OSDN Git Service

Merge branch 'kn/for-each-ref'
authorJunio C Hamano <gitster@pobox.com>
Mon, 3 Aug 2015 18:01:10 +0000 (11:01 -0700)
committerJunio C Hamano <gitster@pobox.com>
Mon, 3 Aug 2015 18:01:11 +0000 (11:01 -0700)
GSoC project to rebuild ref listing by branch and tag based on the
for-each-ref machinery.  This is its first part.

* kn/for-each-ref:
  ref-filter: make 'ref_array_item' use a FLEX_ARRAY for refname
  for-each-ref: introduce filter_refs()
  ref-filter: move code from 'for-each-ref'
  ref-filter: add 'ref-filter.h'
  for-each-ref: rename variables called sort to sorting
  for-each-ref: rename some functions and make them public
  for-each-ref: introduce 'ref_array_clear()'
  for-each-ref: introduce new structures for better organisation
  for-each-ref: rename 'refinfo' to 'ref_array_item'
  for-each-ref: clean up code
  for-each-ref: extract helper functions out of grab_single_ref()

1  2 
Makefile
ref-filter.c

diff --cc Makefile
Simple merge
diff --cc ref-filter.c
index 0000000,3fbbbeb..6e09da8
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,1101 +1,1106 @@@
+ #include "builtin.h"
+ #include "cache.h"
+ #include "parse-options.h"
+ #include "refs.h"
+ #include "wildmatch.h"
+ #include "commit.h"
+ #include "remote.h"
+ #include "color.h"
+ #include "tag.h"
+ #include "quote.h"
+ #include "ref-filter.h"
+ typedef enum { FIELD_STR, FIELD_ULONG, FIELD_TIME } cmp_type;
+ static struct {
+       const char *name;
+       cmp_type cmp_type;
+ } valid_atom[] = {
+       { "refname" },
+       { "objecttype" },
+       { "objectsize", FIELD_ULONG },
+       { "objectname" },
+       { "tree" },
+       { "parent" },
+       { "numparent", FIELD_ULONG },
+       { "object" },
+       { "type" },
+       { "tag" },
+       { "author" },
+       { "authorname" },
+       { "authoremail" },
+       { "authordate", FIELD_TIME },
+       { "committer" },
+       { "committername" },
+       { "committeremail" },
+       { "committerdate", FIELD_TIME },
+       { "tagger" },
+       { "taggername" },
+       { "taggeremail" },
+       { "taggerdate", FIELD_TIME },
+       { "creator" },
+       { "creatordate", FIELD_TIME },
+       { "subject" },
+       { "body" },
+       { "contents" },
+       { "contents:subject" },
+       { "contents:body" },
+       { "contents:signature" },
+       { "upstream" },
+       { "push" },
+       { "symref" },
+       { "flag" },
+       { "HEAD" },
+       { "color" },
+ };
+ /*
+  * An atom is a valid field atom listed above, possibly prefixed with
+  * a "*" to denote deref_tag().
+  *
+  * We parse given format string and sort specifiers, and make a list
+  * of properties that we need to extract out of objects.  ref_array_item
+  * structure will hold an array of values extracted that can be
+  * indexed with the "atom number", which is an index into this
+  * array.
+  */
+ static const char **used_atom;
+ static cmp_type *used_atom_type;
+ static int used_atom_cnt, need_tagged, need_symref;
+ static int need_color_reset_at_eol;
+ /*
+  * Used to parse format string and sort specifiers
+  */
+ int parse_ref_filter_atom(const char *atom, const char *ep)
+ {
+       const char *sp;
+       int i, at;
+       sp = atom;
+       if (*sp == '*' && sp < ep)
+               sp++; /* deref */
+       if (ep <= sp)
+               die("malformed field name: %.*s", (int)(ep-atom), atom);
+       /* Do we have the atom already used elsewhere? */
+       for (i = 0; i < used_atom_cnt; i++) {
+               int len = strlen(used_atom[i]);
+               if (len == ep - atom && !memcmp(used_atom[i], atom, len))
+                       return i;
+       }
+       /* Is the atom a valid one? */
+       for (i = 0; i < ARRAY_SIZE(valid_atom); i++) {
+               int len = strlen(valid_atom[i].name);
+               /*
+                * If the atom name has a colon, strip it and everything after
+                * it off - it specifies the format for this entry, and
+                * shouldn't be used for checking against the valid_atom
+                * table.
+                */
+               const char *formatp = strchr(sp, ':');
+               if (!formatp || ep < formatp)
+                       formatp = ep;
+               if (len == formatp - sp && !memcmp(valid_atom[i].name, sp, len))
+                       break;
+       }
+       if (ARRAY_SIZE(valid_atom) <= i)
+               die("unknown field name: %.*s", (int)(ep-atom), atom);
+       /* Add it in, including the deref prefix */
+       at = used_atom_cnt;
+       used_atom_cnt++;
+       REALLOC_ARRAY(used_atom, used_atom_cnt);
+       REALLOC_ARRAY(used_atom_type, used_atom_cnt);
+       used_atom[at] = xmemdupz(atom, ep - atom);
+       used_atom_type[at] = valid_atom[i].cmp_type;
+       if (*atom == '*')
+               need_tagged = 1;
+       if (!strcmp(used_atom[at], "symref"))
+               need_symref = 1;
+       return at;
+ }
+ /*
+  * In a format string, find the next occurrence of %(atom).
+  */
+ static const char *find_next(const char *cp)
+ {
+       while (*cp) {
+               if (*cp == '%') {
+                       /*
+                        * %( is the start of an atom;
+                        * %% is a quoted per-cent.
+                        */
+                       if (cp[1] == '(')
+                               return cp;
+                       else if (cp[1] == '%')
+                               cp++; /* skip over two % */
+                       /* otherwise this is a singleton, literal % */
+               }
+               cp++;
+       }
+       return NULL;
+ }
+ /*
+  * Make sure the format string is well formed, and parse out
+  * the used atoms.
+  */
+ int verify_ref_format(const char *format)
+ {
+       const char *cp, *sp;
+       need_color_reset_at_eol = 0;
+       for (cp = format; *cp && (sp = find_next(cp)); ) {
+               const char *color, *ep = strchr(sp, ')');
+               int at;
+               if (!ep)
+                       return error("malformed format string %s", sp);
+               /* sp points at "%(" and ep points at the closing ")" */
+               at = parse_ref_filter_atom(sp + 2, ep);
+               cp = ep + 1;
+               if (skip_prefix(used_atom[at], "color:", &color))
+                       need_color_reset_at_eol = !!strcmp(color, "reset");
+       }
+       return 0;
+ }
+ /*
+  * Given an object name, read the object data and size, and return a
+  * "struct object".  If the object data we are returning is also borrowed
+  * by the "struct object" representation, set *eaten as well---it is a
+  * signal from parse_object_buffer to us not to free the buffer.
+  */
+ static void *get_obj(const unsigned char *sha1, struct object **obj, unsigned long *sz, int *eaten)
+ {
+       enum object_type type;
+       void *buf = read_sha1_file(sha1, &type, sz);
+       if (buf)
+               *obj = parse_object_buffer(sha1, type, *sz, buf, eaten);
+       else
+               *obj = NULL;
+       return buf;
+ }
+ static int grab_objectname(const char *name, const unsigned char *sha1,
+                           struct atom_value *v)
+ {
+       if (!strcmp(name, "objectname")) {
+               char *s = xmalloc(41);
+               strcpy(s, sha1_to_hex(sha1));
+               v->s = s;
+               return 1;
+       }
+       if (!strcmp(name, "objectname:short")) {
+               v->s = xstrdup(find_unique_abbrev(sha1, DEFAULT_ABBREV));
+               return 1;
+       }
+       return 0;
+ }
+ /* See grab_values */
+ static void grab_common_values(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
+ {
+       int i;
+       for (i = 0; i < used_atom_cnt; i++) {
+               const char *name = used_atom[i];
+               struct atom_value *v = &val[i];
+               if (!!deref != (*name == '*'))
+                       continue;
+               if (deref)
+                       name++;
+               if (!strcmp(name, "objecttype"))
+                       v->s = typename(obj->type);
+               else if (!strcmp(name, "objectsize")) {
+                       char *s = xmalloc(40);
+                       sprintf(s, "%lu", sz);
+                       v->ul = sz;
+                       v->s = s;
+               }
+               else if (deref)
+                       grab_objectname(name, obj->sha1, v);
+       }
+ }
+ /* See grab_values */
+ static void grab_tag_values(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
+ {
+       int i;
+       struct tag *tag = (struct tag *) obj;
+       for (i = 0; i < used_atom_cnt; i++) {
+               const char *name = used_atom[i];
+               struct atom_value *v = &val[i];
+               if (!!deref != (*name == '*'))
+                       continue;
+               if (deref)
+                       name++;
+               if (!strcmp(name, "tag"))
+                       v->s = tag->tag;
+               else if (!strcmp(name, "type") && tag->tagged)
+                       v->s = typename(tag->tagged->type);
+               else if (!strcmp(name, "object") && tag->tagged) {
+                       char *s = xmalloc(41);
+                       strcpy(s, sha1_to_hex(tag->tagged->sha1));
+                       v->s = s;
+               }
+       }
+ }
+ /* See grab_values */
+ static void grab_commit_values(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
+ {
+       int i;
+       struct commit *commit = (struct commit *) obj;
+       for (i = 0; i < used_atom_cnt; i++) {
+               const char *name = used_atom[i];
+               struct atom_value *v = &val[i];
+               if (!!deref != (*name == '*'))
+                       continue;
+               if (deref)
+                       name++;
+               if (!strcmp(name, "tree")) {
+                       char *s = xmalloc(41);
+                       strcpy(s, sha1_to_hex(commit->tree->object.sha1));
+                       v->s = s;
+               }
+               if (!strcmp(name, "numparent")) {
+                       char *s = xmalloc(40);
+                       v->ul = commit_list_count(commit->parents);
+                       sprintf(s, "%lu", v->ul);
+                       v->s = s;
+               }
+               else if (!strcmp(name, "parent")) {
+                       int num = commit_list_count(commit->parents);
+                       int i;
+                       struct commit_list *parents;
+                       char *s = xmalloc(41 * num + 1);
+                       v->s = s;
+                       for (i = 0, parents = commit->parents;
+                            parents;
+                            parents = parents->next, i = i + 41) {
+                               struct commit *parent = parents->item;
+                               strcpy(s+i, sha1_to_hex(parent->object.sha1));
+                               if (parents->next)
+                                       s[i+40] = ' ';
+                       }
+                       if (!i)
+                               *s = '\0';
+               }
+       }
+ }
+ static const char *find_wholine(const char *who, int wholen, const char *buf, unsigned long sz)
+ {
+       const char *eol;
+       while (*buf) {
+               if (!strncmp(buf, who, wholen) &&
+                   buf[wholen] == ' ')
+                       return buf + wholen + 1;
+               eol = strchr(buf, '\n');
+               if (!eol)
+                       return "";
+               eol++;
+               if (*eol == '\n')
+                       return ""; /* end of header */
+               buf = eol;
+       }
+       return "";
+ }
+ static const char *copy_line(const char *buf)
+ {
+       const char *eol = strchrnul(buf, '\n');
+       return xmemdupz(buf, eol - buf);
+ }
+ static const char *copy_name(const char *buf)
+ {
+       const char *cp;
+       for (cp = buf; *cp && *cp != '\n'; cp++) {
+               if (!strncmp(cp, " <", 2))
+                       return xmemdupz(buf, cp - buf);
+       }
+       return "";
+ }
+ static const char *copy_email(const char *buf)
+ {
+       const char *email = strchr(buf, '<');
+       const char *eoemail;
+       if (!email)
+               return "";
+       eoemail = strchr(email, '>');
+       if (!eoemail)
+               return "";
+       return xmemdupz(email, eoemail + 1 - email);
+ }
+ static char *copy_subject(const char *buf, unsigned long len)
+ {
+       char *r = xmemdupz(buf, len);
+       int i;
+       for (i = 0; i < len; i++)
+               if (r[i] == '\n')
+                       r[i] = ' ';
+       return r;
+ }
+ static void grab_date(const char *buf, struct atom_value *v, const char *atomname)
+ {
+       const char *eoemail = strstr(buf, "> ");
+       char *zone;
+       unsigned long timestamp;
+       long tz;
+       enum date_mode date_mode = DATE_NORMAL;
+       const char *formatp;
+       /*
+        * We got here because atomname ends in "date" or "date<something>";
+        * it's not possible that <something> is not ":<format>" because
+        * parse_ref_filter_atom() wouldn't have allowed it, so we can assume that no
+        * ":" means no format is specified, and use the default.
+        */
+       formatp = strchr(atomname, ':');
+       if (formatp != NULL) {
+               formatp++;
+               date_mode = parse_date_format(formatp);
+       }
+       if (!eoemail)
+               goto bad;
+       timestamp = strtoul(eoemail + 2, &zone, 10);
+       if (timestamp == ULONG_MAX)
+               goto bad;
+       tz = strtol(zone, NULL, 10);
+       if ((tz == LONG_MIN || tz == LONG_MAX) && errno == ERANGE)
+               goto bad;
+       v->s = xstrdup(show_date(timestamp, tz, date_mode));
+       v->ul = timestamp;
+       return;
+  bad:
+       v->s = "";
+       v->ul = 0;
+ }
+ /* See grab_values */
+ static void grab_person(const char *who, struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
+ {
+       int i;
+       int wholen = strlen(who);
+       const char *wholine = NULL;
+       for (i = 0; i < used_atom_cnt; i++) {
+               const char *name = used_atom[i];
+               struct atom_value *v = &val[i];
+               if (!!deref != (*name == '*'))
+                       continue;
+               if (deref)
+                       name++;
+               if (strncmp(who, name, wholen))
+                       continue;
+               if (name[wholen] != 0 &&
+                   strcmp(name + wholen, "name") &&
+                   strcmp(name + wholen, "email") &&
+                   !starts_with(name + wholen, "date"))
+                       continue;
+               if (!wholine)
+                       wholine = find_wholine(who, wholen, buf, sz);
+               if (!wholine)
+                       return; /* no point looking for it */
+               if (name[wholen] == 0)
+                       v->s = copy_line(wholine);
+               else if (!strcmp(name + wholen, "name"))
+                       v->s = copy_name(wholine);
+               else if (!strcmp(name + wholen, "email"))
+                       v->s = copy_email(wholine);
+               else if (starts_with(name + wholen, "date"))
+                       grab_date(wholine, v, name);
+       }
+       /*
+        * For a tag or a commit object, if "creator" or "creatordate" is
+        * requested, do something special.
+        */
+       if (strcmp(who, "tagger") && strcmp(who, "committer"))
+               return; /* "author" for commit object is not wanted */
+       if (!wholine)
+               wholine = find_wholine(who, wholen, buf, sz);
+       if (!wholine)
+               return;
+       for (i = 0; i < used_atom_cnt; i++) {
+               const char *name = used_atom[i];
+               struct atom_value *v = &val[i];
+               if (!!deref != (*name == '*'))
+                       continue;
+               if (deref)
+                       name++;
+               if (starts_with(name, "creatordate"))
+                       grab_date(wholine, v, name);
+               else if (!strcmp(name, "creator"))
+                       v->s = copy_line(wholine);
+       }
+ }
+ static void find_subpos(const char *buf, unsigned long sz,
+                       const char **sub, unsigned long *sublen,
+                       const char **body, unsigned long *bodylen,
+                       unsigned long *nonsiglen,
+                       const char **sig, unsigned long *siglen)
+ {
+       const char *eol;
+       /* skip past header until we hit empty line */
+       while (*buf && *buf != '\n') {
+               eol = strchrnul(buf, '\n');
+               if (*eol)
+                       eol++;
+               buf = eol;
+       }
+       /* skip any empty lines */
+       while (*buf == '\n')
+               buf++;
+       /* parse signature first; we might not even have a subject line */
+       *sig = buf + parse_signature(buf, strlen(buf));
+       *siglen = strlen(*sig);
+       /* subject is first non-empty line */
+       *sub = buf;
+       /* subject goes to first empty line */
+       while (buf < *sig && *buf && *buf != '\n') {
+               eol = strchrnul(buf, '\n');
+               if (*eol)
+                       eol++;
+               buf = eol;
+       }
+       *sublen = buf - *sub;
+       /* drop trailing newline, if present */
+       if (*sublen && (*sub)[*sublen - 1] == '\n')
+               *sublen -= 1;
+       /* skip any empty lines */
+       while (*buf == '\n')
+               buf++;
+       *body = buf;
+       *bodylen = strlen(buf);
+       *nonsiglen = *sig - buf;
+ }
+ /* See grab_values */
+ static void grab_sub_body_contents(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
+ {
+       int i;
+       const char *subpos = NULL, *bodypos = NULL, *sigpos = NULL;
+       unsigned long sublen = 0, bodylen = 0, nonsiglen = 0, siglen = 0;
+       for (i = 0; i < used_atom_cnt; i++) {
+               const char *name = used_atom[i];
+               struct atom_value *v = &val[i];
+               if (!!deref != (*name == '*'))
+                       continue;
+               if (deref)
+                       name++;
+               if (strcmp(name, "subject") &&
+                   strcmp(name, "body") &&
+                   strcmp(name, "contents") &&
+                   strcmp(name, "contents:subject") &&
+                   strcmp(name, "contents:body") &&
+                   strcmp(name, "contents:signature"))
+                       continue;
+               if (!subpos)
+                       find_subpos(buf, sz,
+                                   &subpos, &sublen,
+                                   &bodypos, &bodylen, &nonsiglen,
+                                   &sigpos, &siglen);
+               if (!strcmp(name, "subject"))
+                       v->s = copy_subject(subpos, sublen);
+               else if (!strcmp(name, "contents:subject"))
+                       v->s = copy_subject(subpos, sublen);
+               else if (!strcmp(name, "body"))
+                       v->s = xmemdupz(bodypos, bodylen);
+               else if (!strcmp(name, "contents:body"))
+                       v->s = xmemdupz(bodypos, nonsiglen);
+               else if (!strcmp(name, "contents:signature"))
+                       v->s = xmemdupz(sigpos, siglen);
+               else if (!strcmp(name, "contents"))
+                       v->s = xstrdup(subpos);
+       }
+ }
+ /*
+  * We want to have empty print-string for field requests
+  * that do not apply (e.g. "authordate" for a tag object)
+  */
+ static void fill_missing_values(struct atom_value *val)
+ {
+       int i;
+       for (i = 0; i < used_atom_cnt; i++) {
+               struct atom_value *v = &val[i];
+               if (v->s == NULL)
+                       v->s = "";
+       }
+ }
+ /*
+  * val is a list of atom_value to hold returned values.  Extract
+  * the values for atoms in used_atom array out of (obj, buf, sz).
+  * when deref is false, (obj, buf, sz) is the object that is
+  * pointed at by the ref itself; otherwise it is the object the
+  * ref (which is a tag) refers to.
+  */
+ static void grab_values(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz)
+ {
+       grab_common_values(val, deref, obj, buf, sz);
+       switch (obj->type) {
+       case OBJ_TAG:
+               grab_tag_values(val, deref, obj, buf, sz);
+               grab_sub_body_contents(val, deref, obj, buf, sz);
+               grab_person("tagger", val, deref, obj, buf, sz);
+               break;
+       case OBJ_COMMIT:
+               grab_commit_values(val, deref, obj, buf, sz);
+               grab_sub_body_contents(val, deref, obj, buf, sz);
+               grab_person("author", val, deref, obj, buf, sz);
+               grab_person("committer", val, deref, obj, buf, sz);
+               break;
+       case OBJ_TREE:
+               /* grab_tree_values(val, deref, obj, buf, sz); */
+               break;
+       case OBJ_BLOB:
+               /* grab_blob_values(val, deref, obj, buf, sz); */
+               break;
+       default:
+               die("Eh?  Object of type %d?", obj->type);
+       }
+ }
+ static inline char *copy_advance(char *dst, const char *src)
+ {
+       while (*src)
+               *dst++ = *src++;
+       return dst;
+ }
+ /*
+  * Parse the object referred by ref, and grab needed value.
+  */
+ static void populate_value(struct ref_array_item *ref)
+ {
+       void *buf;
+       struct object *obj;
+       int eaten, i;
+       unsigned long size;
+       const unsigned char *tagged;
+       ref->value = xcalloc(used_atom_cnt, sizeof(struct atom_value));
+       if (need_symref && (ref->flag & REF_ISSYMREF) && !ref->symref) {
+               unsigned char unused1[20];
+               ref->symref = resolve_refdup(ref->refname, RESOLVE_REF_READING,
+                                            unused1, NULL);
+               if (!ref->symref)
+                       ref->symref = "";
+       }
+       /* Fill in specials first */
+       for (i = 0; i < used_atom_cnt; i++) {
+               const char *name = used_atom[i];
+               struct atom_value *v = &ref->value[i];
+               int deref = 0;
+               const char *refname;
+               const char *formatp;
+               struct branch *branch = NULL;
+               if (*name == '*') {
+                       deref = 1;
+                       name++;
+               }
+               if (starts_with(name, "refname"))
+                       refname = ref->refname;
+               else if (starts_with(name, "symref"))
+                       refname = ref->symref ? ref->symref : "";
+               else if (starts_with(name, "upstream")) {
+                       const char *branch_name;
+                       /* only local branches may have an upstream */
+                       if (!skip_prefix(ref->refname, "refs/heads/",
+                                        &branch_name))
+                               continue;
+                       branch = branch_get(branch_name);
+                       refname = branch_get_upstream(branch, NULL);
+                       if (!refname)
+                               continue;
+               } else if (starts_with(name, "push")) {
+                       const char *branch_name;
+                       if (!skip_prefix(ref->refname, "refs/heads/",
+                                        &branch_name))
+                               continue;
+                       branch = branch_get(branch_name);
+                       refname = branch_get_push(branch, NULL);
+                       if (!refname)
+                               continue;
+               } else if (starts_with(name, "color:")) {
+                       char color[COLOR_MAXLEN] = "";
+                       if (color_parse(name + 6, color) < 0)
+                               die(_("unable to parse format"));
+                       v->s = xstrdup(color);
+                       continue;
+               } else if (!strcmp(name, "flag")) {
+                       char buf[256], *cp = buf;
+                       if (ref->flag & REF_ISSYMREF)
+                               cp = copy_advance(cp, ",symref");
+                       if (ref->flag & REF_ISPACKED)
+                               cp = copy_advance(cp, ",packed");
+                       if (cp == buf)
+                               v->s = "";
+                       else {
+                               *cp = '\0';
+                               v->s = xstrdup(buf + 1);
+                       }
+                       continue;
+               } else if (!deref && grab_objectname(name, ref->objectname, v)) {
+                       continue;
+               } else if (!strcmp(name, "HEAD")) {
+                       const char *head;
+                       unsigned char sha1[20];
+                       head = resolve_ref_unsafe("HEAD", RESOLVE_REF_READING,
+                                                 sha1, NULL);
+                       if (!strcmp(ref->refname, head))
+                               v->s = "*";
+                       else
+                               v->s = " ";
+                       continue;
+               } else
+                       continue;
+               formatp = strchr(name, ':');
+               if (formatp) {
+                       int num_ours, num_theirs;
+                       formatp++;
+                       if (!strcmp(formatp, "short"))
+                               refname = shorten_unambiguous_ref(refname,
+                                                     warn_ambiguous_refs);
+                       else if (!strcmp(formatp, "track") &&
+                                (starts_with(name, "upstream") ||
+                                 starts_with(name, "push"))) {
+                               char buf[40];
+                               if (stat_tracking_info(branch, &num_ours,
+                                                      &num_theirs, NULL))
+                                       continue;
+                               if (!num_ours && !num_theirs)
+                                       v->s = "";
+                               else if (!num_ours) {
+                                       sprintf(buf, "[behind %d]", num_theirs);
+                                       v->s = xstrdup(buf);
+                               } else if (!num_theirs) {
+                                       sprintf(buf, "[ahead %d]", num_ours);
+                                       v->s = xstrdup(buf);
+                               } else {
+                                       sprintf(buf, "[ahead %d, behind %d]",
+                                               num_ours, num_theirs);
+                                       v->s = xstrdup(buf);
+                               }
+                               continue;
+                       } else if (!strcmp(formatp, "trackshort") &&
+                                  (starts_with(name, "upstream") ||
+                                   starts_with(name, "push"))) {
+                               assert(branch);
+                               if (stat_tracking_info(branch, &num_ours,
+                                                       &num_theirs, NULL))
+                                       continue;
+                               if (!num_ours && !num_theirs)
+                                       v->s = "=";
+                               else if (!num_ours)
+                                       v->s = "<";
+                               else if (!num_theirs)
+                                       v->s = ">";
+                               else
+                                       v->s = "<>";
+                               continue;
+                       } else
+                               die("unknown %.*s format %s",
+                                   (int)(formatp - name), name, formatp);
+               }
+               if (!deref)
+                       v->s = refname;
+               else {
+                       int len = strlen(refname);
+                       char *s = xmalloc(len + 4);
+                       sprintf(s, "%s^{}", refname);
+                       v->s = s;
+               }
+       }
+       for (i = 0; i < used_atom_cnt; i++) {
+               struct atom_value *v = &ref->value[i];
+               if (v->s == NULL)
+                       goto need_obj;
+       }
+       return;
+  need_obj:
+       buf = get_obj(ref->objectname, &obj, &size, &eaten);
+       if (!buf)
+               die("missing object %s for %s",
+                   sha1_to_hex(ref->objectname), ref->refname);
+       if (!obj)
+               die("parse_object_buffer failed on %s for %s",
+                   sha1_to_hex(ref->objectname), ref->refname);
+       grab_values(ref->value, 0, obj, buf, size);
+       if (!eaten)
+               free(buf);
+       /*
+        * If there is no atom that wants to know about tagged
+        * object, we are done.
+        */
+       if (!need_tagged || (obj->type != OBJ_TAG))
+               return;
+       /*
+        * If it is a tag object, see if we use a value that derefs
+        * the object, and if we do grab the object it refers to.
+        */
+       tagged = ((struct tag *)obj)->tagged->sha1;
+       /*
+        * NEEDSWORK: This derefs tag only once, which
+        * is good to deal with chains of trust, but
+        * is not consistent with what deref_tag() does
+        * which peels the onion to the core.
+        */
+       buf = get_obj(tagged, &obj, &size, &eaten);
+       if (!buf)
+               die("missing object %s for %s",
+                   sha1_to_hex(tagged), ref->refname);
+       if (!obj)
+               die("parse_object_buffer failed on %s for %s",
+                   sha1_to_hex(tagged), ref->refname);
+       grab_values(ref->value, 1, obj, buf, size);
+       if (!eaten)
+               free(buf);
+ }
+ /*
+  * Given a ref, return the value for the atom.  This lazily gets value
+  * out of the object by calling populate value.
+  */
+ static void get_ref_atom_value(struct ref_array_item *ref, int atom, struct atom_value **v)
+ {
+       if (!ref->value) {
+               populate_value(ref);
+               fill_missing_values(ref->value);
+       }
+       *v = &ref->value[atom];
+ }
+ /*
+  * Return 1 if the refname matches one of the patterns, otherwise 0.
+  * A pattern can be path prefix (e.g. a refname "refs/heads/master"
+  * matches a pattern "refs/heads/") or a wildcard (e.g. the same ref
+  * matches "refs/heads/m*",too).
+  */
+ static int match_name_as_path(const char **pattern, const char *refname)
+ {
+       int namelen = strlen(refname);
+       for (; *pattern; pattern++) {
+               const char *p = *pattern;
+               int plen = strlen(p);
+               if ((plen <= namelen) &&
+                   !strncmp(refname, p, plen) &&
+                   (refname[plen] == '\0' ||
+                    refname[plen] == '/' ||
+                    p[plen-1] == '/'))
+                       return 1;
+               if (!wildmatch(p, refname, WM_PATHNAME, NULL))
+                       return 1;
+       }
+       return 0;
+ }
+ /* Allocate space for a new ref_array_item and copy the objectname and flag to it */
+ static struct ref_array_item *new_ref_array_item(const char *refname,
+                                                const unsigned char *objectname,
+                                                int flag)
+ {
+       size_t len = strlen(refname);
+       struct ref_array_item *ref = xcalloc(1, sizeof(struct ref_array_item) + len + 1);
+       memcpy(ref->refname, refname, len);
+       ref->refname[len] = '\0';
+       hashcpy(ref->objectname, objectname);
+       ref->flag = flag;
+       return ref;
+ }
+ /*
+  * A call-back given to for_each_ref().  Filter refs and keep them for
+  * later object processing.
+  */
+ static int ref_filter_handler(const char *refname, const struct object_id *oid, int flag, void *cb_data)
+ {
+       struct ref_filter_cbdata *ref_cbdata = cb_data;
+       struct ref_filter *filter = ref_cbdata->filter;
+       struct ref_array_item *ref;
+       if (flag & REF_BAD_NAME) {
+               warning("ignoring ref with broken name %s", refname);
+               return 0;
+       }
++      if (flag & REF_ISBROKEN) {
++              warning("ignoring broken ref %s", refname);
++              return 0;
++      }
++
+       if (*filter->name_patterns && !match_name_as_path(filter->name_patterns, refname))
+               return 0;
+       /*
+        * We do not open the object yet; sort may only need refname
+        * to do its job and the resulting list may yet to be pruned
+        * by maxcount logic.
+        */
+       ref = new_ref_array_item(refname, oid->hash, flag);
+       REALLOC_ARRAY(ref_cbdata->array->items, ref_cbdata->array->nr + 1);
+       ref_cbdata->array->items[ref_cbdata->array->nr++] = ref;
+       return 0;
+ }
+ /*  Free memory allocated for a ref_array_item */
+ static void free_array_item(struct ref_array_item *item)
+ {
+       free((char *)item->symref);
+       free(item);
+ }
+ /* Free all memory allocated for ref_array */
+ void ref_array_clear(struct ref_array *array)
+ {
+       int i;
+       for (i = 0; i < array->nr; i++)
+               free_array_item(array->items[i]);
+       free(array->items);
+       array->items = NULL;
+       array->nr = array->alloc = 0;
+ }
+ /*
+  * API for filtering a set of refs. Based on the type of refs the user
+  * has requested, we iterate through those refs and apply filters
+  * as per the given ref_filter structure and finally store the
+  * filtered refs in the ref_array structure.
+  */
+ int filter_refs(struct ref_array *array, struct ref_filter *filter, unsigned int type)
+ {
+       struct ref_filter_cbdata ref_cbdata;
+       ref_cbdata.array = array;
+       ref_cbdata.filter = filter;
+       if (type & (FILTER_REFS_ALL | FILTER_REFS_INCLUDE_BROKEN))
+               return for_each_rawref(ref_filter_handler, &ref_cbdata);
+       else if (type & FILTER_REFS_ALL)
+               return for_each_ref(ref_filter_handler, &ref_cbdata);
+       else
+               die("filter_refs: invalid type");
+       return 0;
+ }
+ static int cmp_ref_sorting(struct ref_sorting *s, struct ref_array_item *a, struct ref_array_item *b)
+ {
+       struct atom_value *va, *vb;
+       int cmp;
+       cmp_type cmp_type = used_atom_type[s->atom];
+       get_ref_atom_value(a, s->atom, &va);
+       get_ref_atom_value(b, s->atom, &vb);
+       switch (cmp_type) {
+       case FIELD_STR:
+               cmp = strcmp(va->s, vb->s);
+               break;
+       default:
+               if (va->ul < vb->ul)
+                       cmp = -1;
+               else if (va->ul == vb->ul)
+                       cmp = 0;
+               else
+                       cmp = 1;
+               break;
+       }
+       return (s->reverse) ? -cmp : cmp;
+ }
+ static struct ref_sorting *ref_sorting;
+ static int compare_refs(const void *a_, const void *b_)
+ {
+       struct ref_array_item *a = *((struct ref_array_item **)a_);
+       struct ref_array_item *b = *((struct ref_array_item **)b_);
+       struct ref_sorting *s;
+       for (s = ref_sorting; s; s = s->next) {
+               int cmp = cmp_ref_sorting(s, a, b);
+               if (cmp)
+                       return cmp;
+       }
+       return 0;
+ }
+ void ref_array_sort(struct ref_sorting *sorting, struct ref_array *array)
+ {
+       ref_sorting = sorting;
+       qsort(array->items, array->nr, sizeof(struct ref_array_item *), compare_refs);
+ }
+ static void print_value(struct atom_value *v, int quote_style)
+ {
+       struct strbuf sb = STRBUF_INIT;
+       switch (quote_style) {
+       case QUOTE_NONE:
+               fputs(v->s, stdout);
+               break;
+       case QUOTE_SHELL:
+               sq_quote_buf(&sb, v->s);
+               break;
+       case QUOTE_PERL:
+               perl_quote_buf(&sb, v->s);
+               break;
+       case QUOTE_PYTHON:
+               python_quote_buf(&sb, v->s);
+               break;
+       case QUOTE_TCL:
+               tcl_quote_buf(&sb, v->s);
+               break;
+       }
+       if (quote_style != QUOTE_NONE) {
+               fputs(sb.buf, stdout);
+               strbuf_release(&sb);
+       }
+ }
+ static int hex1(char ch)
+ {
+       if ('0' <= ch && ch <= '9')
+               return ch - '0';
+       else if ('a' <= ch && ch <= 'f')
+               return ch - 'a' + 10;
+       else if ('A' <= ch && ch <= 'F')
+               return ch - 'A' + 10;
+       return -1;
+ }
+ static int hex2(const char *cp)
+ {
+       if (cp[0] && cp[1])
+               return (hex1(cp[0]) << 4) | hex1(cp[1]);
+       else
+               return -1;
+ }
+ static void emit(const char *cp, const char *ep)
+ {
+       while (*cp && (!ep || cp < ep)) {
+               if (*cp == '%') {
+                       if (cp[1] == '%')
+                               cp++;
+                       else {
+                               int ch = hex2(cp + 1);
+                               if (0 <= ch) {
+                                       putchar(ch);
+                                       cp += 3;
+                                       continue;
+                               }
+                       }
+               }
+               putchar(*cp);
+               cp++;
+       }
+ }
+ void show_ref_array_item(struct ref_array_item *info, const char *format, int quote_style)
+ {
+       const char *cp, *sp, *ep;
+       for (cp = format; *cp && (sp = find_next(cp)); cp = ep + 1) {
+               struct atom_value *atomv;
+               ep = strchr(sp, ')');
+               if (cp < sp)
+                       emit(cp, sp);
+               get_ref_atom_value(info, parse_ref_filter_atom(sp + 2, ep), &atomv);
+               print_value(atomv, quote_style);
+       }
+       if (*cp) {
+               sp = cp + strlen(cp);
+               emit(cp, sp);
+       }
+       if (need_color_reset_at_eol) {
+               struct atom_value resetv;
+               char color[COLOR_MAXLEN] = "";
+               if (color_parse("reset", color) < 0)
+                       die("BUG: couldn't parse 'reset' as a color");
+               resetv.s = color;
+               print_value(&resetv, quote_style);
+       }
+       putchar('\n');
+ }
+ /*  If no sorting option is given, use refname to sort as default */
+ struct ref_sorting *ref_default_sorting(void)
+ {
+       static const char cstr_name[] = "refname";
+       struct ref_sorting *sorting = xcalloc(1, sizeof(*sorting));
+       sorting->next = NULL;
+       sorting->atom = parse_ref_filter_atom(cstr_name, cstr_name + strlen(cstr_name));
+       return sorting;
+ }
+ int parse_opt_ref_sorting(const struct option *opt, const char *arg, int unset)
+ {
+       struct ref_sorting **sorting_tail = opt->value;
+       struct ref_sorting *s;
+       int len;
+       if (!arg) /* should --no-sort void the list ? */
+               return -1;
+       s = xcalloc(1, sizeof(*s));
+       s->next = *sorting_tail;
+       *sorting_tail = s;
+       if (*arg == '-') {
+               s->reverse = 1;
+               arg++;
+       }
+       len = strlen(arg);
+       s->atom = parse_ref_filter_atom(arg, arg+len);
+       return 0;
+ }