completion: document tilde expansion failure in tests
[gitweb.git] / pretty.c
index 622275c70b5fb57d08ede37b25b774baee5b72e6..ba3148181eff960ff6de68cafffe77e37719fd1d 100644 (file)
--- a/pretty.c
+++ b/pretty.c
@@ -345,7 +345,7 @@ static int needs_rfc2047_encoding(const char *line, int len,
        return 0;
 }
 
-static void add_rfc2047(struct strbuf *sb, const char *line, int len,
+static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
                       const char *encoding, enum rfc2047_type type)
 {
        static const int max_encoded_length = 76; /* per rfc2047 */
@@ -355,9 +355,22 @@ static void add_rfc2047(struct strbuf *sb, const char *line, int len,
        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
        strbuf_addf(sb, "=?%s?q?", encoding);
        line_len += strlen(encoding) + 5; /* 5 for =??q? */
-       for (i = 0; i < len; i++) {
-               unsigned ch = line[i] & 0xFF;
-               int is_special = is_rfc2047_special(ch, type);
+
+       while (len) {
+               /*
+                * RFC 2047, section 5 (3):
+                *
+                * Each 'encoded-word' MUST represent an integral number of
+                * characters.  A multi-octet character may not be split across
+                * adjacent 'encoded- word's.
+                */
+               const unsigned char *p = (const unsigned char *)line;
+               int chrlen = mbs_chrlen(&line, &len, encoding);
+               int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
+
+               /* "=%02X" * chrlen, or the byte itself */
+               const char *encoded_fmt = is_special ? "=%02X"    : "%c";
+               int         encoded_len = is_special ? 3 * chrlen : 1;
 
                /*
                 * According to RFC 2047, we could encode the special character
@@ -367,22 +380,32 @@ static void add_rfc2047(struct strbuf *sb, const char *line, int len,
                 * causes ' ' to be encoded as '=20', avoiding this problem.
                 */
 
-               if (line_len + 2 + (is_special ? 3 : 1) > max_encoded_length) {
+               if (line_len + encoded_len + 2 > max_encoded_length) {
+                       /* It won't fit with trailing "?=" --- break the line */
                        strbuf_addf(sb, "?=\n =?%s?q?", encoding);
                        line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
                }
 
-               if (is_special) {
-                       strbuf_addf(sb, "=%02X", ch);
-                       line_len += 3;
-               } else {
-                       strbuf_addch(sb, ch);
-                       line_len++;
-               }
+               for (i = 0; i < chrlen; i++)
+                       strbuf_addf(sb, encoded_fmt, p[i]);
+               line_len += encoded_len;
        }
        strbuf_addstr(sb, "?=");
 }
 
+static const char *show_ident_date(const struct ident_split *ident,
+                                  enum date_mode mode)
+{
+       unsigned long date = 0;
+       int tz = 0;
+
+       if (ident->date_begin && ident->date_end)
+               date = strtoul(ident->date_begin, NULL, 10);
+       if (ident->tz_begin && ident->tz_end)
+               tz = strtol(ident->tz_begin, NULL, 10);
+       return show_date(date, tz, mode);
+}
+
 void pp_user_info(const struct pretty_print_context *pp,
                  const char *what, struct strbuf *sb,
                  const char *line, const char *encoding)
@@ -391,12 +414,10 @@ void pp_user_info(const struct pretty_print_context *pp,
        struct strbuf mail;
        struct ident_split ident;
        int linelen;
-       char *line_end, *date;
+       char *line_end;
        const char *mailbuf, *namebuf;
        size_t namelen, maillen;
        int max_length = 78; /* per rfc2822 */
-       unsigned long time;
-       int tz;
 
        if (pp->fmt == CMIT_FMT_ONELINE)
                return;
@@ -428,8 +449,6 @@ void pp_user_info(const struct pretty_print_context *pp,
        strbuf_add(&name, namebuf, namelen);
 
        namelen = name.len + mail.len + 3; /* ' ' + '<' + '>' */
-       time = strtoul(ident.date_begin, &date, 10);
-       tz = strtol(date, NULL, 10);
 
        if (pp->fmt == CMIT_FMT_EMAIL) {
                strbuf_addstr(sb, "From: ");
@@ -462,13 +481,16 @@ void pp_user_info(const struct pretty_print_context *pp,
 
        switch (pp->fmt) {
        case CMIT_FMT_MEDIUM:
-               strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, pp->date_mode));
+               strbuf_addf(sb, "Date:   %s\n",
+                           show_ident_date(&ident, pp->date_mode));
                break;
        case CMIT_FMT_EMAIL:
-               strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
+               strbuf_addf(sb, "Date: %s\n",
+                           show_ident_date(&ident, DATE_RFC2822));
                break;
        case CMIT_FMT_FULLER:
-               strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, pp->date_mode));
+               strbuf_addf(sb, "%sDate: %s\n", what,
+                           show_ident_date(&ident, pp->date_mode));
                break;
        default:
                /* notin' */
@@ -524,10 +546,11 @@ static void add_merge_info(const struct pretty_print_context *pp,
        strbuf_addch(sb, '\n');
 }
 
-static char *get_header(const struct commit *commit, const char *key)
+static char *get_header(const struct commit *commit, const char *msg,
+                       const char *key)
 {
        int key_len = strlen(key);
-       const char *line = commit->buffer;
+       const char *line = msg;
 
        while (line) {
                const char *eol = strchr(line, '\n'), *next;
@@ -583,30 +606,90 @@ static char *replace_encoding_header(char *buf, const char *encoding)
 }
 
 char *logmsg_reencode(const struct commit *commit,
+                     char **commit_encoding,
                      const char *output_encoding)
 {
        static const char *utf8 = "UTF-8";
        const char *use_encoding;
        char *encoding;
+       char *msg = commit->buffer;
        char *out;
 
-       if (!*output_encoding)
-               return NULL;
-       encoding = get_header(commit, "encoding");
+       if (!msg) {
+               enum object_type type;
+               unsigned long size;
+
+               msg = read_sha1_file(commit->object.sha1, &type, &size);
+               if (!msg)
+                       die("Cannot read commit object %s",
+                           sha1_to_hex(commit->object.sha1));
+               if (type != OBJ_COMMIT)
+                       die("Expected commit for '%s', got %s",
+                           sha1_to_hex(commit->object.sha1), typename(type));
+       }
+
+       if (!output_encoding || !*output_encoding) {
+               if (commit_encoding)
+                       *commit_encoding =
+                               get_header(commit, msg, "encoding");
+               return msg;
+       }
+       encoding = get_header(commit, msg, "encoding");
+       if (commit_encoding)
+               *commit_encoding = encoding;
        use_encoding = encoding ? encoding : utf8;
-       if (same_encoding(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 (same_encoding(use_encoding, output_encoding)) {
+               /*
+                * No encoding work to be done. If we have no encoding header
+                * at all, then there's nothing to do, and we can return the
+                * message verbatim (whether newly allocated or not).
+                */
+               if (!encoding)
+                       return msg;
+
+               /*
+                * Otherwise, we still want to munge the encoding header in the
+                * result, which will be done by modifying the buffer. If we
+                * are using a fresh copy, we can reuse it. But if we are using
+                * the cached copy from commit->buffer, we need to duplicate it
+                * to avoid munging commit->buffer.
+                */
+               out = msg;
+               if (out == commit->buffer)
+                       out = xstrdup(out);
+       }
+       else {
+               /*
+                * There's actual encoding work to do. Do the reencoding, which
+                * still leaves the header to be replaced in the next step. At
+                * this point, we are done with msg. If we allocated a fresh
+                * copy, we can free it.
+                */
+               out = reencode_string(msg, output_encoding, use_encoding);
+               if (out && msg != commit->buffer)
+                       free(msg);
+       }
+
+       /*
+        * This replacement actually consumes the buffer we hand it, so we do
+        * not have to worry about freeing the old "out" here.
+        */
        if (out)
                out = replace_encoding_header(out, output_encoding);
 
-       free(encoding);
-       return out;
+       if (!commit_encoding)
+               free(encoding);
+       /*
+        * If the re-encoding failed, out might be NULL here; in that
+        * case we just return the commit message verbatim.
+        */
+       return out ? out : msg;
+}
+
+void logmsg_free(char *msg, const struct commit *commit)
+{
+       if (msg != commit->buffer)
+               free(msg);
 }
 
 static int mailmap_name(const char **email, size_t *email_len,
@@ -625,8 +708,6 @@ static size_t format_person_part(struct strbuf *sb, char part,
 {
        /* currently all placeholders have same length */
        const int placeholder_len = 2;
-       int tz;
-       unsigned long date = 0;
        struct ident_split s;
        const char *name, *mail;
        size_t maillen, namelen;
@@ -653,30 +734,23 @@ static size_t format_person_part(struct strbuf *sb, char part,
        if (!s.date_begin)
                goto skip;
 
-       date = strtoul(s.date_begin, NULL, 10);
-
        if (part == 't') {      /* date, UNIX timestamp */
                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
                return placeholder_len;
        }
 
-       /* parse tz */
-       tz = strtoul(s.tz_begin + 1, NULL, 10);
-       if (*s.tz_begin == '-')
-               tz = -tz;
-
        switch (part) {
        case 'd':       /* date */
-               strbuf_addstr(sb, show_date(date, tz, dmode));
+               strbuf_addstr(sb, show_ident_date(&s, dmode));
                return placeholder_len;
        case 'D':       /* date, RFC2822 style */
-               strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
+               strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
                return placeholder_len;
        case 'r':       /* date, relative */
-               strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
+               strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
                return placeholder_len;
        case 'i':       /* date, ISO 8601 */
-               strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
+               strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
                return placeholder_len;
        }
 
@@ -698,24 +772,38 @@ struct chunk {
        size_t len;
 };
 
+enum flush_type {
+       no_flush,
+       flush_right,
+       flush_left,
+       flush_left_and_steal,
+       flush_both
+};
+
+enum trunc_type {
+       trunc_none,
+       trunc_left,
+       trunc_middle,
+       trunc_right
+};
+
 struct format_commit_context {
        const struct commit *commit;
        const struct pretty_print_context *pretty_ctx;
        unsigned commit_header_parsed:1;
        unsigned commit_message_parsed:1;
-       unsigned commit_signature_parsed:1;
-       struct {
-               char *gpg_output;
-               char good_bad;
-               char *signer;
-       } signature;
+       struct signature_check signature_check;
+       enum flush_type flush_type;
+       enum trunc_type truncate;
        char *message;
+       char *commit_encoding;
        size_t width, indent1, indent2;
+       int auto_color;
+       int padding;
 
        /* These offsets are relative to the start of the commit message. */
        struct chunk author;
        struct chunk committer;
-       struct chunk encoding;
        size_t message_off;
        size_t subject_off;
        size_t body_off;
@@ -762,9 +850,6 @@ static void parse_commit_header(struct format_commit_context *context)
                } else if (!prefixcmp(msg + i, "committer ")) {
                        context->committer.off = i + 10;
                        context->committer.len = eol - i - 10;
-               } else if (!prefixcmp(msg + i, "encoding ")) {
-                       context->encoding.off = i + 9;
-                       context->encoding.len = eol - i - 9;
                }
                i = eol;
        }
@@ -845,23 +930,6 @@ static void parse_commit_message(struct format_commit_context *c)
        c->commit_message_parsed = 1;
 }
 
-static void format_decoration(struct strbuf *sb, const struct commit *commit)
-{
-       struct name_decoration *d;
-       const char *prefix = " (";
-
-       load_ref_decorations(DECORATE_SHORT_REFS);
-       d = lookup_decoration(&name_decoration, &commit->object);
-       while (d) {
-               strbuf_addstr(sb, prefix);
-               prefix = ", ";
-               strbuf_addstr(sb, d->name);
-               d = d->next;
-       }
-       if (prefix[0] == ',')
-               strbuf_addch(sb, ')');
-}
-
 static void strbuf_wrap(struct strbuf *sb, size_t pos,
                        size_t width, size_t indent1, size_t indent2)
 {
@@ -891,59 +959,6 @@ static void rewrap_message_tail(struct strbuf *sb,
        c->indent2 = new_indent2;
 }
 
-static struct {
-       char result;
-       const char *check;
-} signature_check[] = {
-       { 'G', ": Good signature from " },
-       { 'B', ": BAD signature from " },
-};
-
-static void parse_signature_lines(struct format_commit_context *ctx)
-{
-       const char *buf = ctx->signature.gpg_output;
-       int i;
-
-       for (i = 0; i < ARRAY_SIZE(signature_check); i++) {
-               const char *found = strstr(buf, signature_check[i].check);
-               const char *next;
-               if (!found)
-                       continue;
-               ctx->signature.good_bad = signature_check[i].result;
-               found += strlen(signature_check[i].check);
-               next = strchrnul(found, '\n');
-               ctx->signature.signer = xmemdupz(found, next - found);
-               break;
-       }
-}
-
-static void parse_commit_signature(struct format_commit_context *ctx)
-{
-       struct strbuf payload = STRBUF_INIT;
-       struct strbuf signature = STRBUF_INIT;
-       struct strbuf gpg_output = STRBUF_INIT;
-       int status;
-
-       ctx->commit_signature_parsed = 1;
-
-       if (parse_signed_commit(ctx->commit->object.sha1,
-                               &payload, &signature) <= 0)
-               goto out;
-       status = verify_signed_buffer(payload.buf, payload.len,
-                                     signature.buf, signature.len,
-                                     &gpg_output);
-       if (status && !gpg_output.len)
-               goto out;
-       ctx->signature.gpg_output = strbuf_detach(&gpg_output, NULL);
-       parse_signature_lines(ctx);
-
- out:
-       strbuf_release(&gpg_output);
-       strbuf_release(&payload);
-       strbuf_release(&signature);
-}
-
-
 static int format_reflog_person(struct strbuf *sb,
                                char part,
                                struct reflog_walk_info *log,
@@ -961,7 +976,112 @@ static int format_reflog_person(struct strbuf *sb,
        return format_person_part(sb, part, ident, strlen(ident), dmode);
 }
 
-static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
+static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
+                         const char *placeholder,
+                         struct format_commit_context *c)
+{
+       if (placeholder[1] == '(') {
+               const char *begin = placeholder + 2;
+               const char *end = strchr(begin, ')');
+               char color[COLOR_MAXLEN];
+
+               if (!end)
+                       return 0;
+               if (!prefixcmp(begin, "auto,")) {
+                       if (!want_color(c->pretty_ctx->color))
+                               return end - placeholder + 1;
+                       begin += 5;
+               }
+               color_parse_mem(begin,
+                               end - begin,
+                               "--pretty format", color);
+               strbuf_addstr(sb, color);
+               return end - placeholder + 1;
+       }
+       if (!prefixcmp(placeholder + 1, "red")) {
+               strbuf_addstr(sb, GIT_COLOR_RED);
+               return 4;
+       } else if (!prefixcmp(placeholder + 1, "green")) {
+               strbuf_addstr(sb, GIT_COLOR_GREEN);
+               return 6;
+       } else if (!prefixcmp(placeholder + 1, "blue")) {
+               strbuf_addstr(sb, GIT_COLOR_BLUE);
+               return 5;
+       } else if (!prefixcmp(placeholder + 1, "reset")) {
+               strbuf_addstr(sb, GIT_COLOR_RESET);
+               return 6;
+       } else
+               return 0;
+}
+
+static size_t parse_padding_placeholder(struct strbuf *sb,
+                                       const char *placeholder,
+                                       struct format_commit_context *c)
+{
+       const char *ch = placeholder;
+       enum flush_type flush_type;
+       int to_column = 0;
+
+       switch (*ch++) {
+       case '<':
+               flush_type = flush_right;
+               break;
+       case '>':
+               if (*ch == '<') {
+                       flush_type = flush_both;
+                       ch++;
+               } else if (*ch == '>') {
+                       flush_type = flush_left_and_steal;
+                       ch++;
+               } else
+                       flush_type = flush_left;
+               break;
+       default:
+               return 0;
+       }
+
+       /* the next value means "wide enough to that column" */
+       if (*ch == '|') {
+               to_column = 1;
+               ch++;
+       }
+
+       if (*ch == '(') {
+               const char *start = ch + 1;
+               const char *end = start + strcspn(start, ",)");
+               char *next;
+               int width;
+               if (!end || end == start)
+                       return 0;
+               width = strtoul(start, &next, 10);
+               if (next == start || width == 0)
+                       return 0;
+               c->padding = to_column ? -width : width;
+               c->flush_type = flush_type;
+
+               if (*end == ',') {
+                       start = end + 1;
+                       end = strchr(start, ')');
+                       if (!end || end == start)
+                               return 0;
+                       if (!prefixcmp(start, "trunc)"))
+                               c->truncate = trunc_right;
+                       else if (!prefixcmp(start, "ltrunc)"))
+                               c->truncate = trunc_left;
+                       else if (!prefixcmp(start, "mtrunc)"))
+                               c->truncate = trunc_middle;
+                       else
+                               return 0;
+               } else
+                       c->truncate = trunc_none;
+
+               return end - placeholder + 1;
+       }
+       return 0;
+}
+
+static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
+                               const char *placeholder,
                                void *context)
 {
        struct format_commit_context *c = context;
@@ -973,31 +1093,20 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
        /* these are independent of the commit */
        switch (placeholder[0]) {
        case 'C':
-               if (placeholder[1] == '(') {
-                       const char *end = strchr(placeholder + 2, ')');
-                       char color[COLOR_MAXLEN];
-                       if (!end)
-                               return 0;
-                       color_parse_mem(placeholder + 2,
-                                       end - (placeholder + 2),
-                                       "--pretty format", color);
-                       strbuf_addstr(sb, color);
-                       return end - placeholder + 1;
+               if (!prefixcmp(placeholder + 1, "(auto)")) {
+                       c->auto_color = 1;
+                       return 7; /* consumed 7 bytes, "C(auto)" */
+               } else {
+                       int ret = parse_color(sb, placeholder, c);
+                       if (ret)
+                               c->auto_color = 0;
+                       /*
+                        * Otherwise, we decided to treat %C<unknown>
+                        * as a literal string, and the previous
+                        * %C(auto) is still valid.
+                        */
+                       return ret;
                }
-               if (!prefixcmp(placeholder + 1, "red")) {
-                       strbuf_addstr(sb, GIT_COLOR_RED);
-                       return 4;
-               } else if (!prefixcmp(placeholder + 1, "green")) {
-                       strbuf_addstr(sb, GIT_COLOR_GREEN);
-                       return 6;
-               } else if (!prefixcmp(placeholder + 1, "blue")) {
-                       strbuf_addstr(sb, GIT_COLOR_BLUE);
-                       return 5;
-               } else if (!prefixcmp(placeholder + 1, "reset")) {
-                       strbuf_addstr(sb, GIT_COLOR_RESET);
-                       return 6;
-               } else
-                       return 0;
        case 'n':               /* newline */
                strbuf_addch(sb, '\n');
                return 1;
@@ -1035,6 +1144,10 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
                        return end - placeholder + 1;
                } else
                        return 0;
+
+       case '<':
+       case '>':
+               return parse_padding_placeholder(sb, placeholder, c);
        }
 
        /* these depend on the commit */
@@ -1043,13 +1156,19 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 
        switch (placeholder[0]) {
        case 'H':               /* commit hash */
+               strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
+               strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
                return 1;
        case 'h':               /* abbreviated commit hash */
-               if (add_again(sb, &c->abbrev_commit_hash))
+               strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
+               if (add_again(sb, &c->abbrev_commit_hash)) {
+                       strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
                        return 1;
+               }
                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
                                                     c->pretty_ctx->abbrev));
+               strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
                return 1;
        case 'T':               /* tree hash */
@@ -1086,7 +1205,8 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
                strbuf_addstr(sb, get_revision_mark(NULL, commit));
                return 1;
        case 'd':
-               format_decoration(sb, commit);
+               load_ref_decorations(DECORATE_SHORT_REFS);
+               format_decorations(sb, commit, c->auto_color);
                return 1;
        case 'g':               /* reflog info */
                switch(placeholder[1]) {
@@ -1122,23 +1242,29 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
        }
 
        if (placeholder[0] == 'G') {
-               if (!c->commit_signature_parsed)
-                       parse_commit_signature(c);
+               if (!c->signature_check.result)
+                       check_commit_signature(c->commit, &(c->signature_check));
                switch (placeholder[1]) {
                case 'G':
-                       if (c->signature.gpg_output)
-                               strbuf_addstr(sb, c->signature.gpg_output);
+                       if (c->signature_check.gpg_output)
+                               strbuf_addstr(sb, c->signature_check.gpg_output);
                        break;
                case '?':
-                       switch (c->signature.good_bad) {
+                       switch (c->signature_check.result) {
                        case 'G':
                        case 'B':
-                               strbuf_addch(sb, c->signature.good_bad);
+                       case 'U':
+                       case 'N':
+                               strbuf_addch(sb, c->signature_check.result);
                        }
                        break;
                case 'S':
-                       if (c->signature.signer)
-                               strbuf_addstr(sb, c->signature.signer);
+                       if (c->signature_check.signer)
+                               strbuf_addstr(sb, c->signature_check.signer);
+                       break;
+               case 'K':
+                       if (c->signature_check.key)
+                               strbuf_addstr(sb, c->signature_check.key);
                        break;
                }
                return 2;
@@ -1159,7 +1285,8 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
                                   msg + c->committer.off, c->committer.len,
                                   c->pretty_ctx->date_mode);
        case 'e':       /* encoding */
-               strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
+               if (c->commit_encoding)
+                       strbuf_addstr(sb, c->commit_encoding);
                return 1;
        case 'B':       /* raw body */
                /* message_off is always left at the initial newline */
@@ -1185,7 +1312,111 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
        return 0;       /* unknown placeholder */
 }
 
-static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
+static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
+                                   const char *placeholder,
+                                   struct format_commit_context *c)
+{
+       struct strbuf local_sb = STRBUF_INIT;
+       int total_consumed = 0, len, padding = c->padding;
+       if (padding < 0) {
+               const char *start = strrchr(sb->buf, '\n');
+               int occupied;
+               if (!start)
+                       start = sb->buf;
+               occupied = utf8_strnwidth(start, -1, 1);
+               padding = (-padding) - occupied;
+       }
+       while (1) {
+               int modifier = *placeholder == 'C';
+               int consumed = format_commit_one(&local_sb, placeholder, c);
+               total_consumed += consumed;
+
+               if (!modifier)
+                       break;
+
+               placeholder += consumed;
+               if (*placeholder != '%')
+                       break;
+               placeholder++;
+               total_consumed++;
+       }
+       len = utf8_strnwidth(local_sb.buf, -1, 1);
+
+       if (c->flush_type == flush_left_and_steal) {
+               const char *ch = sb->buf + sb->len - 1;
+               while (len > padding && ch > sb->buf) {
+                       const char *p;
+                       if (*ch == ' ') {
+                               ch--;
+                               padding++;
+                               continue;
+                       }
+                       /* check for trailing ansi sequences */
+                       if (*ch != 'm')
+                               break;
+                       p = ch - 1;
+                       while (ch - p < 10 && *p != '\033')
+                               p--;
+                       if (*p != '\033' ||
+                           ch + 1 - p != display_mode_esc_sequence_len(p))
+                               break;
+                       /*
+                        * got a good ansi sequence, put it back to
+                        * local_sb as we're cutting sb
+                        */
+                       strbuf_insert(&local_sb, 0, p, ch + 1 - p);
+                       ch = p - 1;
+               }
+               strbuf_setlen(sb, ch + 1 - sb->buf);
+               c->flush_type = flush_left;
+       }
+
+       if (len > padding) {
+               switch (c->truncate) {
+               case trunc_left:
+                       strbuf_utf8_replace(&local_sb,
+                                           0, len - (padding - 2),
+                                           "..");
+                       break;
+               case trunc_middle:
+                       strbuf_utf8_replace(&local_sb,
+                                           padding / 2 - 1,
+                                           len - (padding - 2),
+                                           "..");
+                       break;
+               case trunc_right:
+                       strbuf_utf8_replace(&local_sb,
+                                           padding - 2, len - (padding - 2),
+                                           "..");
+                       break;
+               case trunc_none:
+                       break;
+               }
+               strbuf_addstr(sb, local_sb.buf);
+       } else {
+               int sb_len = sb->len, offset = 0;
+               if (c->flush_type == flush_left)
+                       offset = padding - len;
+               else if (c->flush_type == flush_both)
+                       offset = (padding - len) / 2;
+               /*
+                * we calculate padding in columns, now
+                * convert it back to chars
+                */
+               padding = padding - len + local_sb.len;
+               strbuf_grow(sb, padding);
+               strbuf_setlen(sb, sb_len + padding);
+               memset(sb->buf + sb_len, ' ', sb->len - sb_len);
+               memcpy(sb->buf + sb_len + offset, local_sb.buf,
+                      local_sb.len);
+       }
+       strbuf_release(&local_sb);
+       c->flush_type = no_flush;
+       return total_consumed;
+}
+
+static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
+                                const char *placeholder,
                                 void *context)
 {
        int consumed;
@@ -1214,7 +1445,10 @@ static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
                placeholder++;
 
        orig_len = sb->len;
-       consumed = format_commit_one(sb, placeholder, context);
+       if (((struct format_commit_context *)context)->flush_type != no_flush)
+               consumed = format_and_pad_commit(sb, placeholder, context);
+       else
+               consumed = format_commit_one(sb, placeholder, context);
        if (magic == NO_MAGIC)
                return consumed;
 
@@ -1264,31 +1498,41 @@ void format_commit_message(const struct commit *commit,
                           const struct pretty_print_context *pretty_ctx)
 {
        struct format_commit_context context;
-       static const char utf8[] = "UTF-8";
        const char *output_enc = pretty_ctx->output_encoding;
+       const char *utf8 = "UTF-8";
 
        memset(&context, 0, sizeof(context));
        context.commit = commit;
        context.pretty_ctx = pretty_ctx;
        context.wrap_start = sb->len;
-       context.message = commit->buffer;
-       if (output_enc) {
-               char *enc = get_header(commit, "encoding");
-               if (strcmp(enc ? enc : utf8, output_enc)) {
-                       context.message = logmsg_reencode(commit, output_enc);
-                       if (!context.message)
-                               context.message = commit->buffer;
-               }
-               free(enc);
-       }
+       context.message = logmsg_reencode(commit,
+                                         &context.commit_encoding,
+                                         output_enc);
 
        strbuf_expand(sb, format, format_commit_item, &context);
        rewrap_message_tail(sb, &context, 0, 0, 0);
 
-       if (context.message != commit->buffer)
-               free(context.message);
-       free(context.signature.gpg_output);
-       free(context.signature.signer);
+       if (output_enc) {
+               if (same_encoding(utf8, output_enc))
+                       output_enc = NULL;
+       } else {
+               if (context.commit_encoding &&
+                   !same_encoding(context.commit_encoding, utf8))
+                       output_enc = context.commit_encoding;
+       }
+
+       if (output_enc) {
+               int outsz;
+               char *out = reencode_string_len(sb->buf, sb->len,
+                                               output_enc, utf8, &outsz);
+               if (out)
+                       strbuf_attach(sb, out, outsz, outsz + 1);
+       }
+
+       free(context.commit_encoding);
+       logmsg_free(context.message, commit);
+       free(context.signature_check.gpg_output);
+       free(context.signature_check.signer);
 }
 
 static void pp_header(const struct pretty_print_context *pp,
@@ -1316,7 +1560,7 @@ static void pp_header(const struct pretty_print_context *pp,
                        continue;
                }
 
-               if (!memcmp(line, "parent ", 7)) {
+               if (!prefixcmp(line, "parent ")) {
                        if (linelen != 48)
                                die("bad parent line in commit");
                        continue;
@@ -1340,11 +1584,11 @@ static void pp_header(const struct pretty_print_context *pp,
                 * FULL shows both authors but not dates.
                 * FULLER shows both authors and dates.
                 */
-               if (!memcmp(line, "author ", 7)) {
+               if (!prefixcmp(line, "author ")) {
                        strbuf_grow(sb, linelen + 80);
                        pp_user_info(pp, "Author", sb, line + 7, encoding);
                }
-               if (!memcmp(line, "committer ", 10) &&
+               if (!prefixcmp(line, "committer ") &&
                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
                        strbuf_grow(sb, linelen + 80);
                        pp_user_info(pp, "Commit", sb, line + 10, encoding);
@@ -1433,7 +1677,7 @@ void pretty_print_commit(const struct pretty_print_context *pp,
 {
        unsigned long beginning_of_body;
        int indent = 4;
-       const char *msg = commit->buffer;
+       const char *msg;
        char *reencoded;
        const char *encoding;
        int need_8bit_cte = pp->need_8bit_cte;
@@ -1444,10 +1688,7 @@ void pretty_print_commit(const struct pretty_print_context *pp,
        }
 
        encoding = get_log_output_encoding();
-       reencoded = logmsg_reencode(commit, encoding);
-       if (reencoded) {
-               msg = reencoded;
-       }
+       msg = reencoded = logmsg_reencode(commit, NULL, encoding);
 
        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
                indent = 0;
@@ -1504,7 +1745,7 @@ void pretty_print_commit(const struct pretty_print_context *pp,
        if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
                strbuf_addch(sb, '\n');
 
-       free(reencoded);
+       logmsg_free(reencoded, commit);
 }
 
 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,