Build-in merge-recursive
authorDaniel Barkalow <barkalow@iabervon.org>
Thu, 7 Feb 2008 16:40:05 +0000 (11:40 -0500)
committerJunio C Hamano <gitster@pobox.com>
Sun, 10 Feb 2008 07:16:51 +0000 (23:16 -0800)
This makes write_tree_from_memory(), which writes the active cache as
a tree and returns the struct tree for it, available to other code. It
also makes available merge_trees(), which does the internal merge of
two trees with a known base, and merge_recursive(), which does the
recursive internal merge of two commits with a list of common
ancestors.

The first two of these will be used by checkout -m, and the third is
presumably useful in general, although the implementation of checkout
-m which entirely matches the behavior of the shell version does not
use it (since it ignores the difference of ancestry between the old
branch and the new branch).

Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
Makefile
builtin-merge-recursive.c [new file with mode: 0644]
builtin.h
git.c
merge-recursive.c [deleted file]
merge-recursive.h [new file with mode: 0644]
index 5aac0c0c876957445cd6e2e7b81a59f40965930b..6f08bccbcfbde4ccd46f1c3e8ca7986567cab889 100644 (file)
--- a/Makefile
+++ b/Makefile
@@ -256,7 +256,6 @@ PROGRAMS = \
        git-upload-pack$X \
        git-pack-redundant$X git-var$X \
        git-merge-tree$X git-imap-send$X \
-       git-merge-recursive$X \
        $(EXTRA_PROGRAMS)
 
 # Empty...
@@ -360,6 +359,7 @@ BUILTIN_OBJS = \
        builtin-merge-base.o \
        builtin-merge-file.o \
        builtin-merge-ours.o \
+       builtin-merge-recursive.o \
        builtin-mv.o \
        builtin-name-rev.o \
        builtin-pack-objects.o \
diff --git a/builtin-merge-recursive.c b/builtin-merge-recursive.c
new file mode 100644 (file)
index 0000000..45d4601
--- /dev/null
@@ -0,0 +1,1762 @@
+/*
+ * Recursive Merge algorithm stolen from git-merge-recursive.py by
+ * Fredrik Kuivinen.
+ * The thieves were Alex Riesen and Johannes Schindelin, in June/July 2006
+ */
+#include "cache.h"
+#include "cache-tree.h"
+#include "commit.h"
+#include "blob.h"
+#include "builtin.h"
+#include "tree-walk.h"
+#include "diff.h"
+#include "diffcore.h"
+#include "run-command.h"
+#include "tag.h"
+#include "unpack-trees.h"
+#include "path-list.h"
+#include "xdiff-interface.h"
+#include "interpolate.h"
+#include "attr.h"
+#include "merge-recursive.h"
+
+static int subtree_merge;
+
+static struct tree *shift_tree_object(struct tree *one, struct tree *two)
+{
+       unsigned char shifted[20];
+
+       /*
+        * NEEDSWORK: this limits the recursion depth to hardcoded
+        * value '2' to avoid excessive overhead.
+        */
+       shift_tree(one->object.sha1, two->object.sha1, shifted, 2);
+       if (!hashcmp(two->object.sha1, shifted))
+               return two;
+       return lookup_tree(shifted);
+}
+
+/*
+ * A virtual commit has
+ * - (const char *)commit->util set to the name, and
+ * - *(int *)commit->object.sha1 set to the virtual id.
+ */
+
+static unsigned commit_list_count(const struct commit_list *l)
+{
+       unsigned c = 0;
+       for (; l; l = l->next )
+               c++;
+       return c;
+}
+
+static struct commit *make_virtual_commit(struct tree *tree, const char *comment)
+{
+       struct commit *commit = xcalloc(1, sizeof(struct commit));
+       static unsigned virtual_id = 1;
+       commit->tree = tree;
+       commit->util = (void*)comment;
+       *(int*)commit->object.sha1 = virtual_id++;
+       /* avoid warnings */
+       commit->object.parsed = 1;
+       return commit;
+}
+
+/*
+ * Since we use get_tree_entry(), which does not put the read object into
+ * the object pool, we cannot rely on a == b.
+ */
+static int sha_eq(const unsigned char *a, const unsigned char *b)
+{
+       if (!a && !b)
+               return 2;
+       return a && b && hashcmp(a, b) == 0;
+}
+
+/*
+ * Since we want to write the index eventually, we cannot reuse the index
+ * for these (temporary) data.
+ */
+struct stage_data
+{
+       struct
+       {
+               unsigned mode;
+               unsigned char sha[20];
+       } stages[4];
+       unsigned processed:1;
+};
+
+static struct path_list current_file_set = {NULL, 0, 0, 1};
+static struct path_list current_directory_set = {NULL, 0, 0, 1};
+
+static int call_depth = 0;
+static int verbosity = 2;
+static int rename_limit = -1;
+static int buffer_output = 1;
+static struct strbuf obuf = STRBUF_INIT;
+
+static int show(int v)
+{
+       return (!call_depth && verbosity >= v) || verbosity >= 5;
+}
+
+static void flush_output(void)
+{
+       if (obuf.len) {
+               fputs(obuf.buf, stdout);
+               strbuf_reset(&obuf);
+       }
+}
+
+static void output(int v, const char *fmt, ...)
+{
+       int len;
+       va_list ap;
+
+       if (!show(v))
+               return;
+
+       strbuf_grow(&obuf, call_depth * 2 + 2);
+       memset(obuf.buf + obuf.len, ' ', call_depth * 2);
+       strbuf_setlen(&obuf, obuf.len + call_depth * 2);
+
+       va_start(ap, fmt);
+       len = vsnprintf(obuf.buf + obuf.len, strbuf_avail(&obuf), fmt, ap);
+       va_end(ap);
+
+       if (len < 0)
+               len = 0;
+       if (len >= strbuf_avail(&obuf)) {
+               strbuf_grow(&obuf, len + 2);
+               va_start(ap, fmt);
+               len = vsnprintf(obuf.buf + obuf.len, strbuf_avail(&obuf), fmt, ap);
+               va_end(ap);
+               if (len >= strbuf_avail(&obuf)) {
+                       die("this should not happen, your snprintf is broken");
+               }
+       }
+       strbuf_setlen(&obuf, obuf.len + len);
+       strbuf_add(&obuf, "\n", 1);
+       if (!buffer_output)
+               flush_output();
+}
+
+static void output_commit_title(struct commit *commit)
+{
+       int i;
+       flush_output();
+       for (i = call_depth; i--;)
+               fputs("  ", stdout);
+       if (commit->util)
+               printf("virtual %s\n", (char *)commit->util);
+       else {
+               printf("%s ", find_unique_abbrev(commit->object.sha1, DEFAULT_ABBREV));
+               if (parse_commit(commit) != 0)
+                       printf("(bad commit)\n");
+               else {
+                       const char *s;
+                       int len;
+                       for (s = commit->buffer; *s; s++)
+                               if (*s == '\n' && s[1] == '\n') {
+                                       s += 2;
+                                       break;
+                               }
+                       for (len = 0; s[len] && '\n' != s[len]; len++)
+                               ; /* do nothing */
+                       printf("%.*s\n", len, s);
+               }
+       }
+}
+
+static int add_cacheinfo(unsigned int mode, const unsigned char *sha1,
+               const char *path, int stage, int refresh, int options)
+{
+       struct cache_entry *ce;
+       ce = make_cache_entry(mode, sha1 ? sha1 : null_sha1, path, stage, refresh);
+       if (!ce)
+               return error("addinfo_cache failed for path '%s'", path);
+       return add_cache_entry(ce, options);
+}
+
+/*
+ * This is a global variable which is used in a number of places but
+ * only written to in the 'merge' function.
+ *
+ * index_only == 1    => Don't leave any non-stage 0 entries in the cache and
+ *                       don't update the working directory.
+ *               0    => Leave unmerged entries in the cache and update
+ *                       the working directory.
+ */
+static int index_only = 0;
+
+static void init_tree_desc_from_tree(struct tree_desc *desc, struct tree *tree)
+{
+       parse_tree(tree);
+       init_tree_desc(desc, tree->buffer, tree->size);
+}
+
+static int git_merge_trees(int index_only,
+                          struct tree *common,
+                          struct tree *head,
+                          struct tree *merge)
+{
+       int rc;
+       struct tree_desc t[3];
+       struct unpack_trees_options opts;
+
+       memset(&opts, 0, sizeof(opts));
+       if (index_only)
+               opts.index_only = 1;
+       else
+               opts.update = 1;
+       opts.merge = 1;
+       opts.head_idx = 2;
+       opts.fn = threeway_merge;
+
+       init_tree_desc_from_tree(t+0, common);
+       init_tree_desc_from_tree(t+1, head);
+       init_tree_desc_from_tree(t+2, merge);
+
+       rc = unpack_trees(3, t, &opts);
+       cache_tree_free(&active_cache_tree);
+       return rc;
+}
+
+static int unmerged_index(void)
+{
+       int i;
+       for (i = 0; i < active_nr; i++) {
+               struct cache_entry *ce = active_cache[i];
+               if (ce_stage(ce))
+                       return 1;
+       }
+       return 0;
+}
+
+struct tree *write_tree_from_memory(void)
+{
+       struct tree *result = NULL;
+
+       if (unmerged_index()) {
+               int i;
+               output(0, "There are unmerged index entries:");
+               for (i = 0; i < active_nr; i++) {
+                       struct cache_entry *ce = active_cache[i];
+                       if (ce_stage(ce))
+                               output(0, "%d %.*s", ce_stage(ce), ce_namelen(ce), ce->name);
+               }
+               return NULL;
+       }
+
+       if (!active_cache_tree)
+               active_cache_tree = cache_tree();
+
+       if (!cache_tree_fully_valid(active_cache_tree) &&
+           cache_tree_update(active_cache_tree,
+                             active_cache, active_nr, 0, 0) < 0)
+               die("error building trees");
+
+       result = lookup_tree(active_cache_tree->sha1);
+
+       return result;
+}
+
+static int save_files_dirs(const unsigned char *sha1,
+               const char *base, int baselen, const char *path,
+               unsigned int mode, int stage)
+{
+       int len = strlen(path);
+       char *newpath = xmalloc(baselen + len + 1);
+       memcpy(newpath, base, baselen);
+       memcpy(newpath + baselen, path, len);
+       newpath[baselen + len] = '\0';
+
+       if (S_ISDIR(mode))
+               path_list_insert(newpath, &current_directory_set);
+       else
+               path_list_insert(newpath, &current_file_set);
+       free(newpath);
+
+       return READ_TREE_RECURSIVE;
+}
+
+static int get_files_dirs(struct tree *tree)
+{
+       int n;
+       if (read_tree_recursive(tree, "", 0, 0, NULL, save_files_dirs) != 0)
+               return 0;
+       n = current_file_set.nr + current_directory_set.nr;
+       return n;
+}
+
+/*
+ * Returns an index_entry instance which doesn't have to correspond to
+ * a real cache entry in Git's index.
+ */
+static struct stage_data *insert_stage_data(const char *path,
+               struct tree *o, struct tree *a, struct tree *b,
+               struct path_list *entries)
+{
+       struct path_list_item *item;
+       struct stage_data *e = xcalloc(1, sizeof(struct stage_data));
+       get_tree_entry(o->object.sha1, path,
+                       e->stages[1].sha, &e->stages[1].mode);
+       get_tree_entry(a->object.sha1, path,
+                       e->stages[2].sha, &e->stages[2].mode);
+       get_tree_entry(b->object.sha1, path,
+                       e->stages[3].sha, &e->stages[3].mode);
+       item = path_list_insert(path, entries);
+       item->util = e;
+       return e;
+}
+
+/*
+ * Create a dictionary mapping file names to stage_data objects. The
+ * dictionary contains one entry for every path with a non-zero stage entry.
+ */
+static struct path_list *get_unmerged(void)
+{
+       struct path_list *unmerged = xcalloc(1, sizeof(struct path_list));
+       int i;
+
+       unmerged->strdup_paths = 1;
+
+       for (i = 0; i < active_nr; i++) {
+               struct path_list_item *item;
+               struct stage_data *e;
+               struct cache_entry *ce = active_cache[i];
+               if (!ce_stage(ce))
+                       continue;
+
+               item = path_list_lookup(ce->name, unmerged);
+               if (!item) {
+                       item = path_list_insert(ce->name, unmerged);
+                       item->util = xcalloc(1, sizeof(struct stage_data));
+               }
+               e = item->util;
+               e->stages[ce_stage(ce)].mode = ce->ce_mode;
+               hashcpy(e->stages[ce_stage(ce)].sha, ce->sha1);
+       }
+
+       return unmerged;
+}
+
+struct rename
+{
+       struct diff_filepair *pair;
+       struct stage_data *src_entry;
+       struct stage_data *dst_entry;
+       unsigned processed:1;
+};
+
+/*
+ * Get information of all renames which occurred between 'o_tree' and
+ * 'tree'. We need the three trees in the merge ('o_tree', 'a_tree' and
+ * 'b_tree') to be able to associate the correct cache entries with
+ * the rename information. 'tree' is always equal to either a_tree or b_tree.
+ */
+static struct path_list *get_renames(struct tree *tree,
+                                       struct tree *o_tree,
+                                       struct tree *a_tree,
+                                       struct tree *b_tree,
+                                       struct path_list *entries)
+{
+       int i;
+       struct path_list *renames;
+       struct diff_options opts;
+
+       renames = xcalloc(1, sizeof(struct path_list));
+       diff_setup(&opts);
+       DIFF_OPT_SET(&opts, RECURSIVE);
+       opts.detect_rename = DIFF_DETECT_RENAME;
+       opts.rename_limit = rename_limit;
+       opts.output_format = DIFF_FORMAT_NO_OUTPUT;
+       if (diff_setup_done(&opts) < 0)
+               die("diff setup failed");
+       diff_tree_sha1(o_tree->object.sha1, tree->object.sha1, "", &opts);
+       diffcore_std(&opts);
+       for (i = 0; i < diff_queued_diff.nr; ++i) {
+               struct path_list_item *item;
+               struct rename *re;
+               struct diff_filepair *pair = diff_queued_diff.queue[i];
+               if (pair->status != 'R') {
+                       diff_free_filepair(pair);
+                       continue;
+               }
+               re = xmalloc(sizeof(*re));
+               re->processed = 0;
+               re->pair = pair;
+               item = path_list_lookup(re->pair->one->path, entries);
+               if (!item)
+                       re->src_entry = insert_stage_data(re->pair->one->path,
+                                       o_tree, a_tree, b_tree, entries);
+               else
+                       re->src_entry = item->util;
+
+               item = path_list_lookup(re->pair->two->path, entries);
+               if (!item)
+                       re->dst_entry = insert_stage_data(re->pair->two->path,
+                                       o_tree, a_tree, b_tree, entries);
+               else
+                       re->dst_entry = item->util;
+               item = path_list_insert(pair->one->path, renames);
+               item->util = re;
+       }
+       opts.output_format = DIFF_FORMAT_NO_OUTPUT;
+       diff_queued_diff.nr = 0;
+       diff_flush(&opts);
+       return renames;
+}
+
+static int update_stages(const char *path, struct diff_filespec *o,
+                        struct diff_filespec *a, struct diff_filespec *b,
+                        int clear)
+{
+       int options = ADD_CACHE_OK_TO_ADD | ADD_CACHE_OK_TO_REPLACE;
+       if (clear)
+               if (remove_file_from_cache(path))
+                       return -1;
+       if (o)
+               if (add_cacheinfo(o->mode, o->sha1, path, 1, 0, options))
+                       return -1;
+       if (a)
+               if (add_cacheinfo(a->mode, a->sha1, path, 2, 0, options))
+                       return -1;
+       if (b)
+               if (add_cacheinfo(b->mode, b->sha1, path, 3, 0, options))
+                       return -1;
+       return 0;
+}
+
+static int remove_path(const char *name)
+{
+       int ret;
+       char *slash, *dirs;
+
+       ret = unlink(name);
+       if (ret)
+               return ret;
+       dirs = xstrdup(name);
+       while ((slash = strrchr(name, '/'))) {
+               *slash = '\0';
+               if (rmdir(name) != 0)
+                       break;
+       }
+       free(dirs);
+       return ret;
+}
+
+static int remove_file(int clean, const char *path, int no_wd)
+{
+       int update_cache = index_only || clean;
+       int update_working_directory = !index_only && !no_wd;
+
+       if (update_cache) {
+               if (remove_file_from_cache(path))
+                       return -1;
+       }
+       if (update_working_directory) {
+               unlink(path);
+               if (errno != ENOENT || errno != EISDIR)
+                       return -1;
+               remove_path(path);
+       }
+       return 0;
+}
+
+static char *unique_path(const char *path, const char *branch)
+{
+       char *newpath = xmalloc(strlen(path) + 1 + strlen(branch) + 8 + 1);
+       int suffix = 0;
+       struct stat st;
+       char *p = newpath + strlen(path);
+       strcpy(newpath, path);
+       *(p++) = '~';
+       strcpy(p, branch);
+       for (; *p; ++p)
+               if ('/' == *p)
+                       *p = '_';
+       while (path_list_has_path(&current_file_set, newpath) ||
+              path_list_has_path(&current_directory_set, newpath) ||
+              lstat(newpath, &st) == 0)
+               sprintf(p, "_%d", suffix++);
+
+       path_list_insert(newpath, &current_file_set);
+       return newpath;
+}
+
+static int mkdir_p(const char *path, unsigned long mode)
+{
+       /* path points to cache entries, so xstrdup before messing with it */
+       char *buf = xstrdup(path);
+       int result = safe_create_leading_directories(buf);
+       free(buf);
+       return result;
+}
+
+static void flush_buffer(int fd, const char *buf, unsigned long size)
+{
+       while (size > 0) {
+               long ret = write_in_full(fd, buf, size);
+               if (ret < 0) {
+                       /* Ignore epipe */
+                       if (errno == EPIPE)
+                               break;
+                       die("merge-recursive: %s", strerror(errno));
+               } else if (!ret) {
+                       die("merge-recursive: disk full?");
+               }
+               size -= ret;
+               buf += ret;
+       }
+}
+
+static int make_room_for_path(const char *path)
+{
+       int status;
+       const char *msg = "failed to create path '%s'%s";
+
+       status = mkdir_p(path, 0777);
+       if (status) {
+               if (status == -3) {
+                       /* something else exists */
+                       error(msg, path, ": perhaps a D/F conflict?");
+                       return -1;
+               }
+               die(msg, path, "");
+       }
+
+       /* Successful unlink is good.. */
+       if (!unlink(path))
+               return 0;
+       /* .. and so is no existing file */
+       if (errno == ENOENT)
+               return 0;
+       /* .. but not some other error (who really cares what?) */
+       return error(msg, path, ": perhaps a D/F conflict?");
+}
+
+static void update_file_flags(const unsigned char *sha,
+                             unsigned mode,
+                             const char *path,
+                             int update_cache,
+                             int update_wd)
+{
+       if (index_only)
+               update_wd = 0;
+
+       if (update_wd) {
+               enum object_type type;
+               void *buf;
+               unsigned long size;
+
+               if (S_ISGITLINK(mode))
+                       die("cannot read object %s '%s': It is a submodule!",
+                           sha1_to_hex(sha), path);
+
+               buf = read_sha1_file(sha, &type, &size);
+               if (!buf)
+                       die("cannot read object %s '%s'", sha1_to_hex(sha), path);
+               if (type != OBJ_BLOB)
+                       die("blob expected for %s '%s'", sha1_to_hex(sha), path);
+
+               if (make_room_for_path(path) < 0) {
+                       update_wd = 0;
+                       goto update_index;
+               }
+               if (S_ISREG(mode) || (!has_symlinks && S_ISLNK(mode))) {
+                       int fd;
+                       if (mode & 0100)
+                               mode = 0777;
+                       else
+                               mode = 0666;
+                       fd = open(path, O_WRONLY | O_TRUNC | O_CREAT, mode);
+                       if (fd < 0)
+                               die("failed to open %s: %s", path, strerror(errno));
+                       flush_buffer(fd, buf, size);
+                       close(fd);
+               } else if (S_ISLNK(mode)) {
+                       char *lnk = xmemdupz(buf, size);
+                       mkdir_p(path, 0777);
+                       unlink(path);
+                       symlink(lnk, path);
+                       free(lnk);
+               } else
+                       die("do not know what to do with %06o %s '%s'",
+                           mode, sha1_to_hex(sha), path);
+       }
+ update_index:
+       if (update_cache)
+               add_cacheinfo(mode, sha, path, 0, update_wd, ADD_CACHE_OK_TO_ADD);
+}
+
+static void update_file(int clean,
+                       const unsigned char *sha,
+                       unsigned mode,
+                       const char *path)
+{
+       update_file_flags(sha, mode, path, index_only || clean, !index_only);
+}
+
+/* Low level file merging, update and removal */
+
+struct merge_file_info
+{
+       unsigned char sha[20];
+       unsigned mode;
+       unsigned clean:1,
+                merge:1;
+};
+
+static void fill_mm(const unsigned char *sha1, mmfile_t *mm)
+{
+       unsigned long size;
+       enum object_type type;
+
+       if (!hashcmp(sha1, null_sha1)) {
+               mm->ptr = xstrdup("");
+               mm->size = 0;
+               return;
+       }
+
+       mm->ptr = read_sha1_file(sha1, &type, &size);
+       if (!mm->ptr || type != OBJ_BLOB)
+               die("unable to read blob object %s", sha1_to_hex(sha1));
+       mm->size = size;
+}
+
+/*
+ * Customizable low-level merge drivers support.
+ */
+
+struct ll_merge_driver;
+typedef int (*ll_merge_fn)(const struct ll_merge_driver *,
+                          const char *path,
+                          mmfile_t *orig,
+                          mmfile_t *src1, const char *name1,
+                          mmfile_t *src2, const char *name2,
+                          mmbuffer_t *result);
+
+struct ll_merge_driver {
+       const char *name;
+       const char *description;
+       ll_merge_fn fn;
+       const char *recursive;
+       struct ll_merge_driver *next;
+       char *cmdline;
+};
+
+/*
+ * Built-in low-levels
+ */
+static int ll_binary_merge(const struct ll_merge_driver *drv_unused,
+                          const char *path_unused,
+                          mmfile_t *orig,
+                          mmfile_t *src1, const char *name1,
+                          mmfile_t *src2, const char *name2,
+                          mmbuffer_t *result)
+{
+       /*
+        * The tentative merge result is "ours" for the final round,
+        * or common ancestor for an internal merge.  Still return
+        * "conflicted merge" status.
+        */
+       mmfile_t *stolen = index_only ? orig : src1;
+
+       result->ptr = stolen->ptr;
+       result->size = stolen->size;
+       stolen->ptr = NULL;
+       return 1;
+}
+
+static int ll_xdl_merge(const struct ll_merge_driver *drv_unused,
+                       const char *path_unused,
+                       mmfile_t *orig,
+                       mmfile_t *src1, const char *name1,
+                       mmfile_t *src2, const char *name2,
+                       mmbuffer_t *result)
+{
+       xpparam_t xpp;
+
+       if (buffer_is_binary(orig->ptr, orig->size) ||
+           buffer_is_binary(src1->ptr, src1->size) ||
+           buffer_is_binary(src2->ptr, src2->size)) {
+               warning("Cannot merge binary files: %s vs. %s\n",
+                       name1, name2);
+               return ll_binary_merge(drv_unused, path_unused,
+                                      orig, src1, name1,
+                                      src2, name2,
+                                      result);
+       }
+
+       memset(&xpp, 0, sizeof(xpp));
+       return xdl_merge(orig,
+                        src1, name1,
+                        src2, name2,
+                        &xpp, XDL_MERGE_ZEALOUS,
+                        result);
+}
+
+static int ll_union_merge(const struct ll_merge_driver *drv_unused,
+                         const char *path_unused,
+                         mmfile_t *orig,
+                         mmfile_t *src1, const char *name1,
+                         mmfile_t *src2, const char *name2,
+                         mmbuffer_t *result)
+{
+       char *src, *dst;
+       long size;
+       const int marker_size = 7;
+
+       int status = ll_xdl_merge(drv_unused, path_unused,
+                                 orig, src1, NULL, src2, NULL, result);
+       if (status <= 0)
+               return status;
+       size = result->size;
+       src = dst = result->ptr;
+       while (size) {
+               char ch;
+               if ((marker_size < size) &&
+                   (*src == '<' || *src == '=' || *src == '>')) {
+                       int i;
+                       ch = *src;
+                       for (i = 0; i < marker_size; i++)
+                               if (src[i] != ch)
+                                       goto not_a_marker;
+                       if (src[marker_size] != '\n')
+                               goto not_a_marker;
+                       src += marker_size + 1;
+                       size -= marker_size + 1;
+                       continue;
+               }
+       not_a_marker:
+               do {
+                       ch = *src++;
+                       *dst++ = ch;
+                       size--;
+               } while (ch != '\n' && size);
+       }
+       result->size = dst - result->ptr;
+       return 0;
+}
+
+#define LL_BINARY_MERGE 0
+#define LL_TEXT_MERGE 1
+#define LL_UNION_MERGE 2
+static struct ll_merge_driver ll_merge_drv[] = {
+       { "binary", "built-in binary merge", ll_binary_merge },
+       { "text", "built-in 3-way text merge", ll_xdl_merge },
+       { "union", "built-in union merge", ll_union_merge },
+};
+
+static void create_temp(mmfile_t *src, char *path)
+{
+       int fd;
+
+       strcpy(path, ".merge_file_XXXXXX");
+       fd = xmkstemp(path);
+       if (write_in_full(fd, src->ptr, src->size) != src->size)
+               die("unable to write temp-file");
+       close(fd);
+}
+
+/*
+ * User defined low-level merge driver support.
+ */
+static int ll_ext_merge(const struct ll_merge_driver *fn,
+                       const char *path,
+                       mmfile_t *orig,
+                       mmfile_t *src1, const char *name1,
+                       mmfile_t *src2, const char *name2,
+                       mmbuffer_t *result)
+{
+       char temp[3][50];
+       char cmdbuf[2048];
+       struct interp table[] = {
+               { "%O" },
+               { "%A" },
+               { "%B" },
+       };
+       struct child_process child;
+       const char *args[20];
+       int status, fd, i;
+       struct stat st;
+
+       if (fn->cmdline == NULL)
+               die("custom merge driver %s lacks command line.", fn->name);
+
+       result->ptr = NULL;
+       result->size = 0;
+       create_temp(orig, temp[0]);
+       create_temp(src1, temp[1]);
+       create_temp(src2, temp[2]);
+
+       interp_set_entry(table, 0, temp[0]);
+       interp_set_entry(table, 1, temp[1]);
+       interp_set_entry(table, 2, temp[2]);
+
+       output(1, "merging %s using %s", path,
+              fn->description ? fn->description : fn->name);
+
+       interpolate(cmdbuf, sizeof(cmdbuf), fn->cmdline, table, 3);
+
+       memset(&child, 0, sizeof(child));
+       child.argv = args;
+       args[0] = "sh";
+       args[1] = "-c";
+       args[2] = cmdbuf;
+       args[3] = NULL;
+
+       status = run_command(&child);
+       if (status < -ERR_RUN_COMMAND_FORK)
+               ; /* failure in run-command */
+       else
+               status = -status;
+       fd = open(temp[1], O_RDONLY);
+       if (fd < 0)
+               goto bad;
+       if (fstat(fd, &st))
+               goto close_bad;
+       result->size = st.st_size;
+       result->ptr = xmalloc(result->size + 1);
+       if (read_in_full(fd, result->ptr, result->size) != result->size) {
+               free(result->ptr);
+               result->ptr = NULL;
+               result->size = 0;
+       }
+ close_bad:
+       close(fd);
+ bad:
+       for (i = 0; i < 3; i++)
+               unlink(temp[i]);
+       return status;
+}
+
+/*
+ * merge.default and merge.driver configuration items
+ */
+static struct ll_merge_driver *ll_user_merge, **ll_user_merge_tail;
+static const char *default_ll_merge;
+
+static int read_merge_config(const char *var, const char *value)
+{
+       struct ll_merge_driver *fn;
+       const char *ep, *name;
+       int namelen;
+
+       if (!strcmp(var, "merge.default")) {
+               if (value)
+                       default_ll_merge = strdup(value);
+               return 0;
+       }
+
+       /*
+        * We are not interested in anything but "merge.<name>.variable";
+        * especially, we do not want to look at variables such as
+        * "merge.summary", "merge.tool", and "merge.verbosity".
+        */
+       if (prefixcmp(var, "merge.") || (ep = strrchr(var, '.')) == var + 5)
+               return 0;
+
+       /*
+        * Find existing one as we might be processing merge.<name>.var2
+        * after seeing merge.<name>.var1.
+        */
+       name = var + 6;
+       namelen = ep - name;
+       for (fn = ll_user_merge; fn; fn = fn->next)
+               if (!strncmp(fn->name, name, namelen) && !fn->name[namelen])
+                       break;
+       if (!fn) {
+               fn = xcalloc(1, sizeof(struct ll_merge_driver));
+               fn->name = xmemdupz(name, namelen);
+               fn->fn = ll_ext_merge;
+               *ll_user_merge_tail = fn;
+               ll_user_merge_tail = &(fn->next);
+       }
+
+       ep++;
+
+       if (!strcmp("name", ep)) {
+               if (!value)
+                       return error("%s: lacks value", var);
+               fn->description = strdup(value);
+               return 0;
+       }
+
+       if (!strcmp("driver", ep)) {
+               if (!value)
+                       return error("%s: lacks value", var);
+               /*
+                * merge.<name>.driver specifies the command line:
+                *
+                *      command-line
+                *
+                * The command-line will be interpolated with the following
+                * tokens and is given to the shell:
+                *
+                *    %O - temporary file name for the merge base.
+                *    %A - temporary file name for our version.
+                *    %B - temporary file name for the other branches' version.
+                *
+                * The external merge driver should write the results in the
+                * file named by %A, and signal that it has done with zero exit
+                * status.
+                */
+               fn->cmdline = strdup(value);
+               return 0;
+       }
+
+       if (!strcmp("recursive", ep)) {
+               if (!value)
+                       return error("%s: lacks value", var);
+               fn->recursive = strdup(value);
+               return 0;
+       }
+
+       return 0;
+}
+
+static void initialize_ll_merge(void)
+{
+       if (ll_user_merge_tail)
+               return;
+       ll_user_merge_tail = &ll_user_merge;
+       git_config(read_merge_config);
+}
+
+static const struct ll_merge_driver *find_ll_merge_driver(const char *merge_attr)
+{
+       struct ll_merge_driver *fn;
+       const char *name;
+       int i;
+
+       initialize_ll_merge();
+
+       if (ATTR_TRUE(merge_attr))
+               return &ll_merge_drv[LL_TEXT_MERGE];
+       else if (ATTR_FALSE(merge_attr))
+               return &ll_merge_drv[LL_BINARY_MERGE];
+       else if (ATTR_UNSET(merge_attr)) {
+               if (!default_ll_merge)
+                       return &ll_merge_drv[LL_TEXT_MERGE];
+               else
+                       name = default_ll_merge;
+       }
+       else
+               name = merge_attr;
+
+       for (fn = ll_user_merge; fn; fn = fn->next)
+               if (!strcmp(fn->name, name))
+                       return fn;
+
+       for (i = 0; i < ARRAY_SIZE(ll_merge_drv); i++)
+               if (!strcmp(ll_merge_drv[i].name, name))
+                       return &ll_merge_drv[i];
+
+       /* default to the 3-way */
+       return &ll_merge_drv[LL_TEXT_MERGE];
+}
+
+static const char *git_path_check_merge(const char *path)
+{
+       static struct git_attr_check attr_merge_check;
+
+       if (!attr_merge_check.attr)
+               attr_merge_check.attr = git_attr("merge", 5);
+
+       if (git_checkattr(path, 1, &attr_merge_check))
+               return NULL;
+       return attr_merge_check.value;
+}
+
+static int ll_merge(mmbuffer_t *result_buf,
+                   struct diff_filespec *o,
+                   struct diff_filespec *a,
+                   struct diff_filespec *b,
+                   const char *branch1,
+                   const char *branch2)
+{
+       mmfile_t orig, src1, src2;
+       char *name1, *name2;
+       int merge_status;
+       const char *ll_driver_name;
+       const struct ll_merge_driver *driver;
+
+       name1 = xstrdup(mkpath("%s:%s", branch1, a->path));
+       name2 = xstrdup(mkpath("%s:%s", branch2, b->path));
+
+       fill_mm(o->sha1, &orig);
+       fill_mm(a->sha1, &src1);
+       fill_mm(b->sha1, &src2);
+
+       ll_driver_name = git_path_check_merge(a->path);
+       driver = find_ll_merge_driver(ll_driver_name);
+
+       if (index_only && driver->recursive)
+               driver = find_ll_merge_driver(driver->recursive);
+       merge_status = driver->fn(driver, a->path,
+                                 &orig, &src1, name1, &src2, name2,
+                                 result_buf);
+
+       free(name1);
+       free(name2);
+       free(orig.ptr);
+       free(src1.ptr);
+       free(src2.ptr);
+       return merge_status;
+}
+
+static struct merge_file_info merge_file(struct diff_filespec *o,
+               struct diff_filespec *a, struct diff_filespec *b,
+               const char *branch1, const char *branch2)
+{
+       struct merge_file_info result;
+       result.merge = 0;
+       result.clean = 1;
+
+       if ((S_IFMT & a->mode) != (S_IFMT & b->mode)) {
+               result.clean = 0;
+               if (S_ISREG(a->mode)) {
+                       result.mode = a->mode;
+                       hashcpy(result.sha, a->sha1);
+               } else {
+                       result.mode = b->mode;
+                       hashcpy(result.sha, b->sha1);
+               }
+       } else {
+               if (!sha_eq(a->sha1, o->sha1) && !sha_eq(b->sha1, o->sha1))
+                       result.merge = 1;
+
+               result.mode = a->mode == o->mode ? b->mode: a->mode;
+
+               if (sha_eq(a->sha1, o->sha1))
+                       hashcpy(result.sha, b->sha1);
+               else if (sha_eq(b->sha1, o->sha1))
+                       hashcpy(result.sha, a->sha1);
+               else if (S_ISREG(a->mode)) {
+                       mmbuffer_t result_buf;
+                       int merge_status;
+
+                       merge_status = ll_merge(&result_buf, o, a, b,
+                                               branch1, branch2);
+
+                       if ((merge_status < 0) || !result_buf.ptr)
+                               die("Failed to execute internal merge");
+
+                       if (write_sha1_file(result_buf.ptr, result_buf.size,
+                                           blob_type, result.sha))
+                               die("Unable to add %s to database",
+                                   a->path);
+
+                       free(result_buf.ptr);
+                       result.clean = (merge_status == 0);
+               } else if (S_ISGITLINK(a->mode)) {
+                       result.clean = 0;
+                       hashcpy(result.sha, a->sha1);
+               } else if (S_ISLNK(a->mode)) {
+                       hashcpy(result.sha, a->sha1);
+
+                       if (!sha_eq(a->sha1, b->sha1))
+                               result.clean = 0;
+               } else {
+                       die("unsupported object type in the tree");
+               }
+       }
+
+       return result;
+}
+
+static void conflict_rename_rename(struct rename *ren1,
+                                  const char *branch1,
+                                  struct rename *ren2,
+                                  const char *branch2)
+{
+       char *del[2];
+       int delp = 0;
+       const char *ren1_dst = ren1->pair->two->path;
+       const char *ren2_dst = ren2->pair->two->path;
+       const char *dst_name1 = ren1_dst;
+       const char *dst_name2 = ren2_dst;
+       if (path_list_has_path(&current_directory_set, ren1_dst)) {
+               dst_name1 = del[delp++] = unique_path(ren1_dst, branch1);
+               output(1, "%s is a directory in %s added as %s instead",
+                      ren1_dst, branch2, dst_name1);
+               remove_file(0, ren1_dst, 0);
+       }
+       if (path_list_has_path(&current_directory_set, ren2_dst)) {
+               dst_name2 = del[delp++] = unique_path(ren2_dst, branch2);
+               output(1, "%s is a directory in %s added as %s instead",
+                      ren2_dst, branch1, dst_name2);
+               remove_file(0, ren2_dst, 0);
+       }
+       if (index_only) {
+               remove_file_from_cache(dst_name1);
+               remove_file_from_cache(dst_name2);
+               /*
+                * Uncomment to leave the conflicting names in the resulting tree
+                *
+                * update_file(0, ren1->pair->two->sha1, ren1->pair->two->mode, dst_name1);
+                * update_file(0, ren2->pair->two->sha1, ren2->pair->two->mode, dst_name2);
+                */
+       } else {
+               update_stages(dst_name1, NULL, ren1->pair->two, NULL, 1);
+               update_stages(dst_name2, NULL, NULL, ren2->pair->two, 1);
+       }
+       while (delp--)
+               free(del[delp]);
+}
+
+static void conflict_rename_dir(struct rename *ren1,
+                               const char *branch1)
+{
+       char *new_path = unique_path(ren1->pair->two->path, branch1);
+       output(1, "Renamed %s to %s instead", ren1->pair->one->path, new_path);
+       remove_file(0, ren1->pair->two->path, 0);
+       update_file(0, ren1->pair->two->sha1, ren1->pair->two->mode, new_path);
+       free(new_path);
+}
+
+static void conflict_rename_rename_2(struct rename *ren1,
+                                    const char *branch1,
+                                    struct rename *ren2,
+                                    const char *branch2)
+{
+       char *new_path1 = unique_path(ren1->pair->two->path, branch1);
+       char *new_path2 = unique_path(ren2->pair->two->path, branch2);
+       output(1, "Renamed %s to %s and %s to %s instead",
+              ren1->pair->one->path, new_path1,
+              ren2->pair->one->path, new_path2);
+       remove_file(0, ren1->pair->two->path, 0);
+       update_file(0, ren1->pair->two->sha1, ren1->pair->two->mode, new_path1);
+       update_file(0, ren2->pair->two->sha1, ren2->pair->two->mode, new_path2);
+       free(new_path2);
+       free(new_path1);
+}
+
+static int process_renames(struct path_list *a_renames,
+                          struct path_list *b_renames,
+                          const char *a_branch,
+                          const char *b_branch)
+{
+       int clean_merge = 1, i, j;
+       struct path_list a_by_dst = {NULL, 0, 0, 0}, b_by_dst = {NULL, 0, 0, 0};
+       const struct rename *sre;
+
+       for (i = 0; i < a_renames->nr; i++) {
+               sre = a_renames->items[i].util;
+               path_list_insert(sre->pair->two->path, &a_by_dst)->util
+                       = sre->dst_entry;
+       }
+       for (i = 0; i < b_renames->nr; i++) {
+               sre = b_renames->items[i].util;
+               path_list_insert(sre->pair->two->path, &b_by_dst)->util
+                       = sre->dst_entry;
+       }
+
+       for (i = 0, j = 0; i < a_renames->nr || j < b_renames->nr;) {
+               int compare;
+               char *src;
+               struct path_list *renames1, *renames2, *renames2Dst;
+               struct rename *ren1 = NULL, *ren2 = NULL;
+               const char *branch1, *branch2;
+               const char *ren1_src, *ren1_dst;
+
+               if (i >= a_renames->nr) {
+                       compare = 1;
+                       ren2 = b_renames->items[j++].util;
+               } else if (j >= b_renames->nr) {
+                       compare = -1;
+                       ren1 = a_renames->items[i++].util;
+               } else {
+                       compare = strcmp(a_renames->items[i].path,
+                                       b_renames->items[j].path);
+                       if (compare <= 0)
+                               ren1 = a_renames->items[i++].util;
+                       if (compare >= 0)
+                               ren2 = b_renames->items[j++].util;
+               }
+
+               /* TODO: refactor, so that 1/2 are not needed */
+               if (ren1) {
+                       renames1 = a_renames;
+                       renames2 = b_renames;
+                       renames2Dst = &b_by_dst;
+                       branch1 = a_branch;
+                       branch2 = b_branch;
+               } else {
+                       struct rename *tmp;
+                       renames1 = b_renames;
+                       renames2 = a_renames;
+                       renames2Dst = &a_by_dst;
+                       branch1 = b_branch;
+                       branch2 = a_branch;
+                       tmp = ren2;
+                       ren2 = ren1;
+                       ren1 = tmp;
+               }
+               src = ren1->pair->one->path;
+
+               ren1->dst_entry->processed = 1;
+               ren1->src_entry->processed = 1;
+
+               if (ren1->processed)
+                       continue;
+               ren1->processed = 1;
+
+               ren1_src = ren1->pair->one->path;
+               ren1_dst = ren1->pair->two->path;
+
+               if (ren2) {
+                       const char *ren2_src = ren2->pair->one->path;
+                       const char *ren2_dst = ren2->pair->two->path;
+                       /* Renamed in 1 and renamed in 2 */
+                       if (strcmp(ren1_src, ren2_src) != 0)
+                               die("ren1.src != ren2.src");
+                       ren2->dst_entry->processed = 1;
+                       ren2->processed = 1;
+                       if (strcmp(ren1_dst, ren2_dst) != 0) {
+                               clean_merge = 0;
+                               output(1, "CONFLICT (rename/rename): "
+                                      "Rename \"%s\"->\"%s\" in branch \"%s\" "
+                                      "rename \"%s\"->\"%s\" in \"%s\"%s",
+                                      src, ren1_dst, branch1,
+                                      src, ren2_dst, branch2,
+                                      index_only ? " (left unresolved)": "");
+                               if (index_only) {
+                                       remove_file_from_cache(src);
+                                       update_file(0, ren1->pair->one->sha1,
+                                                   ren1->pair->one->mode, src);
+                               }
+                               conflict_rename_rename(ren1, branch1, ren2, branch2);
+                       } else {
+                               struct merge_file_info mfi;
+                               remove_file(1, ren1_src, 1);
+                               mfi = merge_file(ren1->pair->one,
+                                                ren1->pair->two,
+                                                ren2->pair->two,
+                                                branch1,
+                                                branch2);
+                               if (mfi.merge || !mfi.clean)
+                                       output(1, "Renamed %s->%s", src, ren1_dst);
+
+                               if (mfi.merge)
+                                       output(2, "Auto-merged %s", ren1_dst);
+
+                               if (!mfi.clean) {
+                                       output(1, "CONFLICT (content): merge conflict in %s",
+                                              ren1_dst);
+                                       clean_merge = 0;
+
+                                       if (!index_only)
+                                               update_stages(ren1_dst,
+                                                             ren1->pair->one,
+                                                             ren1->pair->two,
+                                                             ren2->pair->two,
+                                                             1 /* clear */);
+                               }
+                               update_file(mfi.clean, mfi.sha, mfi.mode, ren1_dst);
+                       }
+               } else {
+                       /* Renamed in 1, maybe changed in 2 */
+                       struct path_list_item *item;
+                       /* we only use sha1 and mode of these */
+                       struct diff_filespec src_other, dst_other;
+                       int try_merge, stage = a_renames == renames1 ? 3: 2;
+
+                       remove_file(1, ren1_src, index_only || stage == 3);
+
+                       hashcpy(src_other.sha1, ren1->src_entry->stages[stage].sha);
+                       src_other.mode = ren1->src_entry->stages[stage].mode;
+                       hashcpy(dst_other.sha1, ren1->dst_entry->stages[stage].sha);
+                       dst_other.mode = ren1->dst_entry->stages[stage].mode;
+
+                       try_merge = 0;
+
+                       if (path_list_has_path(&current_directory_set, ren1_dst)) {
+                               clean_merge = 0;
+                               output(1, "CONFLICT (rename/directory): Renamed %s->%s in %s "
+                                      " directory %s added in %s",
+                                      ren1_src, ren1_dst, branch1,
+                                      ren1_dst, branch2);
+                               conflict_rename_dir(ren1, branch1);
+                       } else if (sha_eq(src_other.sha1, null_sha1)) {
+                               clean_merge = 0;
+                               output(1, "CONFLICT (rename/delete): Renamed %s->%s in %s "
+                                      "and deleted in %s",
+                                      ren1_src, ren1_dst, branch1,
+                                      branch2);
+                               update_file(0, ren1->pair->two->sha1, ren1->pair->two->mode, ren1_dst);
+                       } else if (!sha_eq(dst_other.sha1, null_sha1)) {
+                               const char *new_path;
+                               clean_merge = 0;
+                               try_merge = 1;
+                               output(1, "CONFLICT (rename/add): Renamed %s->%s in %s. "
+                                      "%s added in %s",
+                                      ren1_src, ren1_dst, branch1,
+                                      ren1_dst, branch2);
+                               new_path = unique_path(ren1_dst, branch2);
+                               output(1, "Added as %s instead", new_path);
+                               update_file(0, dst_other.sha1, dst_other.mode, new_path);
+                       } else if ((item = path_list_lookup(ren1_dst, renames2Dst))) {
+                               ren2 = item->util;
+                               clean_merge = 0;
+                               ren2->processed = 1;
+                               output(1, "CONFLICT (rename/rename): Renamed %s->%s in %s. "
+                                      "Renamed %s->%s in %s",
+                                      ren1_src, ren1_dst, branch1,
+                                      ren2->pair->one->path, ren2->pair->two->path, branch2);
+                               conflict_rename_rename_2(ren1, branch1, ren2, branch2);
+                       } else
+                               try_merge = 1;
+
+                       if (try_merge) {
+                               struct diff_filespec *o, *a, *b;
+                               struct merge_file_info mfi;
+                               src_other.path = (char *)ren1_src;
+
+                               o = ren1->pair->one;
+                               if (a_renames == renames1) {
+                                       a = ren1->pair->two;
+                                       b = &src_other;
+                               } else {
+                                       b = ren1->pair->two;
+                                       a = &src_other;
+                               }
+                               mfi = merge_file(o, a, b,
+                                               a_branch, b_branch);
+
+                               if (mfi.clean &&
+                                   sha_eq(mfi.sha, ren1->pair->two->sha1) &&
+                                   mfi.mode == ren1->pair->two->mode)
+                                       /*
+                                        * This messaged is part of
+                                        * t6022 test. If you change
+                                        * it update the test too.
+                                        */
+                                       output(3, "Skipped %s (merged same as existing)", ren1_dst);
+                               else {
+                                       if (mfi.merge || !mfi.clean)
+                                               output(1, "Renamed %s => %s", ren1_src, ren1_dst);
+                                       if (mfi.merge)
+                                               output(2, "Auto-merged %s", ren1_dst);
+                                       if (!mfi.clean) {
+                                               output(1, "CONFLICT (rename/modify): Merge conflict in %s",
+                                                      ren1_dst);
+                                               clean_merge = 0;
+
+                                               if (!index_only)
+                                                       update_stages(ren1_dst,
+                                                                     o, a, b, 1);
+                                       }
+                                       update_file(mfi.clean, mfi.sha, mfi.mode, ren1_dst);
+                               }
+                       }
+               }
+       }
+       path_list_clear(&a_by_dst, 0);
+       path_list_clear(&b_by_dst, 0);
+
+       return clean_merge;
+}
+
+static unsigned char *stage_sha(const unsigned char *sha, unsigned mode)
+{
+       return (is_null_sha1(sha) || mode == 0) ? NULL: (unsigned char *)sha;
+}
+
+/* Per entry merge function */
+static int process_entry(const char *path, struct stage_data *entry,
+                        const char *branch1,
+                        const char *branch2)
+{
+       /*
+       printf("processing entry, clean cache: %s\n", index_only ? "yes": "no");
+       print_index_entry("\tpath: ", entry);
+       */
+       int clean_merge = 1;
+       unsigned o_mode = entry->stages[1].mode;
+       unsigned a_mode = entry->stages[2].mode;
+       unsigned b_mode = entry->stages[3].mode;
+       unsigned char *o_sha = stage_sha(entry->stages[1].sha, o_mode);
+       unsigned char *a_sha = stage_sha(entry->stages[2].sha, a_mode);
+       unsigned char *b_sha = stage_sha(entry->stages[3].sha, b_mode);
+
+       if (o_sha && (!a_sha || !b_sha)) {
+               /* Case A: Deleted in one */
+               if ((!a_sha && !b_sha) ||
+                   (sha_eq(a_sha, o_sha) && !b_sha) ||
+                   (!a_sha && sha_eq(b_sha, o_sha))) {
+                       /* Deleted in both or deleted in one and
+                        * unchanged in the other */
+                       if (a_sha)
+                               output(2, "Removed %s", path);
+                       /* do not touch working file if it did not exist */
+                       remove_file(1, path, !a_sha);
+               } else {
+                       /* Deleted in one and changed in the other */
+                       clean_merge = 0;
+                       if (!a_sha) {
+                               output(1, "CONFLICT (delete/modify): %s deleted in %s "
+                                      "and modified in %s. Version %s of %s left in tree.",
+                                      path, branch1,
+                                      branch2, branch2, path);
+                               update_file(0, b_sha, b_mode, path);
+                       } else {
+                               output(1, "CONFLICT (delete/modify): %s deleted in %s "
+                                      "and modified in %s. Version %s of %s left in tree.",
+                                      path, branch2,
+                                      branch1, branch1, path);
+                               update_file(0, a_sha, a_mode, path);
+                       }
+               }
+
+       } else if ((!o_sha && a_sha && !b_sha) ||
+                  (!o_sha && !a_sha && b_sha)) {
+               /* Case B: Added in one. */
+               const char *add_branch;
+               const char *other_branch;
+               unsigned mode;
+               const unsigned char *sha;
+               const char *conf;
+
+               if (a_sha) {
+                       add_branch = branch1;
+                       other_branch = branch2;
+                       mode = a_mode;
+                       sha = a_sha;
+                       conf = "file/directory";
+               } else {
+                       add_branch = branch2;
+                       other_branch = branch1;
+                       mode = b_mode;
+                       sha = b_sha;
+                       conf = "directory/file";
+               }
+               if (path_list_has_path(&current_directory_set, path)) {
+                       const char *new_path = unique_path(path, add_branch);
+                       clean_merge = 0;
+                       output(1, "CONFLICT (%s): There is a directory with name %s in %s. "
+                              "Added %s as %s",
+                              conf, path, other_branch, path, new_path);
+                       remove_file(0, path, 0);
+                       update_file(0, sha, mode, new_path);
+               } else {
+                       output(2, "Added %s", path);
+                       update_file(1, sha, mode, path);
+               }
+       } else if (a_sha && b_sha) {
+               /* Case C: Added in both (check for same permissions) and */
+               /* case D: Modified in both, but differently. */
+               const char *reason = "content";
+               struct merge_file_info mfi;
+               struct diff_filespec o, a, b;
+
+               if (!o_sha) {
+                       reason = "add/add";
+                       o_sha = (unsigned char *)null_sha1;
+               }
+               output(2, "Auto-merged %s", path);
+               o.path = a.path = b.path = (char *)path;
+               hashcpy(o.sha1, o_sha);
+               o.mode = o_mode;
+               hashcpy(a.sha1, a_sha);
+               a.mode = a_mode;
+               hashcpy(b.sha1, b_sha);
+               b.mode = b_mode;
+
+               mfi = merge_file(&o, &a, &b,
+                                branch1, branch2);
+
+               clean_merge = mfi.clean;
+               if (mfi.clean)
+                       update_file(1, mfi.sha, mfi.mode, path);
+               else if (S_ISGITLINK(mfi.mode))
+                       output(1, "CONFLICT (submodule): Merge conflict in %s "
+                              "- needs %s", path, sha1_to_hex(b.sha1));
+               else {
+                       output(1, "CONFLICT (%s): Merge conflict in %s",
+                                       reason, path);
+
+                       if (index_only)
+                               update_file(0, mfi.sha, mfi.mode, path);
+                       else
+                               update_file_flags(mfi.sha, mfi.mode, path,
+                                             0 /* update_cache */, 1 /* update_working_directory */);
+               }
+       } else if (!o_sha && !a_sha && !b_sha) {
+               /*
+                * this entry was deleted altogether. a_mode == 0 means
+                * we had that path and want to actively remove it.
+                */
+               remove_file(1, path, !a_mode);
+       } else
+               die("Fatal merge failure, shouldn't happen.");
+
+       return clean_merge;
+}
+
+int merge_trees(struct tree *head,
+               struct tree *merge,
+               struct tree *common,
+               const char *branch1,
+               const char *branch2,
+               struct tree **result)
+{
+       int code, clean;
+
+       if (subtree_merge) {
+               merge = shift_tree_object(head, merge);
+               common = shift_tree_object(head, common);
+       }
+
+       if (sha_eq(common->object.sha1, merge->object.sha1)) {
+               output(0, "Already uptodate!");
+               *result = head;
+               return 1;
+       }
+
+       code = git_merge_trees(index_only, common, head, merge);
+
+       if (code != 0)
+               die("merging of trees %s and %s failed",
+                   sha1_to_hex(head->object.sha1),
+                   sha1_to_hex(merge->object.sha1));
+
+       if (unmerged_index()) {
+               struct path_list *entries, *re_head, *re_merge;
+               int i;
+               path_list_clear(&current_file_set, 1);
+               path_list_clear(&current_directory_set, 1);
+               get_files_dirs(head);
+               get_files_dirs(merge);
+
+               entries = get_unmerged();
+               re_head  = get_renames(head, common, head, merge, entries);
+               re_merge = get_renames(merge, common, head, merge, entries);
+               clean = process_renames(re_head, re_merge,
+                               branch1, branch2);
+               for (i = 0; i < entries->nr; i++) {
+                       const char *path = entries->items[i].path;
+                       struct stage_data *e = entries->items[i].util;
+                       if (!e->processed
+                               && !process_entry(path, e, branch1, branch2))
+                               clean = 0;
+               }
+
+               path_list_clear(re_merge, 0);
+               path_list_clear(re_head, 0);
+               path_list_clear(entries, 1);
+
+       }
+       else
+               clean = 1;
+
+       if (index_only)
+               *result = write_tree_from_memory();
+
+       return clean;
+}
+
+static struct commit_list *reverse_commit_list(struct commit_list *list)
+{
+       struct commit_list *next = NULL, *current, *backup;
+       for (current = list; current; current = backup) {
+               backup = current->next;
+               current->next = next;
+               next = current;
+       }
+       return next;
+}
+
+/*
+ * Merge the commits h1 and h2, return the resulting virtual
+ * commit object and a flag indicating the cleanness of the merge.
+ */
+int merge_recursive(struct commit *h1,
+                   struct commit *h2,
+                   const char *branch1,
+                   const char *branch2,
+                   struct commit_list *ca,
+                   struct commit **result)
+{
+       struct commit_list *iter;
+       struct commit *merged_common_ancestors;
+       struct tree *mrtree = mrtree;
+       int clean;
+
+       if (show(4)) {
+               output(4, "Merging:");
+               output_commit_title(h1);
+               output_commit_title(h2);
+       }
+
+       if (!ca) {
+               ca = get_merge_bases(h1, h2, 1);
+               ca = reverse_commit_list(ca);
+       }
+
+       if (show(5)) {
+               output(5, "found %u common ancestor(s):", commit_list_count(ca));
+               for (iter = ca; iter; iter = iter->next)
+                       output_commit_title(iter->item);
+       }
+
+       merged_common_ancestors = pop_commit(&ca);
+       if (merged_common_ancestors == NULL) {
+               /* if there is no common ancestor, make an empty tree */
+               struct tree *tree = xcalloc(1, sizeof(struct tree));
+
+               tree->object.parsed = 1;
+               tree->object.type = OBJ_TREE;
+               pretend_sha1_file(NULL, 0, OBJ_TREE, tree->object.sha1);
+               merged_common_ancestors = make_virtual_commit(tree, "ancestor");
+       }
+
+       for (iter = ca; iter; iter = iter->next) {
+               call_depth++;
+               /*
+                * When the merge fails, the result contains files
+                * with conflict markers. The cleanness flag is
+                * ignored, it was never actually used, as result of
+                * merge_trees has always overwritten it: the committed
+                * "conflicts" were already resolved.
+                */
+               discard_cache();
+               merge_recursive(merged_common_ancestors, iter->item,
+                               "Temporary merge branch 1",
+                               "Temporary merge branch 2",
+                               NULL,
+                               &merged_common_ancestors);
+               call_depth--;
+
+               if (!merged_common_ancestors)
+                       die("merge returned no commit");
+       }
+
+       discard_cache();
+       if (!call_depth) {
+               read_cache();
+               index_only = 0;
+       } else
+               index_only = 1;
+
+       clean = merge_trees(h1->tree, h2->tree, merged_common_ancestors->tree,
+                           branch1, branch2, &mrtree);
+
+       if (index_only) {
+               *result = make_virtual_commit(mrtree, "merged tree");
+               commit_list_insert(h1, &(*result)->parents);
+               commit_list_insert(h2, &(*result)->parents->next);
+       }
+       flush_output();
+       return clean;
+}
+
+static const char *better_branch_name(const char *branch)
+{
+       static char githead_env[8 + 40 + 1];
+       char *name;
+
+       if (strlen(branch) != 40)
+               return branch;
+       sprintf(githead_env, "GITHEAD_%s", branch);
+       name = getenv(githead_env);
+       return name ? name : branch;
+}
+
+static struct commit *get_ref(const char *ref)
+{
+       unsigned char sha1[20];
+       struct object *object;
+
+       if (get_sha1(ref, sha1))
+               die("Could not resolve ref '%s'", ref);
+       object = deref_tag(parse_object(sha1), ref, strlen(ref));
+       if (object->type == OBJ_TREE)
+               return make_virtual_commit((struct tree*)object,
+                       better_branch_name(ref));
+       if (object->type != OBJ_COMMIT)
+               return NULL;
+       if (parse_commit((struct commit *)object))
+               die("Could not parse commit '%s'", sha1_to_hex(object->sha1));
+       return (struct commit *)object;
+}
+
+static int merge_config(const char *var, const char *value)
+{
+       if (!strcasecmp(var, "merge.verbosity")) {
+               verbosity = git_config_int(var, value);
+               return 0;
+       }
+       if (!strcasecmp(var, "diff.renamelimit")) {
+               rename_limit = git_config_int(var, value);
+               return 0;
+       }
+       return git_default_config(var, value);
+}
+
+int cmd_merge_recursive(int argc, const char **argv, const char *prefix)
+{
+       static const char *bases[20];
+       static unsigned bases_count = 0;
+       int i, clean;
+       const char *branch1, *branch2;
+       struct commit *result, *h1, *h2;
+       struct commit_list *ca = NULL;
+       struct lock_file *lock = xcalloc(1, sizeof(struct lock_file));
+       int index_fd;
+
+       if (argv[0]) {
+               int namelen = strlen(argv[0]);
+               if (8 < namelen &&
+                   !strcmp(argv[0] + namelen - 8, "-subtree"))
+                       subtree_merge = 1;
+       }
+
+       git_config(merge_config);
+       if (getenv("GIT_MERGE_VERBOSITY"))
+               verbosity = strtol(getenv("GIT_MERGE_VERBOSITY"), NULL, 10);
+
+       if (argc < 4)
+               die("Usage: %s <base>... -- <head> <remote> ...\n", argv[0]);
+
+       for (i = 1; i < argc; ++i) {
+               if (!strcmp(argv[i], "--"))
+                       break;
+               if (bases_count < sizeof(bases)/sizeof(*bases))
+                       bases[bases_count++] = argv[i];
+       }
+       if (argc - i != 3) /* "--" "<head>" "<remote>" */
+               die("Not handling anything other than two heads merge.");
+       if (verbosity >= 5)
+               buffer_output = 0;
+
+       branch1 = argv[++i];
+       branch2 = argv[++i];
+
+       h1 = get_ref(branch1);
+       h2 = get_ref(branch2);
+
+       branch1 = better_branch_name(branch1);
+       branch2 = better_branch_name(branch2);
+
+       if (show(3))
+               printf("Merging %s with %s\n", branch1, branch2);
+
+       index_fd = hold_locked_index(lock, 1);
+
+       for (i = 0; i < bases_count; i++) {
+               struct commit *ancestor = get_ref(bases[i]);
+               ca = commit_list_insert(ancestor, &ca);
+       }
+       clean = merge_recursive(h1, h2, branch1, branch2, ca, &result);
+
+       if (active_cache_changed &&
+           (write_cache(index_fd, active_cache, active_nr) ||
+            commit_locked_index(lock)))
+                       die ("unable to write %s", get_index_file());
+
+       return clean ? 0: 1;
+}
index cb675c4d7a80b99a06d6d04156e767ea025eb512..428160d0e48b2acf1ac54d6d73910fd93151ca22 100644 (file)
--- a/builtin.h
+++ b/builtin.h
@@ -57,6 +57,7 @@ extern int cmd_mailsplit(int argc, const char **argv, const char *prefix);
 extern int cmd_merge_base(int argc, const char **argv, const char *prefix);
 extern int cmd_merge_ours(int argc, const char **argv, const char *prefix);
 extern int cmd_merge_file(int argc, const char **argv, const char *prefix);
+extern int cmd_merge_recursive(int argc, const char **argv, const char *prefix);
 extern int cmd_mv(int argc, const char **argv, const char *prefix);
 extern int cmd_name_rev(int argc, const char **argv, const char *prefix);
 extern int cmd_pack_objects(int argc, const char **argv, const char *prefix);
diff --git a/git.c b/git.c
index 15fec8974ad5361a2308b3c46424e896bc8e3246..114ea75eef55e2960ff111014a505c3eb678caae 100644 (file)
--- a/git.c
+++ b/git.c
@@ -330,6 +330,7 @@ static void handle_internal_command(int argc, const char **argv)
                { "merge-base", cmd_merge_base, RUN_SETUP },
                { "merge-file", cmd_merge_file },
                { "merge-ours", cmd_merge_ours, RUN_SETUP },
+               { "merge-recursive", cmd_merge_recursive, RUN_SETUP | NEED_WORK_TREE },
                { "mv", cmd_mv, RUN_SETUP | NEED_WORK_TREE },
                { "name-rev", cmd_name_rev, RUN_SETUP },
                { "pack-objects", cmd_pack_objects, RUN_SETUP },
diff --git a/merge-recursive.c b/merge-recursive.c
deleted file mode 100644 (file)
index bdf03b1..0000000
+++ /dev/null
@@ -1,1760 +0,0 @@
-/*
- * Recursive Merge algorithm stolen from git-merge-recursive.py by
- * Fredrik Kuivinen.
- * The thieves were Alex Riesen and Johannes Schindelin, in June/July 2006
- */
-#include "cache.h"
-#include "cache-tree.h"
-#include "commit.h"
-#include "blob.h"
-#include "tree-walk.h"
-#include "diff.h"
-#include "diffcore.h"
-#include "run-command.h"
-#include "tag.h"
-#include "unpack-trees.h"
-#include "path-list.h"
-#include "xdiff-interface.h"
-#include "interpolate.h"
-#include "attr.h"
-
-static int subtree_merge;
-
-static struct tree *shift_tree_object(struct tree *one, struct tree *two)
-{
-       unsigned char shifted[20];
-
-       /*
-        * NEEDSWORK: this limits the recursion depth to hardcoded
-        * value '2' to avoid excessive overhead.
-        */
-       shift_tree(one->object.sha1, two->object.sha1, shifted, 2);
-       if (!hashcmp(two->object.sha1, shifted))
-               return two;
-       return lookup_tree(shifted);
-}
-
-/*
- * A virtual commit has
- * - (const char *)commit->util set to the name, and
- * - *(int *)commit->object.sha1 set to the virtual id.
- */
-
-static unsigned commit_list_count(const struct commit_list *l)
-{
-       unsigned c = 0;
-       for (; l; l = l->next )
-               c++;
-       return c;
-}
-
-static struct commit *make_virtual_commit(struct tree *tree, const char *comment)
-{
-       struct commit *commit = xcalloc(1, sizeof(struct commit));
-       static unsigned virtual_id = 1;
-       commit->tree = tree;
-       commit->util = (void*)comment;
-       *(int*)commit->object.sha1 = virtual_id++;
-       /* avoid warnings */
-       commit->object.parsed = 1;
-       return commit;
-}
-
-/*
- * Since we use get_tree_entry(), which does not put the read object into
- * the object pool, we cannot rely on a == b.
- */
-static int sha_eq(const unsigned char *a, const unsigned char *b)
-{
-       if (!a && !b)
-               return 2;
-       return a && b && hashcmp(a, b) == 0;
-}
-
-/*
- * Since we want to write the index eventually, we cannot reuse the index
- * for these (temporary) data.
- */
-struct stage_data
-{
-       struct
-       {
-               unsigned mode;
-               unsigned char sha[20];
-       } stages[4];
-       unsigned processed:1;
-};
-
-static struct path_list current_file_set = {NULL, 0, 0, 1};
-static struct path_list current_directory_set = {NULL, 0, 0, 1};
-
-static int call_depth = 0;
-static int verbosity = 2;
-static int rename_limit = -1;
-static int buffer_output = 1;
-static struct strbuf obuf = STRBUF_INIT;
-
-static int show(int v)
-{
-       return (!call_depth && verbosity >= v) || verbosity >= 5;
-}
-
-static void flush_output(void)
-{
-       if (obuf.len) {
-               fputs(obuf.buf, stdout);
-               strbuf_reset(&obuf);
-       }
-}
-
-static void output(int v, const char *fmt, ...)
-{
-       int len;
-       va_list ap;
-
-       if (!show(v))
-               return;
-
-       strbuf_grow(&obuf, call_depth * 2 + 2);
-       memset(obuf.buf + obuf.len, ' ', call_depth * 2);
-       strbuf_setlen(&obuf, obuf.len + call_depth * 2);
-
-       va_start(ap, fmt);
-       len = vsnprintf(obuf.buf + obuf.len, strbuf_avail(&obuf), fmt, ap);
-       va_end(ap);
-
-       if (len < 0)
-               len = 0;
-       if (len >= strbuf_avail(&obuf)) {
-               strbuf_grow(&obuf, len + 2);
-               va_start(ap, fmt);
-               len = vsnprintf(obuf.buf + obuf.len, strbuf_avail(&obuf), fmt, ap);
-               va_end(ap);
-               if (len >= strbuf_avail(&obuf)) {
-                       die("this should not happen, your snprintf is broken");
-               }
-       }
-       strbuf_setlen(&obuf, obuf.len + len);
-       strbuf_add(&obuf, "\n", 1);
-       if (!buffer_output)
-               flush_output();
-}
-
-static void output_commit_title(struct commit *commit)
-{
-       int i;
-       flush_output();
-       for (i = call_depth; i--;)
-               fputs("  ", stdout);
-       if (commit->util)
-               printf("virtual %s\n", (char *)commit->util);
-       else {
-               printf("%s ", find_unique_abbrev(commit->object.sha1, DEFAULT_ABBREV));
-               if (parse_commit(commit) != 0)
-                       printf("(bad commit)\n");
-               else {
-                       const char *s;
-                       int len;
-                       for (s = commit->buffer; *s; s++)
-                               if (*s == '\n' && s[1] == '\n') {
-                                       s += 2;
-                                       break;
-                               }
-                       for (len = 0; s[len] && '\n' != s[len]; len++)
-                               ; /* do nothing */
-                       printf("%.*s\n", len, s);
-               }
-       }
-}
-
-static int add_cacheinfo(unsigned int mode, const unsigned char *sha1,
-               const char *path, int stage, int refresh, int options)
-{
-       struct cache_entry *ce;
-       ce = make_cache_entry(mode, sha1 ? sha1 : null_sha1, path, stage, refresh);
-       if (!ce)
-               return error("addinfo_cache failed for path '%s'", path);
-       return add_cache_entry(ce, options);
-}
-
-/*
- * This is a global variable which is used in a number of places but
- * only written to in the 'merge' function.
- *
- * index_only == 1    => Don't leave any non-stage 0 entries in the cache and
- *                       don't update the working directory.
- *               0    => Leave unmerged entries in the cache and update
- *                       the working directory.
- */
-static int index_only = 0;
-
-static void init_tree_desc_from_tree(struct tree_desc *desc, struct tree *tree)
-{
-       parse_tree(tree);
-       init_tree_desc(desc, tree->buffer, tree->size);
-}
-
-static int git_merge_trees(int index_only,
-                          struct tree *common,
-                          struct tree *head,
-                          struct tree *merge)
-{
-       int rc;
-       struct tree_desc t[3];
-       struct unpack_trees_options opts;
-
-       memset(&opts, 0, sizeof(opts));
-       if (index_only)
-               opts.index_only = 1;
-       else
-               opts.update = 1;
-       opts.merge = 1;
-       opts.head_idx = 2;
-       opts.fn = threeway_merge;
-
-       init_tree_desc_from_tree(t+0, common);
-       init_tree_desc_from_tree(t+1, head);
-       init_tree_desc_from_tree(t+2, merge);
-
-       rc = unpack_trees(3, t, &opts);
-       cache_tree_free(&active_cache_tree);
-       return rc;
-}
-
-static int unmerged_index(void)
-{
-       int i;
-       for (i = 0; i < active_nr; i++) {
-               struct cache_entry *ce = active_cache[i];
-               if (ce_stage(ce))
-                       return 1;
-       }
-       return 0;
-}
-
-static struct tree *git_write_tree(void)
-{
-       struct tree *result = NULL;
-
-       if (unmerged_index()) {
-               int i;
-               output(0, "There are unmerged index entries:");
-               for (i = 0; i < active_nr; i++) {
-                       struct cache_entry *ce = active_cache[i];
-                       if (ce_stage(ce))
-                               output(0, "%d %.*s", ce_stage(ce), ce_namelen(ce), ce->name);
-               }
-               return NULL;
-       }
-
-       if (!active_cache_tree)
-               active_cache_tree = cache_tree();
-
-       if (!cache_tree_fully_valid(active_cache_tree) &&
-           cache_tree_update(active_cache_tree,
-                             active_cache, active_nr, 0, 0) < 0)
-               die("error building trees");
-
-       result = lookup_tree(active_cache_tree->sha1);
-
-       return result;
-}
-
-static int save_files_dirs(const unsigned char *sha1,
-               const char *base, int baselen, const char *path,
-               unsigned int mode, int stage)
-{
-       int len = strlen(path);
-       char *newpath = xmalloc(baselen + len + 1);
-       memcpy(newpath, base, baselen);
-       memcpy(newpath + baselen, path, len);
-       newpath[baselen + len] = '\0';
-
-       if (S_ISDIR(mode))
-               path_list_insert(newpath, &current_directory_set);
-       else
-               path_list_insert(newpath, &current_file_set);
-       free(newpath);
-
-       return READ_TREE_RECURSIVE;
-}
-
-static int get_files_dirs(struct tree *tree)
-{
-       int n;
-       if (read_tree_recursive(tree, "", 0, 0, NULL, save_files_dirs) != 0)
-               return 0;
-       n = current_file_set.nr + current_directory_set.nr;
-       return n;
-}
-
-/*
- * Returns an index_entry instance which doesn't have to correspond to
- * a real cache entry in Git's index.
- */
-static struct stage_data *insert_stage_data(const char *path,
-               struct tree *o, struct tree *a, struct tree *b,
-               struct path_list *entries)
-{
-       struct path_list_item *item;
-       struct stage_data *e = xcalloc(1, sizeof(struct stage_data));
-       get_tree_entry(o->object.sha1, path,
-                       e->stages[1].sha, &e->stages[1].mode);
-       get_tree_entry(a->object.sha1, path,
-                       e->stages[2].sha, &e->stages[2].mode);
-       get_tree_entry(b->object.sha1, path,
-                       e->stages[3].sha, &e->stages[3].mode);
-       item = path_list_insert(path, entries);
-       item->util = e;
-       return e;
-}
-
-/*
- * Create a dictionary mapping file names to stage_data objects. The
- * dictionary contains one entry for every path with a non-zero stage entry.
- */
-static struct path_list *get_unmerged(void)
-{
-       struct path_list *unmerged = xcalloc(1, sizeof(struct path_list));
-       int i;
-
-       unmerged->strdup_paths = 1;
-
-       for (i = 0; i < active_nr; i++) {
-               struct path_list_item *item;
-               struct stage_data *e;
-               struct cache_entry *ce = active_cache[i];
-               if (!ce_stage(ce))
-                       continue;
-
-               item = path_list_lookup(ce->name, unmerged);
-               if (!item) {
-                       item = path_list_insert(ce->name, unmerged);
-                       item->util = xcalloc(1, sizeof(struct stage_data));
-               }
-               e = item->util;
-               e->stages[ce_stage(ce)].mode = ce->ce_mode;
-               hashcpy(e->stages[ce_stage(ce)].sha, ce->sha1);
-       }
-
-       return unmerged;
-}
-
-struct rename
-{
-       struct diff_filepair *pair;
-       struct stage_data *src_entry;
-       struct stage_data *dst_entry;
-       unsigned processed:1;
-};
-
-/*
- * Get information of all renames which occurred between 'o_tree' and
- * 'tree'. We need the three trees in the merge ('o_tree', 'a_tree' and
- * 'b_tree') to be able to associate the correct cache entries with
- * the rename information. 'tree' is always equal to either a_tree or b_tree.
- */
-static struct path_list *get_renames(struct tree *tree,
-                                       struct tree *o_tree,
-                                       struct tree *a_tree,
-                                       struct tree *b_tree,
-                                       struct path_list *entries)
-{
-       int i;
-       struct path_list *renames;
-       struct diff_options opts;
-
-       renames = xcalloc(1, sizeof(struct path_list));
-       diff_setup(&opts);
-       DIFF_OPT_SET(&opts, RECURSIVE);
-       opts.detect_rename = DIFF_DETECT_RENAME;
-       opts.rename_limit = rename_limit;
-       opts.output_format = DIFF_FORMAT_NO_OUTPUT;
-       if (diff_setup_done(&opts) < 0)
-               die("diff setup failed");
-       diff_tree_sha1(o_tree->object.sha1, tree->object.sha1, "", &opts);
-       diffcore_std(&opts);
-       for (i = 0; i < diff_queued_diff.nr; ++i) {
-               struct path_list_item *item;
-               struct rename *re;
-               struct diff_filepair *pair = diff_queued_diff.queue[i];
-               if (pair->status != 'R') {
-                       diff_free_filepair(pair);
-                       continue;
-               }
-               re = xmalloc(sizeof(*re));
-               re->processed = 0;
-               re->pair = pair;
-               item = path_list_lookup(re->pair->one->path, entries);
-               if (!item)
-                       re->src_entry = insert_stage_data(re->pair->one->path,
-                                       o_tree, a_tree, b_tree, entries);
-               else
-                       re->src_entry = item->util;
-
-               item = path_list_lookup(re->pair->two->path, entries);
-               if (!item)
-                       re->dst_entry = insert_stage_data(re->pair->two->path,
-                                       o_tree, a_tree, b_tree, entries);
-               else
-                       re->dst_entry = item->util;
-               item = path_list_insert(pair->one->path, renames);
-               item->util = re;
-       }
-       opts.output_format = DIFF_FORMAT_NO_OUTPUT;
-       diff_queued_diff.nr = 0;
-       diff_flush(&opts);
-       return renames;
-}
-
-static int update_stages(const char *path, struct diff_filespec *o,
-                        struct diff_filespec *a, struct diff_filespec *b,
-                        int clear)
-{
-       int options = ADD_CACHE_OK_TO_ADD | ADD_CACHE_OK_TO_REPLACE;
-       if (clear)
-               if (remove_file_from_cache(path))
-                       return -1;
-       if (o)
-               if (add_cacheinfo(o->mode, o->sha1, path, 1, 0, options))
-                       return -1;
-       if (a)
-               if (add_cacheinfo(a->mode, a->sha1, path, 2, 0, options))
-                       return -1;
-       if (b)
-               if (add_cacheinfo(b->mode, b->sha1, path, 3, 0, options))
-                       return -1;
-       return 0;
-}
-
-static int remove_path(const char *name)
-{
-       int ret;
-       char *slash, *dirs;
-
-       ret = unlink(name);
-       if (ret)
-               return ret;
-       dirs = xstrdup(name);
-       while ((slash = strrchr(name, '/'))) {
-               *slash = '\0';
-               if (rmdir(name) != 0)
-                       break;
-       }
-       free(dirs);
-       return ret;
-}
-
-static int remove_file(int clean, const char *path, int no_wd)
-{
-       int update_cache = index_only || clean;
-       int update_working_directory = !index_only && !no_wd;
-
-       if (update_cache) {
-               if (remove_file_from_cache(path))
-                       return -1;
-       }
-       if (update_working_directory) {
-               unlink(path);
-               if (errno != ENOENT || errno != EISDIR)
-                       return -1;
-               remove_path(path);
-       }
-       return 0;
-}
-
-static char *unique_path(const char *path, const char *branch)
-{
-       char *newpath = xmalloc(strlen(path) + 1 + strlen(branch) + 8 + 1);
-       int suffix = 0;
-       struct stat st;
-       char *p = newpath + strlen(path);
-       strcpy(newpath, path);
-       *(p++) = '~';
-       strcpy(p, branch);
-       for (; *p; ++p)
-               if ('/' == *p)
-                       *p = '_';
-       while (path_list_has_path(&current_file_set, newpath) ||
-              path_list_has_path(&current_directory_set, newpath) ||
-              lstat(newpath, &st) == 0)
-               sprintf(p, "_%d", suffix++);
-
-       path_list_insert(newpath, &current_file_set);
-       return newpath;
-}
-
-static int mkdir_p(const char *path, unsigned long mode)
-{
-       /* path points to cache entries, so xstrdup before messing with it */
-       char *buf = xstrdup(path);
-       int result = safe_create_leading_directories(buf);
-       free(buf);
-       return result;
-}
-
-static void flush_buffer(int fd, const char *buf, unsigned long size)
-{
-       while (size > 0) {
-               long ret = write_in_full(fd, buf, size);
-               if (ret < 0) {
-                       /* Ignore epipe */
-                       if (errno == EPIPE)
-                               break;
-                       die("merge-recursive: %s", strerror(errno));
-               } else if (!ret) {
-                       die("merge-recursive: disk full?");
-               }
-               size -= ret;
-               buf += ret;
-       }
-}
-
-static int make_room_for_path(const char *path)
-{
-       int status;
-       const char *msg = "failed to create path '%s'%s";
-
-       status = mkdir_p(path, 0777);
-       if (status) {
-               if (status == -3) {
-                       /* something else exists */
-                       error(msg, path, ": perhaps a D/F conflict?");
-                       return -1;
-               }
-               die(msg, path, "");
-       }
-
-       /* Successful unlink is good.. */
-       if (!unlink(path))
-               return 0;
-       /* .. and so is no existing file */
-       if (errno == ENOENT)
-               return 0;
-       /* .. but not some other error (who really cares what?) */
-       return error(msg, path, ": perhaps a D/F conflict?");
-}
-
-static void update_file_flags(const unsigned char *sha,
-                             unsigned mode,
-                             const char *path,
-                             int update_cache,
-                             int update_wd)
-{
-       if (index_only)
-               update_wd = 0;
-
-       if (update_wd) {
-               enum object_type type;
-               void *buf;
-               unsigned long size;
-
-               if (S_ISGITLINK(mode))
-                       die("cannot read object %s '%s': It is a submodule!",
-                           sha1_to_hex(sha), path);
-
-               buf = read_sha1_file(sha, &type, &size);
-               if (!buf)
-                       die("cannot read object %s '%s'", sha1_to_hex(sha), path);
-               if (type != OBJ_BLOB)
-                       die("blob expected for %s '%s'", sha1_to_hex(sha), path);
-
-               if (make_room_for_path(path) < 0) {
-                       update_wd = 0;
-                       goto update_index;
-               }
-               if (S_ISREG(mode) || (!has_symlinks && S_ISLNK(mode))) {
-                       int fd;
-                       if (mode & 0100)
-                               mode = 0777;
-                       else
-                               mode = 0666;
-                       fd = open(path, O_WRONLY | O_TRUNC | O_CREAT, mode);
-                       if (fd < 0)
-                               die("failed to open %s: %s", path, strerror(errno));
-                       flush_buffer(fd, buf, size);
-                       close(fd);
-               } else if (S_ISLNK(mode)) {
-                       char *lnk = xmemdupz(buf, size);
-                       mkdir_p(path, 0777);
-                       unlink(path);
-                       symlink(lnk, path);
-                       free(lnk);
-               } else
-                       die("do not know what to do with %06o %s '%s'",
-                           mode, sha1_to_hex(sha), path);
-       }
- update_index:
-       if (update_cache)
-               add_cacheinfo(mode, sha, path, 0, update_wd, ADD_CACHE_OK_TO_ADD);
-}
-
-static void update_file(int clean,
-                       const unsigned char *sha,
-                       unsigned mode,
-                       const char *path)
-{
-       update_file_flags(sha, mode, path, index_only || clean, !index_only);
-}
-
-/* Low level file merging, update and removal */
-
-struct merge_file_info
-{
-       unsigned char sha[20];
-       unsigned mode;
-       unsigned clean:1,
-                merge:1;
-};
-
-static void fill_mm(const unsigned char *sha1, mmfile_t *mm)
-{
-       unsigned long size;
-       enum object_type type;
-
-       if (!hashcmp(sha1, null_sha1)) {
-               mm->ptr = xstrdup("");
-               mm->size = 0;
-               return;
-       }
-
-       mm->ptr = read_sha1_file(sha1, &type, &size);
-       if (!mm->ptr || type != OBJ_BLOB)
-               die("unable to read blob object %s", sha1_to_hex(sha1));
-       mm->size = size;
-}
-
-/*
- * Customizable low-level merge drivers support.
- */
-
-struct ll_merge_driver;
-typedef int (*ll_merge_fn)(const struct ll_merge_driver *,
-                          const char *path,
-                          mmfile_t *orig,
-                          mmfile_t *src1, const char *name1,
-                          mmfile_t *src2, const char *name2,
-                          mmbuffer_t *result);
-
-struct ll_merge_driver {
-       const char *name;
-       const char *description;
-       ll_merge_fn fn;
-       const char *recursive;
-       struct ll_merge_driver *next;
-       char *cmdline;
-};
-
-/*
- * Built-in low-levels
- */
-static int ll_binary_merge(const struct ll_merge_driver *drv_unused,
-                          const char *path_unused,
-                          mmfile_t *orig,
-                          mmfile_t *src1, const char *name1,
-                          mmfile_t *src2, const char *name2,
-                          mmbuffer_t *result)
-{
-       /*
-        * The tentative merge result is "ours" for the final round,
-        * or common ancestor for an internal merge.  Still return
-        * "conflicted merge" status.
-        */
-       mmfile_t *stolen = index_only ? orig : src1;
-
-       result->ptr = stolen->ptr;
-       result->size = stolen->size;
-       stolen->ptr = NULL;
-       return 1;
-}
-
-static int ll_xdl_merge(const struct ll_merge_driver *drv_unused,
-                       const char *path_unused,
-                       mmfile_t *orig,
-                       mmfile_t *src1, const char *name1,
-                       mmfile_t *src2, const char *name2,
-                       mmbuffer_t *result)
-{
-       xpparam_t xpp;
-
-       if (buffer_is_binary(orig->ptr, orig->size) ||
-           buffer_is_binary(src1->ptr, src1->size) ||
-           buffer_is_binary(src2->ptr, src2->size)) {
-               warning("Cannot merge binary files: %s vs. %s\n",
-                       name1, name2);
-               return ll_binary_merge(drv_unused, path_unused,
-                                      orig, src1, name1,
-                                      src2, name2,
-                                      result);
-       }
-
-       memset(&xpp, 0, sizeof(xpp));
-       return xdl_merge(orig,
-                        src1, name1,
-                        src2, name2,
-                        &xpp, XDL_MERGE_ZEALOUS,
-                        result);
-}
-
-static int ll_union_merge(const struct ll_merge_driver *drv_unused,
-                         const char *path_unused,
-                         mmfile_t *orig,
-                         mmfile_t *src1, const char *name1,
-                         mmfile_t *src2, const char *name2,
-                         mmbuffer_t *result)
-{
-       char *src, *dst;
-       long size;
-       const int marker_size = 7;
-
-       int status = ll_xdl_merge(drv_unused, path_unused,
-                                 orig, src1, NULL, src2, NULL, result);
-       if (status <= 0)
-               return status;
-       size = result->size;
-       src = dst = result->ptr;
-       while (size) {
-               char ch;
-               if ((marker_size < size) &&
-                   (*src == '<' || *src == '=' || *src == '>')) {
-                       int i;
-                       ch = *src;
-                       for (i = 0; i < marker_size; i++)
-                               if (src[i] != ch)
-                                       goto not_a_marker;
-                       if (src[marker_size] != '\n')
-                               goto not_a_marker;
-                       src += marker_size + 1;
-                       size -= marker_size + 1;
-                       continue;
-               }
-       not_a_marker:
-               do {
-                       ch = *src++;
-                       *dst++ = ch;
-                       size--;
-               } while (ch != '\n' && size);
-       }
-       result->size = dst - result->ptr;
-       return 0;
-}
-
-#define LL_BINARY_MERGE 0
-#define LL_TEXT_MERGE 1
-#define LL_UNION_MERGE 2
-static struct ll_merge_driver ll_merge_drv[] = {
-       { "binary", "built-in binary merge", ll_binary_merge },
-       { "text", "built-in 3-way text merge", ll_xdl_merge },
-       { "union", "built-in union merge", ll_union_merge },
-};
-
-static void create_temp(mmfile_t *src, char *path)
-{
-       int fd;
-
-       strcpy(path, ".merge_file_XXXXXX");
-       fd = xmkstemp(path);
-       if (write_in_full(fd, src->ptr, src->size) != src->size)
-               die("unable to write temp-file");
-       close(fd);
-}
-
-/*
- * User defined low-level merge driver support.
- */
-static int ll_ext_merge(const struct ll_merge_driver *fn,
-                       const char *path,
-                       mmfile_t *orig,
-                       mmfile_t *src1, const char *name1,
-                       mmfile_t *src2, const char *name2,
-                       mmbuffer_t *result)
-{
-       char temp[3][50];
-       char cmdbuf[2048];
-       struct interp table[] = {
-               { "%O" },
-               { "%A" },
-               { "%B" },
-       };
-       struct child_process child;
-       const char *args[20];
-       int status, fd, i;
-       struct stat st;
-
-       if (fn->cmdline == NULL)
-               die("custom merge driver %s lacks command line.", fn->name);
-
-       result->ptr = NULL;
-       result->size = 0;
-       create_temp(orig, temp[0]);
-       create_temp(src1, temp[1]);
-       create_temp(src2, temp[2]);
-
-       interp_set_entry(table, 0, temp[0]);
-       interp_set_entry(table, 1, temp[1]);
-       interp_set_entry(table, 2, temp[2]);
-
-       output(1, "merging %s using %s", path,
-              fn->description ? fn->description : fn->name);
-
-       interpolate(cmdbuf, sizeof(cmdbuf), fn->cmdline, table, 3);
-
-       memset(&child, 0, sizeof(child));
-       child.argv = args;
-       args[0] = "sh";
-       args[1] = "-c";
-       args[2] = cmdbuf;
-       args[3] = NULL;
-
-       status = run_command(&child);
-       if (status < -ERR_RUN_COMMAND_FORK)
-               ; /* failure in run-command */
-       else
-               status = -status;
-       fd = open(temp[1], O_RDONLY);
-       if (fd < 0)
-               goto bad;
-       if (fstat(fd, &st))
-               goto close_bad;
-       result->size = st.st_size;
-       result->ptr = xmalloc(result->size + 1);
-       if (read_in_full(fd, result->ptr, result->size) != result->size) {
-               free(result->ptr);
-               result->ptr = NULL;
-               result->size = 0;
-       }
- close_bad:
-       close(fd);
- bad:
-       for (i = 0; i < 3; i++)
-               unlink(temp[i]);
-       return status;
-}
-
-/*
- * merge.default and merge.driver configuration items
- */
-static struct ll_merge_driver *ll_user_merge, **ll_user_merge_tail;
-static const char *default_ll_merge;
-
-static int read_merge_config(const char *var, const char *value)
-{
-       struct ll_merge_driver *fn;
-       const char *ep, *name;
-       int namelen;
-
-       if (!strcmp(var, "merge.default")) {
-               if (value)
-                       default_ll_merge = strdup(value);
-               return 0;
-       }
-
-       /*
-        * We are not interested in anything but "merge.<name>.variable";
-        * especially, we do not want to look at variables such as
-        * "merge.summary", "merge.tool", and "merge.verbosity".
-        */
-       if (prefixcmp(var, "merge.") || (ep = strrchr(var, '.')) == var + 5)
-               return 0;
-
-       /*
-        * Find existing one as we might be processing merge.<name>.var2
-        * after seeing merge.<name>.var1.
-        */
-       name = var + 6;
-       namelen = ep - name;
-       for (fn = ll_user_merge; fn; fn = fn->next)
-               if (!strncmp(fn->name, name, namelen) && !fn->name[namelen])
-                       break;
-       if (!fn) {
-               fn = xcalloc(1, sizeof(struct ll_merge_driver));
-               fn->name = xmemdupz(name, namelen);
-               fn->fn = ll_ext_merge;
-               *ll_user_merge_tail = fn;
-               ll_user_merge_tail = &(fn->next);
-       }
-
-       ep++;
-
-       if (!strcmp("name", ep)) {
-               if (!value)
-                       return error("%s: lacks value", var);
-               fn->description = strdup(value);
-               return 0;
-       }
-
-       if (!strcmp("driver", ep)) {
-               if (!value)
-                       return error("%s: lacks value", var);
-               /*
-                * merge.<name>.driver specifies the command line:
-                *
-                *      command-line
-                *
-                * The command-line will be interpolated with the following
-                * tokens and is given to the shell:
-                *
-                *    %O - temporary file name for the merge base.
-                *    %A - temporary file name for our version.
-                *    %B - temporary file name for the other branches' version.
-                *
-                * The external merge driver should write the results in the
-                * file named by %A, and signal that it has done with zero exit
-                * status.
-                */
-               fn->cmdline = strdup(value);
-               return 0;
-       }
-
-       if (!strcmp("recursive", ep)) {
-               if (!value)
-                       return error("%s: lacks value", var);
-               fn->recursive = strdup(value);
-               return 0;
-       }
-
-       return 0;
-}
-
-static void initialize_ll_merge(void)
-{
-       if (ll_user_merge_tail)
-               return;
-       ll_user_merge_tail = &ll_user_merge;
-       git_config(read_merge_config);
-}
-
-static const struct ll_merge_driver *find_ll_merge_driver(const char *merge_attr)
-{
-       struct ll_merge_driver *fn;
-       const char *name;
-       int i;
-
-       initialize_ll_merge();
-
-       if (ATTR_TRUE(merge_attr))
-               return &ll_merge_drv[LL_TEXT_MERGE];
-       else if (ATTR_FALSE(merge_attr))
-               return &ll_merge_drv[LL_BINARY_MERGE];
-       else if (ATTR_UNSET(merge_attr)) {
-               if (!default_ll_merge)
-                       return &ll_merge_drv[LL_TEXT_MERGE];
-               else
-                       name = default_ll_merge;
-       }
-       else
-               name = merge_attr;
-
-       for (fn = ll_user_merge; fn; fn = fn->next)
-               if (!strcmp(fn->name, name))
-                       return fn;
-
-       for (i = 0; i < ARRAY_SIZE(ll_merge_drv); i++)
-               if (!strcmp(ll_merge_drv[i].name, name))
-                       return &ll_merge_drv[i];
-
-       /* default to the 3-way */
-       return &ll_merge_drv[LL_TEXT_MERGE];
-}
-
-static const char *git_path_check_merge(const char *path)
-{
-       static struct git_attr_check attr_merge_check;
-
-       if (!attr_merge_check.attr)
-               attr_merge_check.attr = git_attr("merge", 5);
-
-       if (git_checkattr(path, 1, &attr_merge_check))
-               return NULL;
-       return attr_merge_check.value;
-}
-
-static int ll_merge(mmbuffer_t *result_buf,
-                   struct diff_filespec *o,
-                   struct diff_filespec *a,
-                   struct diff_filespec *b,
-                   const char *branch1,
-                   const char *branch2)
-{
-       mmfile_t orig, src1, src2;
-       char *name1, *name2;
-       int merge_status;
-       const char *ll_driver_name;
-       const struct ll_merge_driver *driver;
-
-       name1 = xstrdup(mkpath("%s:%s", branch1, a->path));
-       name2 = xstrdup(mkpath("%s:%s", branch2, b->path));
-
-       fill_mm(o->sha1, &orig);
-       fill_mm(a->sha1, &src1);
-       fill_mm(b->sha1, &src2);
-
-       ll_driver_name = git_path_check_merge(a->path);
-       driver = find_ll_merge_driver(ll_driver_name);
-
-       if (index_only && driver->recursive)
-               driver = find_ll_merge_driver(driver->recursive);
-       merge_status = driver->fn(driver, a->path,
-                                 &orig, &src1, name1, &src2, name2,
-                                 result_buf);
-
-       free(name1);
-       free(name2);
-       free(orig.ptr);
-       free(src1.ptr);
-       free(src2.ptr);
-       return merge_status;
-}
-
-static struct merge_file_info merge_file(struct diff_filespec *o,
-               struct diff_filespec *a, struct diff_filespec *b,
-               const char *branch1, const char *branch2)
-{
-       struct merge_file_info result;
-       result.merge = 0;
-       result.clean = 1;
-
-       if ((S_IFMT & a->mode) != (S_IFMT & b->mode)) {
-               result.clean = 0;
-               if (S_ISREG(a->mode)) {
-                       result.mode = a->mode;
-                       hashcpy(result.sha, a->sha1);
-               } else {
-                       result.mode = b->mode;
-                       hashcpy(result.sha, b->sha1);
-               }
-       } else {
-               if (!sha_eq(a->sha1, o->sha1) && !sha_eq(b->sha1, o->sha1))
-                       result.merge = 1;
-
-               result.mode = a->mode == o->mode ? b->mode: a->mode;
-
-               if (sha_eq(a->sha1, o->sha1))
-                       hashcpy(result.sha, b->sha1);
-               else if (sha_eq(b->sha1, o->sha1))
-                       hashcpy(result.sha, a->sha1);
-               else if (S_ISREG(a->mode)) {
-                       mmbuffer_t result_buf;
-                       int merge_status;
-
-                       merge_status = ll_merge(&result_buf, o, a, b,
-                                               branch1, branch2);
-
-                       if ((merge_status < 0) || !result_buf.ptr)
-                               die("Failed to execute internal merge");
-
-                       if (write_sha1_file(result_buf.ptr, result_buf.size,
-                                           blob_type, result.sha))
-                               die("Unable to add %s to database",
-                                   a->path);
-
-                       free(result_buf.ptr);
-                       result.clean = (merge_status == 0);
-               } else if (S_ISGITLINK(a->mode)) {
-                       result.clean = 0;
-                       hashcpy(result.sha, a->sha1);
-               } else if (S_ISLNK(a->mode)) {
-                       hashcpy(result.sha, a->sha1);
-
-                       if (!sha_eq(a->sha1, b->sha1))
-                               result.clean = 0;
-               } else {
-                       die("unsupported object type in the tree");
-               }
-       }
-
-       return result;
-}
-
-static void conflict_rename_rename(struct rename *ren1,
-                                  const char *branch1,
-                                  struct rename *ren2,
-                                  const char *branch2)
-{
-       char *del[2];
-       int delp = 0;
-       const char *ren1_dst = ren1->pair->two->path;
-       const char *ren2_dst = ren2->pair->two->path;
-       const char *dst_name1 = ren1_dst;
-       const char *dst_name2 = ren2_dst;
-       if (path_list_has_path(&current_directory_set, ren1_dst)) {
-               dst_name1 = del[delp++] = unique_path(ren1_dst, branch1);
-               output(1, "%s is a directory in %s added as %s instead",
-                      ren1_dst, branch2, dst_name1);
-               remove_file(0, ren1_dst, 0);
-       }
-       if (path_list_has_path(&current_directory_set, ren2_dst)) {
-               dst_name2 = del[delp++] = unique_path(ren2_dst, branch2);
-               output(1, "%s is a directory in %s added as %s instead",
-                      ren2_dst, branch1, dst_name2);
-               remove_file(0, ren2_dst, 0);
-       }
-       if (index_only) {
-               remove_file_from_cache(dst_name1);
-               remove_file_from_cache(dst_name2);
-               /*
-                * Uncomment to leave the conflicting names in the resulting tree
-                *
-                * update_file(0, ren1->pair->two->sha1, ren1->pair->two->mode, dst_name1);
-                * update_file(0, ren2->pair->two->sha1, ren2->pair->two->mode, dst_name2);
-                */
-       } else {
-               update_stages(dst_name1, NULL, ren1->pair->two, NULL, 1);
-               update_stages(dst_name2, NULL, NULL, ren2->pair->two, 1);
-       }
-       while (delp--)
-               free(del[delp]);
-}
-
-static void conflict_rename_dir(struct rename *ren1,
-                               const char *branch1)
-{
-       char *new_path = unique_path(ren1->pair->two->path, branch1);
-       output(1, "Renamed %s to %s instead", ren1->pair->one->path, new_path);
-       remove_file(0, ren1->pair->two->path, 0);
-       update_file(0, ren1->pair->two->sha1, ren1->pair->two->mode, new_path);
-       free(new_path);
-}
-
-static void conflict_rename_rename_2(struct rename *ren1,
-                                    const char *branch1,
-                                    struct rename *ren2,
-                                    const char *branch2)
-{
-       char *new_path1 = unique_path(ren1->pair->two->path, branch1);
-       char *new_path2 = unique_path(ren2->pair->two->path, branch2);
-       output(1, "Renamed %s to %s and %s to %s instead",
-              ren1->pair->one->path, new_path1,
-              ren2->pair->one->path, new_path2);
-       remove_file(0, ren1->pair->two->path, 0);
-       update_file(0, ren1->pair->two->sha1, ren1->pair->two->mode, new_path1);
-       update_file(0, ren2->pair->two->sha1, ren2->pair->two->mode, new_path2);
-       free(new_path2);
-       free(new_path1);
-}
-
-static int process_renames(struct path_list *a_renames,
-                          struct path_list *b_renames,
-                          const char *a_branch,
-                          const char *b_branch)
-{
-       int clean_merge = 1, i, j;
-       struct path_list a_by_dst = {NULL, 0, 0, 0}, b_by_dst = {NULL, 0, 0, 0};
-       const struct rename *sre;
-
-       for (i = 0; i < a_renames->nr; i++) {
-               sre = a_renames->items[i].util;
-               path_list_insert(sre->pair->two->path, &a_by_dst)->util
-                       = sre->dst_entry;
-       }
-       for (i = 0; i < b_renames->nr; i++) {
-               sre = b_renames->items[i].util;
-               path_list_insert(sre->pair->two->path, &b_by_dst)->util
-                       = sre->dst_entry;
-       }
-
-       for (i = 0, j = 0; i < a_renames->nr || j < b_renames->nr;) {
-               int compare;
-               char *src;
-               struct path_list *renames1, *renames2, *renames2Dst;
-               struct rename *ren1 = NULL, *ren2 = NULL;
-               const char *branch1, *branch2;
-               const char *ren1_src, *ren1_dst;
-
-               if (i >= a_renames->nr) {
-                       compare = 1;
-                       ren2 = b_renames->items[j++].util;
-               } else if (j >= b_renames->nr) {
-                       compare = -1;
-                       ren1 = a_renames->items[i++].util;
-               } else {
-                       compare = strcmp(a_renames->items[i].path,
-                                       b_renames->items[j].path);
-                       if (compare <= 0)
-                               ren1 = a_renames->items[i++].util;
-                       if (compare >= 0)
-                               ren2 = b_renames->items[j++].util;
-               }
-
-               /* TODO: refactor, so that 1/2 are not needed */
-               if (ren1) {
-                       renames1 = a_renames;
-                       renames2 = b_renames;
-                       renames2Dst = &b_by_dst;
-                       branch1 = a_branch;
-                       branch2 = b_branch;
-               } else {
-                       struct rename *tmp;
-                       renames1 = b_renames;
-                       renames2 = a_renames;
-                       renames2Dst = &a_by_dst;
-                       branch1 = b_branch;
-                       branch2 = a_branch;
-                       tmp = ren2;
-                       ren2 = ren1;
-                       ren1 = tmp;
-               }
-               src = ren1->pair->one->path;
-
-               ren1->dst_entry->processed = 1;
-               ren1->src_entry->processed = 1;
-
-               if (ren1->processed)
-                       continue;
-               ren1->processed = 1;
-
-               ren1_src = ren1->pair->one->path;
-               ren1_dst = ren1->pair->two->path;
-
-               if (ren2) {
-                       const char *ren2_src = ren2->pair->one->path;
-                       const char *ren2_dst = ren2->pair->two->path;
-                       /* Renamed in 1 and renamed in 2 */
-                       if (strcmp(ren1_src, ren2_src) != 0)
-                               die("ren1.src != ren2.src");
-                       ren2->dst_entry->processed = 1;
-                       ren2->processed = 1;
-                       if (strcmp(ren1_dst, ren2_dst) != 0) {
-                               clean_merge = 0;
-                               output(1, "CONFLICT (rename/rename): "
-                                      "Rename \"%s\"->\"%s\" in branch \"%s\" "
-                                      "rename \"%s\"->\"%s\" in \"%s\"%s",
-                                      src, ren1_dst, branch1,
-                                      src, ren2_dst, branch2,
-                                      index_only ? " (left unresolved)": "");
-                               if (index_only) {
-                                       remove_file_from_cache(src);
-                                       update_file(0, ren1->pair->one->sha1,
-                                                   ren1->pair->one->mode, src);
-                               }
-                               conflict_rename_rename(ren1, branch1, ren2, branch2);
-                       } else {
-                               struct merge_file_info mfi;
-                               remove_file(1, ren1_src, 1);
-                               mfi = merge_file(ren1->pair->one,
-                                                ren1->pair->two,
-                                                ren2->pair->two,
-                                                branch1,
-                                                branch2);
-                               if (mfi.merge || !mfi.clean)
-                                       output(1, "Renamed %s->%s", src, ren1_dst);
-
-                               if (mfi.merge)
-                                       output(2, "Auto-merged %s", ren1_dst);
-
-                               if (!mfi.clean) {
-                                       output(1, "CONFLICT (content): merge conflict in %s",
-                                              ren1_dst);
-                                       clean_merge = 0;
-
-                                       if (!index_only)
-                                               update_stages(ren1_dst,
-                                                             ren1->pair->one,
-                                                             ren1->pair->two,
-                                                             ren2->pair->two,
-                                                             1 /* clear */);
-                               }
-                               update_file(mfi.clean, mfi.sha, mfi.mode, ren1_dst);
-                       }
-               } else {
-                       /* Renamed in 1, maybe changed in 2 */
-                       struct path_list_item *item;
-                       /* we only use sha1 and mode of these */
-                       struct diff_filespec src_other, dst_other;
-                       int try_merge, stage = a_renames == renames1 ? 3: 2;
-
-                       remove_file(1, ren1_src, index_only || stage == 3);
-
-                       hashcpy(src_other.sha1, ren1->src_entry->stages[stage].sha);
-                       src_other.mode = ren1->src_entry->stages[stage].mode;
-                       hashcpy(dst_other.sha1, ren1->dst_entry->stages[stage].sha);
-                       dst_other.mode = ren1->dst_entry->stages[stage].mode;
-
-                       try_merge = 0;
-
-                       if (path_list_has_path(&current_directory_set, ren1_dst)) {
-                               clean_merge = 0;
-                               output(1, "CONFLICT (rename/directory): Renamed %s->%s in %s "
-                                      " directory %s added in %s",
-                                      ren1_src, ren1_dst, branch1,
-                                      ren1_dst, branch2);
-                               conflict_rename_dir(ren1, branch1);
-                       } else if (sha_eq(src_other.sha1, null_sha1)) {
-                               clean_merge = 0;
-                               output(1, "CONFLICT (rename/delete): Renamed %s->%s in %s "
-                                      "and deleted in %s",
-                                      ren1_src, ren1_dst, branch1,
-                                      branch2);
-                               update_file(0, ren1->pair->two->sha1, ren1->pair->two->mode, ren1_dst);
-                       } else if (!sha_eq(dst_other.sha1, null_sha1)) {
-                               const char *new_path;
-                               clean_merge = 0;
-                               try_merge = 1;
-                               output(1, "CONFLICT (rename/add): Renamed %s->%s in %s. "
-                                      "%s added in %s",
-                                      ren1_src, ren1_dst, branch1,
-                                      ren1_dst, branch2);
-                               new_path = unique_path(ren1_dst, branch2);
-                               output(1, "Added as %s instead", new_path);
-                               update_file(0, dst_other.sha1, dst_other.mode, new_path);
-                       } else if ((item = path_list_lookup(ren1_dst, renames2Dst))) {
-                               ren2 = item->util;
-                               clean_merge = 0;
-                               ren2->processed = 1;
-                               output(1, "CONFLICT (rename/rename): Renamed %s->%s in %s. "
-                                      "Renamed %s->%s in %s",
-                                      ren1_src, ren1_dst, branch1,
-                                      ren2->pair->one->path, ren2->pair->two->path, branch2);
-                               conflict_rename_rename_2(ren1, branch1, ren2, branch2);
-                       } else
-                               try_merge = 1;
-
-                       if (try_merge) {
-                               struct diff_filespec *o, *a, *b;
-                               struct merge_file_info mfi;
-                               src_other.path = (char *)ren1_src;
-
-                               o = ren1->pair->one;
-                               if (a_renames == renames1) {
-                                       a = ren1->pair->two;
-                                       b = &src_other;
-                               } else {
-                                       b = ren1->pair->two;
-                                       a = &src_other;
-                               }
-                               mfi = merge_file(o, a, b,
-                                               a_branch, b_branch);
-
-                               if (mfi.clean &&
-                                   sha_eq(mfi.sha, ren1->pair->two->sha1) &&
-                                   mfi.mode == ren1->pair->two->mode)
-                                       /*
-                                        * This messaged is part of
-                                        * t6022 test. If you change
-                                        * it update the test too.
-                                        */
-                                       output(3, "Skipped %s (merged same as existing)", ren1_dst);
-                               else {
-                                       if (mfi.merge || !mfi.clean)
-                                               output(1, "Renamed %s => %s", ren1_src, ren1_dst);
-                                       if (mfi.merge)
-                                               output(2, "Auto-merged %s", ren1_dst);
-                                       if (!mfi.clean) {
-                                               output(1, "CONFLICT (rename/modify): Merge conflict in %s",
-                                                      ren1_dst);
-                                               clean_merge = 0;
-
-                                               if (!index_only)
-                                                       update_stages(ren1_dst,
-                                                                     o, a, b, 1);
-                                       }
-                                       update_file(mfi.clean, mfi.sha, mfi.mode, ren1_dst);
-                               }
-                       }
-               }
-       }
-       path_list_clear(&a_by_dst, 0);
-       path_list_clear(&b_by_dst, 0);
-
-       return clean_merge;
-}
-
-static unsigned char *stage_sha(const unsigned char *sha, unsigned mode)
-{
-       return (is_null_sha1(sha) || mode == 0) ? NULL: (unsigned char *)sha;
-}
-
-/* Per entry merge function */
-static int process_entry(const char *path, struct stage_data *entry,
-                        const char *branch1,
-                        const char *branch2)
-{
-       /*
-       printf("processing entry, clean cache: %s\n", index_only ? "yes": "no");
-       print_index_entry("\tpath: ", entry);
-       */
-       int clean_merge = 1;
-       unsigned o_mode = entry->stages[1].mode;
-       unsigned a_mode = entry->stages[2].mode;
-       unsigned b_mode = entry->stages[3].mode;
-       unsigned char *o_sha = stage_sha(entry->stages[1].sha, o_mode);
-       unsigned char *a_sha = stage_sha(entry->stages[2].sha, a_mode);
-       unsigned char *b_sha = stage_sha(entry->stages[3].sha, b_mode);
-
-       if (o_sha && (!a_sha || !b_sha)) {
-               /* Case A: Deleted in one */
-               if ((!a_sha && !b_sha) ||
-                   (sha_eq(a_sha, o_sha) && !b_sha) ||
-                   (!a_sha && sha_eq(b_sha, o_sha))) {
-                       /* Deleted in both or deleted in one and
-                        * unchanged in the other */
-                       if (a_sha)
-                               output(2, "Removed %s", path);
-                       /* do not touch working file if it did not exist */
-                       remove_file(1, path, !a_sha);
-               } else {
-                       /* Deleted in one and changed in the other */
-                       clean_merge = 0;
-                       if (!a_sha) {
-                               output(1, "CONFLICT (delete/modify): %s deleted in %s "
-                                      "and modified in %s. Version %s of %s left in tree.",
-                                      path, branch1,
-                                      branch2, branch2, path);
-                               update_file(0, b_sha, b_mode, path);
-                       } else {
-                               output(1, "CONFLICT (delete/modify): %s deleted in %s "
-                                      "and modified in %s. Version %s of %s left in tree.",
-                                      path, branch2,
-                                      branch1, branch1, path);
-                               update_file(0, a_sha, a_mode, path);
-                       }
-               }
-
-       } else if ((!o_sha && a_sha && !b_sha) ||
-                  (!o_sha && !a_sha && b_sha)) {
-               /* Case B: Added in one. */
-               const char *add_branch;
-               const char *other_branch;
-               unsigned mode;
-               const unsigned char *sha;
-               const char *conf;
-
-               if (a_sha) {
-                       add_branch = branch1;
-                       other_branch = branch2;
-                       mode = a_mode;
-                       sha = a_sha;
-                       conf = "file/directory";
-               } else {
-                       add_branch = branch2;
-                       other_branch = branch1;
-                       mode = b_mode;
-                       sha = b_sha;
-                       conf = "directory/file";
-               }
-               if (path_list_has_path(&current_directory_set, path)) {
-                       const char *new_path = unique_path(path, add_branch);
-                       clean_merge = 0;
-                       output(1, "CONFLICT (%s): There is a directory with name %s in %s. "
-                              "Added %s as %s",
-                              conf, path, other_branch, path, new_path);
-                       remove_file(0, path, 0);
-                       update_file(0, sha, mode, new_path);
-               } else {
-                       output(2, "Added %s", path);
-                       update_file(1, sha, mode, path);
-               }
-       } else if (a_sha && b_sha) {
-               /* Case C: Added in both (check for same permissions) and */
-               /* case D: Modified in both, but differently. */
-               const char *reason = "content";
-               struct merge_file_info mfi;
-               struct diff_filespec o, a, b;
-
-               if (!o_sha) {
-                       reason = "add/add";
-                       o_sha = (unsigned char *)null_sha1;
-               }
-               output(2, "Auto-merged %s", path);
-               o.path = a.path = b.path = (char *)path;
-               hashcpy(o.sha1, o_sha);
-               o.mode = o_mode;
-               hashcpy(a.sha1, a_sha);
-               a.mode = a_mode;
-               hashcpy(b.sha1, b_sha);
-               b.mode = b_mode;
-
-               mfi = merge_file(&o, &a, &b,
-                                branch1, branch2);
-
-               clean_merge = mfi.clean;
-               if (mfi.clean)
-                       update_file(1, mfi.sha, mfi.mode, path);
-               else if (S_ISGITLINK(mfi.mode))
-                       output(1, "CONFLICT (submodule): Merge conflict in %s "
-                              "- needs %s", path, sha1_to_hex(b.sha1));
-               else {
-                       output(1, "CONFLICT (%s): Merge conflict in %s",
-                                       reason, path);
-
-                       if (index_only)
-                               update_file(0, mfi.sha, mfi.mode, path);
-                       else
-                               update_file_flags(mfi.sha, mfi.mode, path,
-                                             0 /* update_cache */, 1 /* update_working_directory */);
-               }
-       } else if (!o_sha && !a_sha && !b_sha) {
-               /*
-                * this entry was deleted altogether. a_mode == 0 means
-                * we had that path and want to actively remove it.
-                */
-               remove_file(1, path, !a_mode);
-       } else
-               die("Fatal merge failure, shouldn't happen.");
-
-       return clean_merge;
-}
-
-static int merge_trees(struct tree *head,
-                      struct tree *merge,
-                      struct tree *common,
-                      const char *branch1,
-                      const char *branch2,
-                      struct tree **result)
-{
-       int code, clean;
-
-       if (subtree_merge) {
-               merge = shift_tree_object(head, merge);
-               common = shift_tree_object(head, common);
-       }
-
-       if (sha_eq(common->object.sha1, merge->object.sha1)) {
-               output(0, "Already uptodate!");
-               *result = head;
-               return 1;
-       }
-
-       code = git_merge_trees(index_only, common, head, merge);
-
-       if (code != 0)
-               die("merging of trees %s and %s failed",
-                   sha1_to_hex(head->object.sha1),
-                   sha1_to_hex(merge->object.sha1));
-
-       if (unmerged_index()) {
-               struct path_list *entries, *re_head, *re_merge;
-               int i;
-               path_list_clear(&current_file_set, 1);
-               path_list_clear(&current_directory_set, 1);
-               get_files_dirs(head);
-               get_files_dirs(merge);
-
-               entries = get_unmerged();
-               re_head  = get_renames(head, common, head, merge, entries);
-               re_merge = get_renames(merge, common, head, merge, entries);
-               clean = process_renames(re_head, re_merge,
-                               branch1, branch2);
-               for (i = 0; i < entries->nr; i++) {
-                       const char *path = entries->items[i].path;
-                       struct stage_data *e = entries->items[i].util;
-                       if (!e->processed
-                               && !process_entry(path, e, branch1, branch2))
-                               clean = 0;
-               }
-
-               path_list_clear(re_merge, 0);
-               path_list_clear(re_head, 0);
-               path_list_clear(entries, 1);
-
-       }
-       else
-               clean = 1;
-
-       if (index_only)
-               *result = git_write_tree();
-
-       return clean;
-}
-
-static struct commit_list *reverse_commit_list(struct commit_list *list)
-{
-       struct commit_list *next = NULL, *current, *backup;
-       for (current = list; current; current = backup) {
-               backup = current->next;
-               current->next = next;
-               next = current;
-       }
-       return next;
-}
-
-/*
- * Merge the commits h1 and h2, return the resulting virtual
- * commit object and a flag indicating the cleanness of the merge.
- */
-static int merge(struct commit *h1,
-                struct commit *h2,
-                const char *branch1,
-                const char *branch2,
-                struct commit_list *ca,
-                struct commit **result)
-{
-       struct commit_list *iter;
-       struct commit *merged_common_ancestors;
-       struct tree *mrtree = mrtree;
-       int clean;
-
-       if (show(4)) {
-               output(4, "Merging:");
-               output_commit_title(h1);
-               output_commit_title(h2);
-       }
-
-       if (!ca) {
-               ca = get_merge_bases(h1, h2, 1);
-               ca = reverse_commit_list(ca);
-       }
-
-       if (show(5)) {
-               output(5, "found %u common ancestor(s):", commit_list_count(ca));
-               for (iter = ca; iter; iter = iter->next)
-                       output_commit_title(iter->item);
-       }
-
-       merged_common_ancestors = pop_commit(&ca);
-       if (merged_common_ancestors == NULL) {
-               /* if there is no common ancestor, make an empty tree */
-               struct tree *tree = xcalloc(1, sizeof(struct tree));
-
-               tree->object.parsed = 1;
-               tree->object.type = OBJ_TREE;
-               pretend_sha1_file(NULL, 0, OBJ_TREE, tree->object.sha1);
-               merged_common_ancestors = make_virtual_commit(tree, "ancestor");
-       }
-
-       for (iter = ca; iter; iter = iter->next) {
-               call_depth++;
-               /*
-                * When the merge fails, the result contains files
-                * with conflict markers. The cleanness flag is
-                * ignored, it was never actually used, as result of
-                * merge_trees has always overwritten it: the committed
-                * "conflicts" were already resolved.
-                */
-               discard_cache();
-               merge(merged_common_ancestors, iter->item,
-                     "Temporary merge branch 1",
-                     "Temporary merge branch 2",
-                     NULL,
-                     &merged_common_ancestors);
-               call_depth--;
-
-               if (!merged_common_ancestors)
-                       die("merge returned no commit");
-       }
-
-       discard_cache();
-       if (!call_depth) {
-               read_cache();
-               index_only = 0;
-       } else
-               index_only = 1;
-
-       clean = merge_trees(h1->tree, h2->tree, merged_common_ancestors->tree,
-                           branch1, branch2, &mrtree);
-
-       if (index_only) {
-               *result = make_virtual_commit(mrtree, "merged tree");
-               commit_list_insert(h1, &(*result)->parents);
-               commit_list_insert(h2, &(*result)->parents->next);
-       }
-       flush_output();
-       return clean;
-}
-
-static const char *better_branch_name(const char *branch)
-{
-       static char githead_env[8 + 40 + 1];
-       char *name;
-
-       if (strlen(branch) != 40)
-               return branch;
-       sprintf(githead_env, "GITHEAD_%s", branch);
-       name = getenv(githead_env);
-       return name ? name : branch;
-}
-
-static struct commit *get_ref(const char *ref)
-{
-       unsigned char sha1[20];
-       struct object *object;
-
-       if (get_sha1(ref, sha1))
-               die("Could not resolve ref '%s'", ref);
-       object = deref_tag(parse_object(sha1), ref, strlen(ref));
-       if (object->type == OBJ_TREE)
-               return make_virtual_commit((struct tree*)object,
-                       better_branch_name(ref));
-       if (object->type != OBJ_COMMIT)
-               return NULL;
-       if (parse_commit((struct commit *)object))
-               die("Could not parse commit '%s'", sha1_to_hex(object->sha1));
-       return (struct commit *)object;
-}
-
-static int merge_config(const char *var, const char *value)
-{
-       if (!strcasecmp(var, "merge.verbosity")) {
-               verbosity = git_config_int(var, value);
-               return 0;
-       }
-       if (!strcasecmp(var, "diff.renamelimit")) {
-               rename_limit = git_config_int(var, value);
-               return 0;
-       }
-       return git_default_config(var, value);
-}
-
-int main(int argc, char *argv[])
-{
-       static const char *bases[20];
-       static unsigned bases_count = 0;
-       int i, clean;
-       const char *branch1, *branch2;
-       struct commit *result, *h1, *h2;
-       struct commit_list *ca = NULL;
-       struct lock_file *lock = xcalloc(1, sizeof(struct lock_file));
-       int index_fd;
-
-       if (argv[0]) {
-               int namelen = strlen(argv[0]);
-               if (8 < namelen &&
-                   !strcmp(argv[0] + namelen - 8, "-subtree"))
-                       subtree_merge = 1;
-       }
-
-       git_config(merge_config);
-       if (getenv("GIT_MERGE_VERBOSITY"))
-               verbosity = strtol(getenv("GIT_MERGE_VERBOSITY"), NULL, 10);
-
-       if (argc < 4)
-               die("Usage: %s <base>... -- <head> <remote> ...\n", argv[0]);
-
-       for (i = 1; i < argc; ++i) {
-               if (!strcmp(argv[i], "--"))
-                       break;
-               if (bases_count < sizeof(bases)/sizeof(*bases))
-                       bases[bases_count++] = argv[i];
-       }
-       if (argc - i != 3) /* "--" "<head>" "<remote>" */
-               die("Not handling anything other than two heads merge.");
-       if (verbosity >= 5)
-               buffer_output = 0;
-
-       branch1 = argv[++i];
-       branch2 = argv[++i];
-
-       h1 = get_ref(branch1);
-       h2 = get_ref(branch2);
-
-       branch1 = better_branch_name(branch1);
-       branch2 = better_branch_name(branch2);
-
-       if (show(3))
-               printf("Merging %s with %s\n", branch1, branch2);
-
-       index_fd = hold_locked_index(lock, 1);
-
-       for (i = 0; i < bases_count; i++) {
-               struct commit *ancestor = get_ref(bases[i]);
-               ca = commit_list_insert(ancestor, &ca);
-       }
-       clean = merge(h1, h2, branch1, branch2, ca, &result);
-
-       if (active_cache_changed &&
-           (write_cache(index_fd, active_cache, active_nr) ||
-            commit_locked_index(lock)))
-                       die ("unable to write %s", get_index_file());
-
-       return clean ? 0: 1;
-}
diff --git a/merge-recursive.h b/merge-recursive.h
new file mode 100644 (file)
index 0000000..f37630a
--- /dev/null
@@ -0,0 +1,20 @@
+#ifndef MERGE_RECURSIVE_H
+#define MERGE_RECURSIVE_H
+
+int merge_recursive(struct commit *h1,
+                   struct commit *h2,
+                   const char *branch1,
+                   const char *branch2,
+                   struct commit_list *ancestors,
+                   struct commit **result);
+
+int merge_trees(struct tree *head,
+               struct tree *merge,
+               struct tree *common,
+               const char *branch1,
+               const char *branch2,
+               struct tree **result);
+
+struct tree *write_tree_from_memory(void);
+
+#endif