3ALT */
+		if (!df_conflict_remote && remote_match && !head_match)
+			return merged_entry(head, index);
+	}
+
+	/* #1 */
+	if (!head && !remote && any_anc_missing)
+		return 0;
+
+	/* Below are "no merge" cases, which require that the index be
+	 * up-to-date to avoid the files getting overwritten with
+	 * conflict resolution files. 
+	 */
+	if (index) {
+		verify_uptodate(index);
+	}
+
+	/* #2, #3, #4, #6, #7, #9, #11. */
 	count = 0;
-	if (a) { *dst++ = a; count++; }
-	if (b) { *dst++ = b; count++; }
-	if (c) { *dst++ = c; count++; }
+	if (!head_match || !remote_match) {
+		for (i = 1; i < head_idx; i++) {
+			if (stages[i]) {
+				keep_entry(stages[i]);
+				count++;
+				break;
+			}
+		}
+	}
+	if (head) { count += keep_entry(head); }
+	if (remote) { count += keep_entry(remote); }
 	return count;
 }