Merge git://repo.or.cz/git-gui
[gitweb.git] / commit.c
index 1fbdd2d51b8e798ac73c4b020f8fa47b1d6fb168..8b8fb04d1f6107da15cb142485d80cabe7a3828e 100644 (file)
--- a/commit.c
+++ b/commit.c
@@ -3,70 +3,13 @@
 #include "commit.h"
 #include "pkt-line.h"
 #include "utf8.h"
-#include "interpolate.h"
 #include "diff.h"
 #include "revision.h"
 
 int save_commit_buffer = 1;
 
-struct sort_node
-{
-       /*
-        * the number of children of the associated commit
-        * that also occur in the list being sorted.
-        */
-       unsigned int indegree;
-
-       /*
-        * reference to original list item that we will re-use
-        * on output.
-        */
-       struct commit_list * list_item;
-
-};
-
 const char *commit_type = "commit";
 
-static struct cmt_fmt_map {
-       const char *n;
-       size_t cmp_len;
-       enum cmit_fmt v;
-} cmt_fmts[] = {
-       { "raw",        1,      CMIT_FMT_RAW },
-       { "medium",     1,      CMIT_FMT_MEDIUM },
-       { "short",      1,      CMIT_FMT_SHORT },
-       { "email",      1,      CMIT_FMT_EMAIL },
-       { "full",       5,      CMIT_FMT_FULL },
-       { "fuller",     5,      CMIT_FMT_FULLER },
-       { "oneline",    1,      CMIT_FMT_ONELINE },
-       { "format:",    7,      CMIT_FMT_USERFORMAT},
-};
-
-static char *user_format;
-
-enum cmit_fmt get_commit_format(const char *arg)
-{
-       int i;
-
-       if (!arg || !*arg)
-               return CMIT_FMT_DEFAULT;
-       if (*arg == '=')
-               arg++;
-       if (!prefixcmp(arg, "format:")) {
-               if (user_format)
-                       free(user_format);
-               user_format = xstrdup(arg + 7);
-               return CMIT_FMT_USERFORMAT;
-       }
-       for (i = 0; i < ARRAY_SIZE(cmt_fmts); i++) {
-               if (!strncmp(arg, cmt_fmts[i].n, cmt_fmts[i].cmp_len) &&
-                   !strncmp(arg, cmt_fmts[i].n, strlen(arg)))
-                       return cmt_fmts[i].v;
-       }
-
-       die("invalid --pretty format: %s", arg);
-}
-
 static struct commit *check_commit(struct object *obj,
                                   const unsigned char *sha1,
                                   int quiet)
@@ -105,19 +48,32 @@ struct commit *lookup_commit(const unsigned char *sha1)
        return check_commit(obj, sha1, 0);
 }
 
-static unsigned long parse_commit_date(const char *buf)
+static unsigned long parse_commit_date(const char *buf, const char *tail)
 {
        unsigned long date;
+       const char *dateptr;
 
+       if (buf + 6 >= tail)
+               return 0;
        if (memcmp(buf, "author", 6))
                return 0;
-       while (*buf++ != '\n')
+       while (buf < tail && *buf++ != '\n')
                /* nada */;
+       if (buf + 9 >= tail)
+               return 0;
        if (memcmp(buf, "committer", 9))
                return 0;
-       while (*buf++ != '>')
+       while (buf < tail && *buf++ != '>')
                /* nada */;
-       date = strtoul(buf, NULL, 10);
+       if (buf >= tail)
+               return 0;
+       dateptr = buf;
+       while (buf < tail && *buf++ != '\n')
+               /* nada */;
+       if (buf >= tail)
+               return 0;
+       /* dateptr < buf && buf[-1] == '\n', so strtoul will stop at buf-1 */
+       date = strtoul(dateptr, NULL, 10);
        if (date == ULONG_MAX)
                date = 0;
        return date;
@@ -293,9 +249,9 @@ int parse_commit_buffer(struct commit *item, void *buffer, unsigned long size)
                return 0;
        item->object.parsed = 1;
        tail += size;
-       if (tail <= bufptr + 5 || memcmp(bufptr, "tree ", 5))
+       if (tail <= bufptr + 46 || memcmp(bufptr, "tree ", 5) || bufptr[45] != '\n')
                return error("bogus commit object %s", sha1_to_hex(item->object.sha1));
-       if (tail <= bufptr + 45 || get_sha1_hex(bufptr + 5, parent) < 0)
+       if (get_sha1_hex(bufptr + 5, parent) < 0)
                return error("bad tree pointer in commit %s",
                             sha1_to_hex(item->object.sha1));
        item->tree = lookup_tree(parent);
@@ -332,7 +288,7 @@ int parse_commit_buffer(struct commit *item, void *buffer, unsigned long size)
                        n_refs++;
                }
        }
-       item->date = parse_commit_date(bufptr);
+       item->date = parse_commit_date(bufptr, tail);
 
        if (track_object_refs) {
                unsigned i = 0;
@@ -460,813 +416,6 @@ void clear_commit_marks(struct commit *commit, unsigned int mark)
        }
 }
 
-/*
- * Generic support for pretty-printing the header
- */
-static int get_one_line(const char *msg, unsigned long len)
-{
-       int ret = 0;
-
-       while (len--) {
-               char c = *msg++;
-               if (!c)
-                       break;
-               ret++;
-               if (c == '\n')
-                       break;
-       }
-       return ret;
-}
-
-/* High bit set, or ISO-2022-INT */
-static int non_ascii(int ch)
-{
-       ch = (ch & 0xff);
-       return ((ch & 0x80) || (ch == 0x1b));
-}
-
-static int is_rfc2047_special(char ch)
-{
-       return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
-}
-
-static int add_rfc2047(char *buf, const char *line, int len,
-                      const char *encoding)
-{
-       char *bp = buf;
-       int i, needquote;
-       char q_encoding[128];
-       const char *q_encoding_fmt = "=?%s?q?";
-
-       for (i = needquote = 0; !needquote && i < len; i++) {
-               int ch = line[i];
-               if (non_ascii(ch))
-                       needquote++;
-               if ((i + 1 < len) &&
-                   (ch == '=' && line[i+1] == '?'))
-                       needquote++;
-       }
-       if (!needquote)
-               return sprintf(buf, "%.*s", len, line);
-
-       i = snprintf(q_encoding, sizeof(q_encoding), q_encoding_fmt, encoding);
-       if (sizeof(q_encoding) < i)
-               die("Insanely long encoding name %s", encoding);
-       memcpy(bp, q_encoding, i);
-       bp += i;
-       for (i = 0; i < len; i++) {
-               unsigned ch = line[i] & 0xFF;
-               /*
-                * We encode ' ' using '=20' even though rfc2047
-                * allows using '_' for readability.  Unfortunately,
-                * many programs do not understand this and just
-                * leave the underscore in place.
-                */
-               if (is_rfc2047_special(ch) || ch == ' ') {
-                       sprintf(bp, "=%02X", ch);
-                       bp += 3;
-               }
-               else
-                       *bp++ = ch;
-       }
-       memcpy(bp, "?=", 2);
-       bp += 2;
-       return bp - buf;
-}
-
-static unsigned long bound_rfc2047(unsigned long len, const char *encoding)
-{
-       /* upper bound of q encoded string of length 'len' */
-       unsigned long elen = strlen(encoding);
-
-       return len * 3 + elen + 100;
-}
-
-static int add_user_info(const char *what, enum cmit_fmt fmt, char *buf,
-                        const char *line, enum date_mode dmode,
-                        const char *encoding)
-{
-       char *date;
-       int namelen;
-       unsigned long time;
-       int tz, ret;
-       const char *filler = "    ";
-
-       if (fmt == CMIT_FMT_ONELINE)
-               return 0;
-       date = strchr(line, '>');
-       if (!date)
-               return 0;
-       namelen = ++date - line;
-       time = strtoul(date, &date, 10);
-       tz = strtol(date, NULL, 10);
-
-       if (fmt == CMIT_FMT_EMAIL) {
-               char *name_tail = strchr(line, '<');
-               int display_name_length;
-               if (!name_tail)
-                       return 0;
-               while (line < name_tail && isspace(name_tail[-1]))
-                       name_tail--;
-               display_name_length = name_tail - line;
-               filler = "";
-               strcpy(buf, "From: ");
-               ret = strlen(buf);
-               ret += add_rfc2047(buf + ret, line, display_name_length,
-                                  encoding);
-               memcpy(buf + ret, name_tail, namelen - display_name_length);
-               ret += namelen - display_name_length;
-               buf[ret++] = '\n';
-       }
-       else {
-               ret = sprintf(buf, "%s: %.*s%.*s\n", what,
-                             (fmt == CMIT_FMT_FULLER) ? 4 : 0,
-                             filler, namelen, line);
-       }
-       switch (fmt) {
-       case CMIT_FMT_MEDIUM:
-               ret += sprintf(buf + ret, "Date:   %s\n",
-                              show_date(time, tz, dmode));
-               break;
-       case CMIT_FMT_EMAIL:
-               ret += sprintf(buf + ret, "Date: %s\n",
-                              show_date(time, tz, DATE_RFC2822));
-               break;
-       case CMIT_FMT_FULLER:
-               ret += sprintf(buf + ret, "%sDate: %s\n", what,
-                              show_date(time, tz, dmode));
-               break;
-       default:
-               /* notin' */
-               break;
-       }
-       return ret;
-}
-
-static int is_empty_line(const char *line, int *len_p)
-{
-       int len = *len_p;
-       while (len && isspace(line[len-1]))
-               len--;
-       *len_p = len;
-       return !len;
-}
-
-static int add_merge_info(enum cmit_fmt fmt, char *buf, const struct commit *commit, int abbrev)
-{
-       struct commit_list *parent = commit->parents;
-       int offset;
-
-       if ((fmt == CMIT_FMT_ONELINE) || (fmt == CMIT_FMT_EMAIL) ||
-           !parent || !parent->next)
-               return 0;
-
-       offset = sprintf(buf, "Merge:");
-
-       while (parent) {
-               struct commit *p = parent->item;
-               const char *hex = NULL;
-               const char *dots;
-               if (abbrev)
-                       hex = find_unique_abbrev(p->object.sha1, abbrev);
-               if (!hex)
-                       hex = sha1_to_hex(p->object.sha1);
-               dots = (abbrev && strlen(hex) != 40) ?  "..." : "";
-               parent = parent->next;
-
-               offset += sprintf(buf + offset, " %s%s", hex, dots);
-       }
-       buf[offset++] = '\n';
-       return offset;
-}
-
-static char *get_header(const struct commit *commit, const char *key)
-{
-       int key_len = strlen(key);
-       const char *line = commit->buffer;
-
-       for (;;) {
-               const char *eol = strchr(line, '\n'), *next;
-
-               if (line == eol)
-                       return NULL;
-               if (!eol) {
-                       eol = line + strlen(line);
-                       next = NULL;
-               } else
-                       next = eol + 1;
-               if (eol - line > key_len &&
-                   !strncmp(line, key, key_len) &&
-                   line[key_len] == ' ') {
-                       int len = eol - line - key_len;
-                       char *ret = xmalloc(len);
-                       memcpy(ret, line + key_len + 1, len - 1);
-                       ret[len - 1] = '\0';
-                       return ret;
-               }
-               line = next;
-       }
-}
-
-static char *replace_encoding_header(char *buf, const char *encoding)
-{
-       char *encoding_header = strstr(buf, "\nencoding ");
-       char *header_end = strstr(buf, "\n\n");
-       char *end_of_encoding_header;
-       int encoding_header_pos;
-       int encoding_header_len;
-       int new_len;
-       int need_len;
-       int buflen = strlen(buf) + 1;
-
-       if (!header_end)
-               header_end = buf + buflen;
-       if (!encoding_header || encoding_header >= header_end)
-               return buf;
-       encoding_header++;
-       end_of_encoding_header = strchr(encoding_header, '\n');
-       if (!end_of_encoding_header)
-               return buf; /* should not happen but be defensive */
-       end_of_encoding_header++;
-
-       encoding_header_len = end_of_encoding_header - encoding_header;
-       encoding_header_pos = encoding_header - buf;
-
-       if (is_encoding_utf8(encoding)) {
-               /* we have re-coded to UTF-8; drop the header */
-               memmove(encoding_header, end_of_encoding_header,
-                       buflen - (encoding_header_pos + encoding_header_len));
-               return buf;
-       }
-       new_len = strlen(encoding);
-       need_len = new_len + strlen("encoding \n");
-       if (encoding_header_len < need_len) {
-               buf = xrealloc(buf, buflen + (need_len - encoding_header_len));
-               encoding_header = buf + encoding_header_pos;
-               end_of_encoding_header = encoding_header + encoding_header_len;
-       }
-       memmove(end_of_encoding_header + (need_len - encoding_header_len),
-               end_of_encoding_header,
-               buflen - (encoding_header_pos + encoding_header_len));
-       memcpy(encoding_header + 9, encoding, strlen(encoding));
-       encoding_header[9 + new_len] = '\n';
-       return buf;
-}
-
-static char *logmsg_reencode(const struct commit *commit,
-                            const char *output_encoding)
-{
-       static const char *utf8 = "utf-8";
-       const char *use_encoding;
-       char *encoding;
-       char *out;
-
-       if (!*output_encoding)
-               return NULL;
-       encoding = get_header(commit, "encoding");
-       use_encoding = encoding ? encoding : utf8;
-       if (!strcmp(use_encoding, output_encoding))
-               if (encoding) /* we'll strip encoding header later */
-                       out = xstrdup(commit->buffer);
-               else
-                       return NULL; /* nothing to do */
-       else
-               out = reencode_string(commit->buffer,
-                                     output_encoding, use_encoding);
-       if (out)
-               out = replace_encoding_header(out, output_encoding);
-
-       free(encoding);
-       return out;
-}
-
-static void fill_person(struct interp *table, const char *msg, int len)
-{
-       int start, end, tz = 0;
-       unsigned long date;
-       char *ep;
-
-       /* parse name */
-       for (end = 0; end < len && msg[end] != '<'; end++)
-               ; /* do nothing */
-       start = end + 1;
-       while (end > 0 && isspace(msg[end - 1]))
-               end--;
-       table[0].value = xstrndup(msg, end);
-
-       if (start >= len)
-               return;
-
-       /* parse email */
-       for (end = start + 1; end < len && msg[end] != '>'; end++)
-               ; /* do nothing */
-
-       if (end >= len)
-               return;
-
-       table[1].value = xstrndup(msg + start, end - start);
-
-       /* parse date */
-       for (start = end + 1; start < len && isspace(msg[start]); start++)
-               ; /* do nothing */
-       if (start >= len)
-               return;
-       date = strtoul(msg + start, &ep, 10);
-       if (msg + start == ep)
-               return;
-
-       table[5].value = xstrndup(msg + start, ep - (msg + start));
-
-       /* parse tz */
-       for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
-               ; /* do nothing */
-       if (start + 1 < len) {
-               tz = strtoul(msg + start + 1, NULL, 10);
-               if (msg[start] == '-')
-                       tz = -tz;
-       }
-
-       interp_set_entry(table, 2, show_date(date, tz, DATE_NORMAL));
-       interp_set_entry(table, 3, show_date(date, tz, DATE_RFC2822));
-       interp_set_entry(table, 4, show_date(date, tz, DATE_RELATIVE));
-       interp_set_entry(table, 6, show_date(date, tz, DATE_ISO8601));
-}
-
-static long format_commit_message(const struct commit *commit,
-               const char *msg, char **buf_p, unsigned long *space_p)
-{
-       struct interp table[] = {
-               { "%H" },       /* commit hash */
-               { "%h" },       /* abbreviated commit hash */
-               { "%T" },       /* tree hash */
-               { "%t" },       /* abbreviated tree hash */
-               { "%P" },       /* parent hashes */
-               { "%p" },       /* abbreviated parent hashes */
-               { "%an" },      /* author name */
-               { "%ae" },      /* author email */
-               { "%ad" },      /* author date */
-               { "%aD" },      /* author date, RFC2822 style */
-               { "%ar" },      /* author date, relative */
-               { "%at" },      /* author date, UNIX timestamp */
-               { "%ai" },      /* author date, ISO 8601 */
-               { "%cn" },      /* committer name */
-               { "%ce" },      /* committer email */
-               { "%cd" },      /* committer date */
-               { "%cD" },      /* committer date, RFC2822 style */
-               { "%cr" },      /* committer date, relative */
-               { "%ct" },      /* committer date, UNIX timestamp */
-               { "%ci" },      /* committer date, ISO 8601 */
-               { "%e" },       /* encoding */
-               { "%s" },       /* subject */
-               { "%b" },       /* body */
-               { "%Cred" },    /* red */
-               { "%Cgreen" },  /* green */
-               { "%Cblue" },   /* blue */
-               { "%Creset" },  /* reset color */
-               { "%n" },       /* newline */
-               { "%m" },       /* left/right/bottom */
-       };
-       enum interp_index {
-               IHASH = 0, IHASH_ABBREV,
-               ITREE, ITREE_ABBREV,
-               IPARENTS, IPARENTS_ABBREV,
-               IAUTHOR_NAME, IAUTHOR_EMAIL,
-               IAUTHOR_DATE, IAUTHOR_DATE_RFC2822, IAUTHOR_DATE_RELATIVE,
-               IAUTHOR_TIMESTAMP, IAUTHOR_ISO8601,
-               ICOMMITTER_NAME, ICOMMITTER_EMAIL,
-               ICOMMITTER_DATE, ICOMMITTER_DATE_RFC2822,
-               ICOMMITTER_DATE_RELATIVE, ICOMMITTER_TIMESTAMP,
-               ICOMMITTER_ISO8601,
-               IENCODING,
-               ISUBJECT,
-               IBODY,
-               IRED, IGREEN, IBLUE, IRESET_COLOR,
-               INEWLINE,
-               ILEFT_RIGHT,
-       };
-       struct commit_list *p;
-       char parents[1024];
-       int i;
-       enum { HEADER, SUBJECT, BODY } state;
-
-       if (ILEFT_RIGHT + 1 != ARRAY_SIZE(table))
-               die("invalid interp table!");
-
-       /* these are independent of the commit */
-       interp_set_entry(table, IRED, "\033[31m");
-       interp_set_entry(table, IGREEN, "\033[32m");
-       interp_set_entry(table, IBLUE, "\033[34m");
-       interp_set_entry(table, IRESET_COLOR, "\033[m");
-       interp_set_entry(table, INEWLINE, "\n");
-
-       /* these depend on the commit */
-       if (!commit->object.parsed)
-               parse_object(commit->object.sha1);
-       interp_set_entry(table, IHASH, sha1_to_hex(commit->object.sha1));
-       interp_set_entry(table, IHASH_ABBREV,
-                       find_unique_abbrev(commit->object.sha1,
-                               DEFAULT_ABBREV));
-       interp_set_entry(table, ITREE, sha1_to_hex(commit->tree->object.sha1));
-       interp_set_entry(table, ITREE_ABBREV,
-                       find_unique_abbrev(commit->tree->object.sha1,
-                               DEFAULT_ABBREV));
-       interp_set_entry(table, ILEFT_RIGHT,
-                        (commit->object.flags & BOUNDARY)
-                        ? "-"
-                        : (commit->object.flags & SYMMETRIC_LEFT)
-                        ? "<"
-                        : ">");
-
-       parents[1] = 0;
-       for (i = 0, p = commit->parents;
-                       p && i < sizeof(parents) - 1;
-                       p = p->next)
-               i += snprintf(parents + i, sizeof(parents) - i - 1, " %s",
-                       sha1_to_hex(p->item->object.sha1));
-       interp_set_entry(table, IPARENTS, parents + 1);
-
-       parents[1] = 0;
-       for (i = 0, p = commit->parents;
-                       p && i < sizeof(parents) - 1;
-                       p = p->next)
-               i += snprintf(parents + i, sizeof(parents) - i - 1, " %s",
-                       find_unique_abbrev(p->item->object.sha1,
-                               DEFAULT_ABBREV));
-       interp_set_entry(table, IPARENTS_ABBREV, parents + 1);
-
-       for (i = 0, state = HEADER; msg[i] && state < BODY; i++) {
-               int eol;
-               for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
-                       ; /* do nothing */
-
-               if (state == SUBJECT) {
-                       table[ISUBJECT].value = xstrndup(msg + i, eol - i);
-                       i = eol;
-               }
-               if (i == eol) {
-                       state++;
-                       /* strip empty lines */
-                       while (msg[eol + 1] == '\n')
-                               eol++;
-               } else if (!prefixcmp(msg + i, "author "))
-                       fill_person(table + IAUTHOR_NAME,
-                                       msg + i + 7, eol - i - 7);
-               else if (!prefixcmp(msg + i, "committer "))
-                       fill_person(table + ICOMMITTER_NAME,
-                                       msg + i + 10, eol - i - 10);
-               else if (!prefixcmp(msg + i, "encoding "))
-                       table[IENCODING].value =
-                               xstrndup(msg + i + 9, eol - i - 9);
-               i = eol;
-       }
-       if (msg[i])
-               table[IBODY].value = xstrdup(msg + i);
-       for (i = 0; i < ARRAY_SIZE(table); i++)
-               if (!table[i].value)
-                       interp_set_entry(table, i, "<unknown>");
-
-       do {
-               char *buf = *buf_p;
-               unsigned long space = *space_p;
-
-               space = interpolate(buf, space, user_format,
-                                   table, ARRAY_SIZE(table));
-               if (!space)
-                       break;
-               buf = xrealloc(buf, space);
-               *buf_p = buf;
-               *space_p = space;
-       } while (1);
-       interp_clear_table(table, ARRAY_SIZE(table));
-
-       return strlen(*buf_p);
-}
-
-static void pp_header(enum cmit_fmt fmt,
-                     int abbrev,
-                     enum date_mode dmode,
-                     const char *encoding,
-                     const struct commit *commit,
-                     const char **msg_p,
-                     unsigned long *len_p,
-                     unsigned long *ofs_p,
-                     char **buf_p,
-                     unsigned long *space_p)
-{
-       int parents_shown = 0;
-
-       for (;;) {
-               const char *line = *msg_p;
-               char *dst;
-               int linelen = get_one_line(*msg_p, *len_p);
-               unsigned long len;
-
-               if (!linelen)
-                       return;
-               *msg_p += linelen;
-               *len_p -= linelen;
-
-               if (linelen == 1)
-                       /* End of header */
-                       return;
-
-               ALLOC_GROW(*buf_p, linelen + *ofs_p + 20, *space_p);
-               dst = *buf_p + *ofs_p;
-
-               if (fmt == CMIT_FMT_RAW) {
-                       memcpy(dst, line, linelen);
-                       *ofs_p += linelen;
-                       continue;
-               }
-
-               if (!memcmp(line, "parent ", 7)) {
-                       if (linelen != 48)
-                               die("bad parent line in commit");
-                       continue;
-               }
-
-               if (!parents_shown) {
-                       struct commit_list *parent;
-                       int num;
-                       for (parent = commit->parents, num = 0;
-                            parent;
-                            parent = parent->next, num++)
-                               ;
-                       /* with enough slop */
-                       num = *ofs_p + num * 50 + 20;
-                       ALLOC_GROW(*buf_p, num, *space_p);
-                       dst = *buf_p + *ofs_p;
-                       *ofs_p += add_merge_info(fmt, dst, commit, abbrev);
-                       parents_shown = 1;
-               }
-
-               /*
-                * MEDIUM == DEFAULT shows only author with dates.
-                * FULL shows both authors but not dates.
-                * FULLER shows both authors and dates.
-                */
-               if (!memcmp(line, "author ", 7)) {
-                       len = linelen;
-                       if (fmt == CMIT_FMT_EMAIL)
-                               len = bound_rfc2047(linelen, encoding);
-                       ALLOC_GROW(*buf_p, *ofs_p + len + 80, *space_p);
-                       dst = *buf_p + *ofs_p;
-                       *ofs_p += add_user_info("Author", fmt, dst,
-                                               line + 7, dmode, encoding);
-               }
-
-               if (!memcmp(line, "committer ", 10) &&
-                   (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
-                       len = linelen;
-                       if (fmt == CMIT_FMT_EMAIL)
-                               len = bound_rfc2047(linelen, encoding);
-                       ALLOC_GROW(*buf_p, *ofs_p + len + 80, *space_p);
-                       dst = *buf_p + *ofs_p;
-                       *ofs_p += add_user_info("Commit", fmt, dst,
-                                               line + 10, dmode, encoding);
-               }
-       }
-}
-
-static void pp_title_line(enum cmit_fmt fmt,
-                         const char **msg_p,
-                         unsigned long *len_p,
-                         unsigned long *ofs_p,
-                         char **buf_p,
-                         unsigned long *space_p,
-                         int indent,
-                         const char *subject,
-                         const char *after_subject,
-                         const char *encoding,
-                         int plain_non_ascii)
-{
-       char *title;
-       unsigned long title_alloc, title_len;
-       unsigned long len;
-
-       title_len = 0;
-       title_alloc = 80;
-       title = xmalloc(title_alloc);
-       for (;;) {
-               const char *line = *msg_p;
-               int linelen = get_one_line(line, *len_p);
-               *msg_p += linelen;
-               *len_p -= linelen;
-
-               if (!linelen || is_empty_line(line, &linelen))
-                       break;
-
-               if (title_alloc <= title_len + linelen + 2) {
-                       title_alloc = title_len + linelen + 80;
-                       title = xrealloc(title, title_alloc);
-               }
-               len = 0;
-               if (title_len) {
-                       if (fmt == CMIT_FMT_EMAIL) {
-                               len++;
-                               title[title_len++] = '\n';
-                       }
-                       len++;
-                       title[title_len++] = ' ';
-               }
-               memcpy(title + title_len, line, linelen);
-               title_len += linelen;
-       }
-
-       /* Enough slop for the MIME header and rfc2047 */
-       len = bound_rfc2047(title_len, encoding)+ 1000;
-       if (subject)
-               len += strlen(subject);
-       if (after_subject)
-               len += strlen(after_subject);
-       if (encoding)
-               len += strlen(encoding);
-       ALLOC_GROW(*buf_p, title_len + *ofs_p + len, *space_p);
-
-       if (subject) {
-               len = strlen(subject);
-               memcpy(*buf_p + *ofs_p, subject, len);
-               *ofs_p += len;
-               *ofs_p += add_rfc2047(*buf_p + *ofs_p,
-                                     title, title_len, encoding);
-       } else {
-               memcpy(*buf_p + *ofs_p, title, title_len);
-               *ofs_p += title_len;
-       }
-       (*buf_p)[(*ofs_p)++] = '\n';
-       if (plain_non_ascii) {
-               const char *header_fmt =
-                       "MIME-Version: 1.0\n"
-                       "Content-Type: text/plain; charset=%s\n"
-                       "Content-Transfer-Encoding: 8bit\n";
-               *ofs_p += snprintf(*buf_p + *ofs_p,
-                                  *space_p - *ofs_p,
-                                  header_fmt, encoding);
-       }
-       if (after_subject) {
-               len = strlen(after_subject);
-               memcpy(*buf_p + *ofs_p, after_subject, len);
-               *ofs_p += len;
-       }
-       free(title);
-       if (fmt == CMIT_FMT_EMAIL) {
-               ALLOC_GROW(*buf_p, *ofs_p + 20, *space_p);
-               (*buf_p)[(*ofs_p)++] = '\n';
-       }
-}
-
-static void pp_remainder(enum cmit_fmt fmt,
-                        const char **msg_p,
-                        unsigned long *len_p,
-                        unsigned long *ofs_p,
-                        char **buf_p,
-                        unsigned long *space_p,
-                        int indent)
-{
-       int first = 1;
-       for (;;) {
-               const char *line = *msg_p;
-               int linelen = get_one_line(line, *len_p);
-               *msg_p += linelen;
-               *len_p -= linelen;
-
-               if (!linelen)
-                       break;
-
-               if (is_empty_line(line, &linelen)) {
-                       if (first)
-                               continue;
-                       if (fmt == CMIT_FMT_SHORT)
-                               break;
-               }
-               first = 0;
-
-               ALLOC_GROW(*buf_p, *ofs_p + linelen + indent + 20, *space_p);
-               if (indent) {
-                       memset(*buf_p + *ofs_p, ' ', indent);
-                       *ofs_p += indent;
-               }
-               memcpy(*buf_p + *ofs_p, line, linelen);
-               *ofs_p += linelen;
-               (*buf_p)[(*ofs_p)++] = '\n';
-       }
-}
-
-unsigned long pretty_print_commit(enum cmit_fmt fmt,
-                                 const struct commit *commit,
-                                 unsigned long len,
-                                 char **buf_p, unsigned long *space_p,
-                                 int abbrev, const char *subject,
-                                 const char *after_subject,
-                                 enum date_mode dmode)
-{
-       unsigned long offset = 0;
-       unsigned long beginning_of_body;
-       int indent = 4;
-       const char *msg = commit->buffer;
-       int plain_non_ascii = 0;
-       char *reencoded;
-       const char *encoding;
-       char *buf;
-
-       if (fmt == CMIT_FMT_USERFORMAT)
-               return format_commit_message(commit, msg, buf_p, space_p);
-
-       encoding = (git_log_output_encoding
-                   ? git_log_output_encoding
-                   : git_commit_encoding);
-       if (!encoding)
-               encoding = "utf-8";
-       reencoded = logmsg_reencode(commit, encoding);
-       if (reencoded) {
-               msg = reencoded;
-               len = strlen(reencoded);
-       }
-
-       if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
-               indent = 0;
-
-       /* After-subject is used to pass in Content-Type: multipart
-        * MIME header; in that case we do not have to do the
-        * plaintext content type even if the commit message has
-        * non 7-bit ASCII character.  Otherwise, check if we need
-        * to say this is not a 7-bit ASCII.
-        */
-       if (fmt == CMIT_FMT_EMAIL && !after_subject) {
-               int i, ch, in_body;
-
-               for (in_body = i = 0; (ch = msg[i]) && i < len; i++) {
-                       if (!in_body) {
-                               /* author could be non 7-bit ASCII but
-                                * the log may be so; skip over the
-                                * header part first.
-                                */
-                               if (ch == '\n' &&
-                                   i + 1 < len && msg[i+1] == '\n')
-                                       in_body = 1;
-                       }
-                       else if (non_ascii(ch)) {
-                               plain_non_ascii = 1;
-                               break;
-                       }
-               }
-       }
-
-       pp_header(fmt, abbrev, dmode, encoding,
-                 commit, &msg, &len,
-                 &offset, buf_p, space_p);
-       if (fmt != CMIT_FMT_ONELINE && !subject) {
-               ALLOC_GROW(*buf_p, offset + 20, *space_p);
-               (*buf_p)[offset++] = '\n';
-       }
-
-       /* Skip excess blank lines at the beginning of body, if any... */
-       for (;;) {
-               int linelen = get_one_line(msg, len);
-               int ll = linelen;
-               if (!linelen)
-                       break;
-               if (!is_empty_line(msg, &ll))
-                       break;
-               msg += linelen;
-               len -= linelen;
-       }
-
-       /* These formats treat the title line specially. */
-       if (fmt == CMIT_FMT_ONELINE
-           || fmt == CMIT_FMT_EMAIL)
-               pp_title_line(fmt, &msg, &len, &offset,
-                             buf_p, space_p, indent,
-                             subject, after_subject, encoding,
-                             plain_non_ascii);
-
-       beginning_of_body = offset;
-       if (fmt != CMIT_FMT_ONELINE)
-               pp_remainder(fmt, &msg, &len, &offset,
-                            buf_p, space_p, indent);
-
-       while (offset && isspace((*buf_p)[offset-1]))
-               offset--;
-
-       ALLOC_GROW(*buf_p, offset + 20, *space_p);
-       buf = *buf_p;
-
-       /* Make sure there is an EOLN for the non-oneline case */
-       if (fmt != CMIT_FMT_ONELINE)
-               buf[offset++] = '\n';
-
-       /*
-        * The caller may append additional body text in e-mail
-        * format.  Make sure we did not strip the blank line
-        * between the header and the body.
-        */
-       if (fmt == CMIT_FMT_EMAIL && offset <= beginning_of_body)
-               buf[offset++] = '\n';
-       buf[offset] = '\0';
-       free(reencoded);
-       return offset;
-}
-
 struct commit *pop_commit(struct commit_list **stack)
 {
        struct commit_list *top = *stack;
@@ -1279,125 +428,95 @@ struct commit *pop_commit(struct commit_list **stack)
        return item;
 }
 
-void topo_sort_default_setter(struct commit *c, void *data)
-{
-       c->util = data;
-}
-
-void *topo_sort_default_getter(struct commit *c)
-{
-       return c->util;
-}
-
 /*
  * Performs an in-place topological sort on the list supplied.
  */
 void sort_in_topological_order(struct commit_list ** list, int lifo)
 {
-       sort_in_topological_order_fn(list, lifo, topo_sort_default_setter,
-                                    topo_sort_default_getter);
-}
-
-void sort_in_topological_order_fn(struct commit_list ** list, int lifo,
-                                 topo_sort_set_fn_t setter,
-                                 topo_sort_get_fn_t getter)
-{
-       struct commit_list * next = *list;
-       struct commit_list * work = NULL, **insert;
-       struct commit_list ** pptr = list;
-       struct sort_node * nodes;
-       struct sort_node * next_nodes;
-       int count = 0;
-
-       /* determine the size of the list */
-       while (next) {
-               next = next->next;
-               count++;
-       }
+       struct commit_list *next, *orig = *list;
+       struct commit_list *work, **insert;
+       struct commit_list **pptr;
 
-       if (!count)
+       if (!orig)
                return;
-       /* allocate an array to help sort the list */
-       nodes = xcalloc(count, sizeof(*nodes));
-       /* link the list to the array */
-       next_nodes = nodes;
-       next=*list;
-       while (next) {
-               next_nodes->list_item = next;
-               setter(next->item, next_nodes);
-               next_nodes++;
-               next = next->next;
+       *list = NULL;
+
+       /* Mark them and clear the indegree */
+       for (next = orig; next; next = next->next) {
+               struct commit *commit = next->item;
+               commit->object.flags |= TOPOSORT;
+               commit->indegree = 0;
        }
+
        /* update the indegree */
-       next=*list;
-       while (next) {
+       for (next = orig; next; next = next->next) {
                struct commit_list * parents = next->item->parents;
                while (parents) {
-                       struct commit * parent=parents->item;
-                       struct sort_node * pn = (struct sort_node *) getter(parent);
+                       struct commit *parent = parents->item;
 
-                       if (pn)
-                               pn->indegree++;
-                       parents=parents->next;
+                       if (parent->object.flags & TOPOSORT)
+                               parent->indegree++;
+                       parents = parents->next;
                }
-               next=next->next;
        }
+
        /*
-         * find the tips
-         *
-         * tips are nodes not reachable from any other node in the list
-         *
-         * the tips serve as a starting set for the work queue.
-         */
-       next=*list;
+        * find the tips
+        *
+        * tips are nodes not reachable from any other node in the list
+        *
+        * the tips serve as a starting set for the work queue.
+        */
+       work = NULL;
        insert = &work;
-       while (next) {
-               struct sort_node * node = (struct sort_node *) getter(next->item);
+       for (next = orig; next; next = next->next) {
+               struct commit *commit = next->item;
 
-               if (node->indegree == 0) {
-                       insert = &commit_list_insert(next->item, insert)->next;
-               }
-               next=next->next;
+               if (!commit->indegree)
+                       insert = &commit_list_insert(commit, insert)->next;
        }
 
        /* process the list in topological order */
        if (!lifo)
                sort_by_date(&work);
+
+       pptr = list;
+       *list = NULL;
        while (work) {
-               struct commit * work_item = pop_commit(&work);
-               struct sort_node * work_node = (struct sort_node *) getter(work_item);
-               struct commit_list * parents = work_item->parents;
+               struct commit *commit;
+               struct commit_list *parents, *work_item;
 
-               while (parents) {
-                       struct commit * parent=parents->item;
-                       struct sort_node * pn = (struct sort_node *) getter(parent);
-
-                       if (pn) {
-                               /*
-                                * parents are only enqueued for emission
-                                 * when all their children have been emitted thereby
-                                 * guaranteeing topological order.
-                                 */
-                               pn->indegree--;
-                               if (!pn->indegree) {
-                                       if (!lifo)
-                                               insert_by_date(parent, &work);
-                                       else
-                                               commit_list_insert(parent, &work);
-                               }
+               work_item = work;
+               work = work_item->next;
+               work_item->next = NULL;
+
+               commit = work_item->item;
+               for (parents = commit->parents; parents ; parents = parents->next) {
+                       struct commit *parent=parents->item;
+
+                       if (!(parent->object.flags & TOPOSORT))
+                               continue;
+
+                       /*
+                        * parents are only enqueued for emission
+                        * when all their children have been emitted thereby
+                        * guaranteeing topological order.
+                        */
+                       if (!--parent->indegree) {
+                               if (!lifo)
+                                       insert_by_date(parent, &work);
+                               else
+                                       commit_list_insert(parent, &work);
                        }
-                       parents=parents->next;
                }
                /*
-                 * work_item is a commit all of whose children
-                 * have already been emitted. we can emit it now.
-                 */
-               *pptr = work_node->list_item;
-               pptr = &(*pptr)->next;
-               *pptr = NULL;
-               setter(work_item, NULL);
+                * work_item is a commit all of whose children
+                * have already been emitted. we can emit it now.
+                */
+               commit->object.flags &= ~TOPOSORT;
+               *pptr = work_item;
+               pptr = &work_item->next;
        }
-       free(nodes);
 }
 
 /* merge-base stuff */
@@ -1487,8 +606,7 @@ static struct commit_list *merge_bases(struct commit *one, struct commit *two)
 }
 
 struct commit_list *get_merge_bases(struct commit *one,
-                                   struct commit *two,
-                                    int cleanup)
+                                       struct commit *two, int cleanup)
 {
        struct commit_list *list;
        struct commit **rslt;