Code

[PATCH] Rewrite read-tree
authorDaniel Barkalow <barkalow@iabervon.org>
Mon, 5 Sep 2005 06:04:48 +0000 (02:04 -0400)
committerJunio C Hamano <junkio@cox.net>
Sun, 11 Sep 2005 01:27:41 +0000 (18:27 -0700)
Adds support for multiple ancestors, removes --emu23, much simplification.

Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
read-tree.c
t/t1005-read-tree-m-2way-emu23.sh [deleted file]

index ce2bb88f0aae8919f188a0dbc5f3338523151740..0241c9c6f55c2345d5604ddbb93baee53d11ea9a 100644 (file)
  */
 #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_entry *ce)
        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 cache_entry *merge, struct cache_entry *
                 */
                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)
+static int keep_entry(struct cache_entry *ce)
 {
-       /* 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;
+       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)
+static int threeway_merge(struct cache_entry **stages)
 {
-       struct cache_entry *old = stages[0];
-       struct cache_entry *a = stages[1], *b = stages[2], *c = stages[3];
-       struct cache_entry *merge;
+       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);
+
+       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;
+                       }
                }
-               /* 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 */
+
+       /* 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;
 }
 
@@ -316,14 +457,14 @@ static int threeway_merge(struct cache_entry *stages[4],
  * "carry forward" rule, please see <Documentation/git-read-tree.txt>.
  *
  */
-static int twoway_merge(struct cache_entry **src, struct cache_entry **dst,
-                       struct cache_entry **next, int tail)
+static int twoway_merge(struct cache_entry **src)
 {
        struct cache_entry *current = src[0];
        struct cache_entry *oldtree = src[1], *newtree = src[2];
 
-       if (src[3])
-               return -1;
+       if (merge_size != 2)
+               return error("Cannot do a twoway merge of %d trees\n",
+                            merge_size);
 
        if (current) {
                if ((!oldtree && !newtree) || /* 4 and 5 */
@@ -334,78 +475,32 @@ static int twoway_merge(struct cache_entry **src, struct cache_entry **dst,
                    (oldtree && newtree &&
                     !same(oldtree, newtree) && /* 18 and 19*/
                     same(current, newtree))) {
-                       *dst++ = current;
-                       return 1;
+                       return keep_entry(current);
                }
                else if (oldtree && !newtree && same(current, oldtree)) {
                        /* 10 or 11 */
-                       return deleted_entry(oldtree, current, dst);
+                       return deleted_entry(oldtree, current);
                }
                else if (oldtree && newtree &&
                         same(current, oldtree) && !same(current, newtree)) {
                        /* 20 or 21 */
-                       return merged_entry(newtree, current, dst);
+                       return merged_entry(newtree, current);
                }
-               else
+               else {
                        /* all other failures */
+                       if (oldtree)
+                               reject_merge(oldtree);
+                       if (current)
+                               reject_merge(current);
+                       if (newtree)
+                               reject_merge(newtree);
                        return -1;
+               }
        }
        else if (newtree)
-               return merged_entry(newtree, current, dst);
+               return merged_entry(newtree, current);
        else
-               return deleted_entry(oldtree, current, dst);
-}
-
-/*
- * Two-way merge emulated with three-way merge.
- *
- * This treats "read-tree -m H M" by transforming it internally
- * into "read-tree -m H I+H M", where I+H is a tree that would
- * contain the contents of the current index file, overlayed on
- * top of H.  Unlike the traditional two-way merge, this leaves
- * the stages in the resulting index file and lets the user resolve
- * the merge conflicts using standard tools for three-way merge.
- *
- * This function is just to set-up such an arrangement, and the
- * actual merge uses threeway_merge() function.
- */
-static void setup_emu23(void)
-{
-       /* stage0 contains I, stage1 H, stage2 M.
-        * move stage2 to stage3, and create stage2 entries
-        * by scanning stage0 and stage1 entries.
-        */
-       int i, namelen, size;
-       struct cache_entry *ce, *stage2;
-
-       for (i = 0; i < active_nr; i++) {
-               ce = active_cache[i];
-               if (ce_stage(ce) != 2)
-                       continue;
-               /* hoist them up to stage 3 */
-               namelen = ce_namelen(ce);
-               ce->ce_flags = create_ce_flags(namelen, 3);
-       }
-
-       for (i = 0; i < active_nr; i++) {
-               ce = active_cache[i];
-               if (ce_stage(ce) > 1)
-                       continue;
-               namelen = ce_namelen(ce);
-               size = cache_entry_size(namelen);
-               stage2 = xmalloc(size);
-               memcpy(stage2, ce, size);
-               stage2->ce_flags = create_ce_flags(namelen, 2);
-               if (add_cache_entry(stage2, ADD_CACHE_OK_TO_ADD) < 0)
-                       die("cannot merge index and our head tree");
-
-               /* We are done with this name, so skip to next name */
-               while (i < active_nr &&
-                      ce_namelen(active_cache[i]) == namelen &&
-                      !memcmp(active_cache[i]->name, ce->name, namelen))
-                       i++;
-               i--; /* compensate for the loop control */
-       }
+               return deleted_entry(oldtree, current);
 }
 
 /*
@@ -414,78 +509,21 @@ static void setup_emu23(void)
  * The rule is:
  * - take the stat information from stage0, take the data from stage1
  */
-static int oneway_merge(struct cache_entry **src, struct cache_entry **dst,
-                       struct cache_entry **next, int tail)
+static int oneway_merge(struct cache_entry **src)
 {
        struct cache_entry *old = src[0];
        struct cache_entry *a = src[1];
 
-       if (src[2] || src[3])
-               return -1;
+       if (merge_size != 1)
+               return error("Cannot do a oneway merge of %d trees\n",
+                            merge_size);
 
        if (!a)
                return 0;
        if (old && same(old, a)) {
-               *dst++ = old;
-               return 1;
+               return keep_entry(old);
        }
-       return merged_entry(a, NULL, dst);
-}
-
-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);
-               }
-       }
-}
-
-typedef int (*merge_fn_t)(struct cache_entry **, struct cache_entry **, struct cache_entry **, int);
-
-static void merge_cache(struct cache_entry **src, int nr, merge_fn_t fn)
-{
-       struct cache_entry **dst = src;
-       int tail = nr;
-
-       while (nr) {
-               int entries;
-               struct cache_entry *name, *ce, *stages[4] = { NULL, };
-
-               name = ce = *src;
-               for (;;) {
-                       int stage = ce_stage(ce);
-                       stages[stage] = ce;
-                       ce = *++src;
-                       active_nr--;
-                       if (!--nr)
-                               break;
-                       if (!path_matches(ce, name))
-                               break;
-               }
-
-               entries = fn(stages, dst, src, tail);
-               if (entries < 0)
-                       reject_merge(name);
-               dst += entries;
-               active_nr += entries;
-       }
-       check_updates(active_cache, active_nr);
+       return merged_entry(a, NULL);
 }
 
 static int read_cache_unmerged(void)
@@ -516,8 +554,9 @@ static struct cache_file cache_file;
 
 int main(int argc, char **argv)
 {
-       int i, newfd, merge, reset, emu23;
+       int i, newfd, reset, stage = 0;
        unsigned char sha1[20];
+       merge_fn_t fn = NULL;
 
        newfd = hold_index_file_for_update(&cache_file, get_index_file());
        if (newfd < 0)
@@ -525,7 +564,6 @@ int main(int argc, char **argv)
 
        merge = 0;
        reset = 0;
-       emu23 = 0;
        for (i = 1; i < argc; i++) {
                const char *arg = argv[i];
 
@@ -537,7 +575,7 @@ int main(int argc, char **argv)
 
                /* This differs from "-m" in that we'll silently ignore unmerged entries */
                if (!strcmp(arg, "--reset")) {
-                       if (stage || merge || emu23)
+                       if (stage || merge)
                                usage(read_tree_usage);
                        reset = 1;
                        merge = 1;
@@ -546,9 +584,14 @@ int main(int argc, char **argv)
                        continue;
                }
 
+               if (!strcmp(arg, "--head")) {
+                       head_idx = stage - 1;
+                       fn = threeway_merge;
+               }
+
                /* "-m" stands for "merge", meaning we start in stage 1 */
                if (!strcmp(arg, "-m")) {
-                       if (stage || merge || emu23)
+                       if (stage || merge)
                                usage(read_tree_usage);
                        if (read_cache_unmerged())
                                die("you need to resolve your current index first");
@@ -557,45 +600,41 @@ int main(int argc, char **argv)
                        continue;
                }
 
-               /* "-emu23" uses 3-way merge logic to perform fast-forward */
-               if (!strcmp(arg, "--emu23")) {
-                       if (stage || merge || emu23)
-                               usage(read_tree_usage);
-                       if (read_cache_unmerged())
-                               die("you need to resolve your current index first");
-                       merge = emu23 = stage = 1;
-                       continue;
-               }
-
                if (get_sha1(arg, sha1) < 0)
                        usage(read_tree_usage);
-               if (stage > 3)
-                       usage(read_tree_usage);
-               if (unpack_tree(sha1) < 0)
+               if (list_tree(sha1) < 0)
                        die("failed to unpack tree object %s", arg);
                stage++;
        }
        if (update && !merge)
                usage(read_tree_usage);
-       if (merge) {
-               static const merge_fn_t merge_function[] = {
-                       [1] = oneway_merge,
-                       [2] = twoway_merge,
-                       [3] = threeway_merge,
-               };
-               merge_fn_t fn;
-
-               if (stage < 2 || stage > 4)
+       if (merge && !fn) {
+               if (stage < 2)
                        die("just how do you expect me to merge %d trees?", stage-1);
-               if (emu23 && stage != 3)
-                       die("--emu23 takes only two trees");
-               fn = merge_function[stage-1];
-               if (stage == 3 && emu23) { 
-                       setup_emu23();
-                       fn = merge_function[3];
+               switch (stage - 1) {
+               case 1:
+                       fn = oneway_merge;
+                       break;
+               case 2:
+                       fn = twoway_merge;
+                       break;
+               case 3:
+                       fn = threeway_merge;
+                       break;
+               default:
+                       fn = threeway_merge;
+                       break;
                }
-               merge_cache(active_cache, active_nr, fn);
        }
+
+       if (head_idx < 0) {
+               if (stage - 1 >= 3)
+                       head_idx = stage - 2;
+               else
+                       head_idx = 1;
+       }
+
+       unpack_trees(fn);
        if (write_cache(newfd, active_cache, active_nr) ||
            commit_index_file(&cache_file))
                die("unable to write new index file");
diff --git a/t/t1005-read-tree-m-2way-emu23.sh b/t/t1005-read-tree-m-2way-emu23.sh
deleted file mode 100755 (executable)
index 258a372..0000000
+++ /dev/null
@@ -1,422 +0,0 @@
-#!/bin/sh
-#
-# Copyright (c) 2005 Junio C Hamano
-#
-
-test_description='Two way merge with read-tree --emu23 $H $M
-
-This test tries two-way merge (aka fast forward with carry forward).
-
-There is the head (called H) and another commit (called M), which is
-simply ahead of H.  The index and the work tree contains a state that
-is derived from H, but may also have local changes.  This test checks
-all the combinations described in the two-tree merge "carry forward"
-rules, found in <Documentation/git-rev-tree.txt>.
-
-In the test, these paths are used:
-        bozbar  - in H, stays in M, modified from bozbar to gnusto
-        frotz   - not in H added in M
-        nitfol  - in H, stays in M unmodified
-        rezrov  - in H, deleted in M
-        yomin   - not in H nor M
-'
-. ./test-lib.sh
-
-read_tree_twoway () {
-    git-read-tree --emu23 "$1" "$2" &&
-    git-ls-files --stage &&
-    git-merge-index git-merge-one-file -a &&
-    git-ls-files --stage
-}
-
-_x40='[0-9a-f][0-9a-f][0-9a-f][0-9a-f][0-9a-f]'
-_x40="$_x40$_x40$_x40$_x40$_x40$_x40$_x40$_x40"
-compare_change () {
-       sed -n >current \
-           -e '/^--- /d; /^+++ /d; /^@@ /d;' \
-           -e 's/^\([-+][0-7][0-7][0-7][0-7][0-7][0-7]\) '"$_x40"' /\1 X /p' \
-           "$1"
-       diff -u expected current
-}
-
-check_cache_at () {
-       clean_if_empty=`git-diff-files "$1"`
-       case "$clean_if_empty" in
-       '')  echo "$1: clean" ;;
-       ?*)  echo "$1: dirty" ;;
-       esac
-       case "$2,$clean_if_empty" in
-       clean,)         :     ;;
-       clean,?*)       false ;;
-       dirty,)         false ;;
-       dirty,?*)       :     ;;
-       esac
-}
-
-check_stages () {
-    cat >expected_stages
-    git-ls-files --stage | sed -e "s/ $_x40 / X /" >current_stages
-    diff -u expected_stages current_stages
-}
-
-cat >bozbar-old <<\EOF
-This is a sample file used in two-way fast forward merge
-tests.  Its second line ends with a magic word bozbar
-which will be modified by the merged head to gnusto.
-It has some extra lines so that external tools can
-successfully merge independent changes made to later
-lines (such as this one), avoiding line conflicts.
-EOF
-
-sed -e 's/bozbar/gnusto (earlier bozbar)/' bozbar-old >bozbar-new
-
-test_expect_success \
-    setup \
-    'echo frotz >frotz &&
-     echo nitfol >nitfol &&
-     cat bozbar-old >bozbar &&
-     echo rezrov >rezrov &&
-     echo yomin >yomin &&
-     git-update-index --add nitfol bozbar rezrov &&
-     treeH=`git-write-tree` &&
-     echo treeH $treeH &&
-     git-ls-tree $treeH &&
-
-     cat bozbar-new >bozbar &&
-     git-update-index --add frotz bozbar --force-remove rezrov &&
-     git-ls-files --stage >M.out &&
-     treeM=`git-write-tree` &&
-     echo treeM $treeM &&
-     git-ls-tree $treeM &&
-     git-diff-tree $treeH $treeM'
-
-# "read-tree -m H I+H M" but I is empty so this is "read-tree -m H H M".
-#
-# bozbar [O && A && B && O==A && O!=B (#14) ==> B] take M by read-tree
-# frotz  [!O && !A && B (#2) ==> B]                take M by read-tree
-# nitfol [O && A && B && O==A && O==B (#15) ==> B] take M by read-tree
-# rezrov [O && A && !B && O==A (#10) ==> no merge] removed by script
-#
-# Earlier one did not have #2ALT so taking M was done by the script,
-# which also updated the work tree and making frotz clean.  With #2ALT,
-# this is resolved by read-tree itself and the path is left dirty
-# because we are not testing "read-tree -u --emu23".
-test_expect_success \
-    '1, 2, 3 - no carry forward' \
-    'rm -f .git/index &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >1-3.out &&
-     diff -u M.out 1-3.out &&
-     check_cache_at bozbar dirty &&
-     check_cache_at frotz dirty && # same as pure 2-way again.
-     check_cache_at nitfol dirty'
-
-echo '+100644 X 0      yomin' >expected
-
-test_expect_success \
-    '4 - carry forward local addition.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     git-update-index --add yomin &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >4.out || return 1
-     diff -u M.out 4.out >4diff.out
-     compare_change 4diff.out expected &&
-     check_cache_at yomin clean'
-
-# "read-tree -m H I+H M" where !H && !M; so (I+H) not being up-to-date
-# should not matter.  Thanks to #3ALT, this is now possible.
-test_expect_success \
-    '5 - carry forward local addition.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo yomin >yomin &&
-     git-update-index --add yomin &&
-     echo yomin yomin >yomin &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >5.out || return 1
-     diff -u M.out 5.out >5diff.out
-     compare_change 5diff.out expected &&
-     check_cache_at yomin dirty'
-
-# "read-tree -m H I+H M" where !H && M && (I+H) == M, so this should
-# succeed (even the entry is clean), now thanks to #5ALT.
-test_expect_success \
-    '6 - local addition already has the same.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     git-update-index --add frotz &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >6.out &&
-     diff -u M.out 6.out &&
-     check_cache_at frotz clean'
-
-# Exactly the same pattern as above but with dirty cache.  This also
-# should succeed, now thanks to #5ALT.
-test_expect_success \
-    '7 - local addition already has the same.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo frotz >frotz &&
-     git-update-index --add frotz &&
-     echo frotz frotz >frotz &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >7.out &&
-     diff -u M.out 7.out &&
-     check_cache_at frotz dirty'
-
-test_expect_success \
-    '8 - conflicting addition.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo frotz frotz >frotz &&
-     git-update-index --add frotz &&
-     if read_tree_twoway $treeH $treeM; then false; else :; fi'
-
-test_expect_success \
-    '9 - conflicting addition.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo frotz frotz >frotz &&
-     git-update-index --add frotz &&
-     echo frotz >frotz &&
-     if read_tree_twoway $treeH $treeM; then false; else :; fi'
-
-test_expect_success \
-    '10 - path removed.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo rezrov >rezrov &&
-     git-update-index --add rezrov &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >10.out &&
-     diff -u M.out 10.out'
-
-test_expect_success \
-    '11 - dirty path removed.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo rezrov >rezrov &&
-     git-update-index --add rezrov &&
-     echo rezrov rezrov >rezrov &&
-     if read_tree_twoway $treeH $treeM; then false; else :; fi'
-
-test_expect_success \
-    '12 - unmatching local changes being removed.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo rezrov rezrov >rezrov &&
-     git-update-index --add rezrov &&
-     if read_tree_twoway $treeH $treeM; then false; else :; fi'
-
-test_expect_success \
-    '13 - unmatching local changes being removed.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo rezrov rezrov >rezrov &&
-     git-update-index --add rezrov &&
-     echo rezrov >rezrov &&
-     if read_tree_twoway $treeH $treeM; then false; else :; fi'
-
-cat >expected <<EOF
--100644 X 0    nitfol
-+100644 X 0    nitfol
-EOF
-
-test_expect_success \
-    '14 - unchanged in two heads.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo nitfol nitfol >nitfol &&
-     git-update-index --add nitfol &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >14.out || return 1
-     diff -u M.out 14.out >14diff.out
-     compare_change 14diff.out expected &&
-     check_cache_at nitfol clean'
-
-test_expect_success \
-    '15 - unchanged in two heads.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo nitfol nitfol >nitfol &&
-     git-update-index --add nitfol &&
-     echo nitfol nitfol nitfol >nitfol &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >15.out || return 1
-     diff -u M.out 15.out >15diff.out
-     compare_change 15diff.out expected &&
-     check_cache_at nitfol dirty'
-
-# This is different from straight 2-way merge in that it leaves
-# three stages of bozbar in the index file without failing, so
-# the user can run git-diff-stages to examine the situation.
-# With #2ALT, frotz is resolved internally.
-test_expect_success \
-    '16 - conflicting local change.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo bozbar bozbar >bozbar &&
-     git-update-index --add bozbar &&
-     git-read-tree --emu23 $treeH $treeM &&
-     check_stages' <<\EOF
-100644 X 1     bozbar
-100644 X 2     bozbar
-100644 X 3     bozbar
-100644 X 0     frotz
-100644 X 0     nitfol
-100644 X 1     rezrov
-100644 X 2     rezrov
-EOF
-
-test_expect_success \
-    '17 - conflicting local change.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     echo bozbar bozbar >bozbar &&
-     git-update-index --add bozbar &&
-     echo bozbar bozbar bozbar >bozbar &&
-     if read_tree_twoway $treeH $treeM; then false; else :; fi'
-
-test_expect_success \
-    '18 - local change already having a good result.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     cat bozbar-new >bozbar &&
-     git-update-index --add bozbar &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >18.out &&
-     diff -u M.out 18.out &&
-     check_cache_at bozbar clean'
-
-test_expect_success \
-    '19 - local change already having a good result, further modified.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     cat bozbar-new >bozbar &&
-     git-update-index --add bozbar &&
-     echo gnusto gnusto >bozbar &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >19.out &&
-     diff -u M.out 19.out &&
-     check_cache_at bozbar dirty'
-
-test_expect_success \
-    '20 - no local change, use new tree.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     cat bozbar-old >bozbar &&
-     git-update-index --add bozbar &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >20.out &&
-     diff -u M.out 20.out &&
-     check_cache_at bozbar dirty'
-
-test_expect_success \
-    '21 - no local change, dirty cache.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     cat bozbar-old >bozbar &&
-     git-update-index --add bozbar &&
-     echo gnusto gnusto >bozbar &&
-     if read_tree_twoway $treeH $treeM; then false; else :; fi'
-
-echo '-100644 X 0      bozbar
-+100644 X 0    bozbar' >expected
-
-# This fails with straight two-way fast forward, but emu23
-# can merge them.
-test_expect_success \
-    '22 - local change cache updated.' \
-    'rm -f .git/index &&
-     git-read-tree $treeH &&
-     git-checkout-index -u -f -q -a &&
-     sed -e "s/such as/SUCH AS/" bozbar-old >bozbar &&
-     git-update-index --add bozbar &&
-     read_tree_twoway $treeH $treeM &&
-     git-ls-files --stage >22.out || return 1
-     diff -u M.out 22.out >22diff.out
-     compare_change 22diff.out &&
-     check_cache_at bozbar clean'
-
-# Also make sure we did not break DF vs DF/DF case.
-test_expect_success \
-    'DF vs DF/DF case setup.' \
-    'rm -f .git/index &&
-     echo DF >DF &&
-     git-update-index --add DF &&
-     treeDF=`git-write-tree` &&
-     echo treeDF $treeDF &&
-     git-ls-tree $treeDF &&
-     git-ls-files --stage >DF.out
-
-     rm -f DF &&
-     mkdir DF &&
-     echo DF/DF >DF/DF &&
-     git-update-index --add --remove DF DF/DF &&
-     treeDFDF=`git-write-tree` &&
-     echo treeDFDF $treeDFDF &&
-     git-ls-tree $treeDFDF &&
-     git-ls-files --stage >DFDF.out'
-
-test_expect_success \
-    'DF vs DF/DF case test (#1)' \
-    'rm -f .git/index &&
-     rm -fr DF &&
-     echo DF >DF &&
-     git-update-index --add DF &&
-     read_tree_twoway $treeDF $treeDFDF &&
-     git-ls-files --stage >DFDFcheck.out &&
-     diff -u DFDF.out DFDFcheck.out &&
-     check_cache_at DF/DF clean && # different from pure 2-way
-     :'
-
-# The other way around
-test_expect_success \
-    'DF vs DF/DF case test (#2)' \
-    'rm -f .git/index &&
-     rm -fr DF &&
-     mkdir DF &&
-     echo DF/DF >DF/DF &&
-     git-update-index --add DF/DF &&
-     read_tree_twoway $treeDFDF $treeDF &&
-     git-ls-files --stage >DFDFcheck.out &&
-     diff -u DF.out DFDFcheck.out &&
-     check_cache_at DF clean && # different from pure 2-way
-     :'
-
-# Emu23 can grok I having more than H.  Make sure we did not
-# botch the conflict tests (fixed).
-test_expect_success \
-    'DF vs DF/DF case test (#3).' \
-    'rm -f .git/index &&
-     rm -fr DF &&
-     mkdir DF &&
-     echo DF/DF >DF/DF &&
-     git-update-index --add DF/DF &&
-     # This should fail because I and H have a conflict
-     # at DF.
-     if git-read-tree --emu23 $treeDF $treeDFDF
-     then false
-     else true
-     fi'
-
-test_done