pretty-formats.txt: wrap long lines
[gitweb.git] / pretty.c
index 91bb2d3ef675b12ac5425ae71600e13c1adc34f9..e59688bcf5d07b562d11d25fc2d2d70a4388fda7 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
@@ -367,18 +380,15 @@ 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, "?=");
 }
@@ -387,56 +397,79 @@ void pp_user_info(const struct pretty_print_context *pp,
                  const char *what, struct strbuf *sb,
                  const char *line, const char *encoding)
 {
+       struct strbuf name;
+       struct strbuf mail;
+       struct ident_split ident;
+       int linelen;
+       char *line_end, *date;
+       const char *mailbuf, *namebuf;
+       size_t namelen, maillen;
        int max_length = 78; /* per rfc2822 */
-       char *date;
-       int namelen;
        unsigned long time;
        int tz;
 
        if (pp->fmt == CMIT_FMT_ONELINE)
                return;
-       date = strchr(line, '>');
-       if (!date)
+
+       line_end = strchr(line, '\n');
+       if (!line_end) {
+               line_end = strchr(line, '\0');
+               if (!line_end)
+                       return;
+       }
+
+       linelen = ++line_end - line;
+       if (split_ident_line(&ident, line, linelen))
                return;
-       namelen = ++date - line;
-       time = strtoul(date, &date, 10);
+
+
+       mailbuf = ident.mail_begin;
+       maillen = ident.mail_end - ident.mail_begin;
+       namebuf = ident.name_begin;
+       namelen = ident.name_end - ident.name_begin;
+
+       if (pp->mailmap)
+               map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
+
+       strbuf_init(&mail, 0);
+       strbuf_init(&name, 0);
+
+       strbuf_add(&mail, mailbuf, maillen);
+       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) {
-               char *name_tail = strchr(line, '<');
-               int display_name_length;
-               if (!name_tail)
-                       return;
-               while (line < name_tail && isspace(name_tail[-1]))
-                       name_tail--;
-               display_name_length = name_tail - line;
                strbuf_addstr(sb, "From: ");
-               if (needs_rfc2047_encoding(line, display_name_length, RFC2047_ADDRESS)) {
-                       add_rfc2047(sb, line, display_name_length,
-                                               encoding, RFC2047_ADDRESS);
+               if (needs_rfc2047_encoding(name.buf, name.len, RFC2047_ADDRESS)) {
+                       add_rfc2047(sb, name.buf, name.len,
+                                   encoding, RFC2047_ADDRESS);
                        max_length = 76; /* per rfc2047 */
-               } else if (needs_rfc822_quoting(line, display_name_length)) {
+               } else if (needs_rfc822_quoting(name.buf, name.len)) {
                        struct strbuf quoted = STRBUF_INIT;
-                       add_rfc822_quoted(&quoted, line, display_name_length);
+                       add_rfc822_quoted(&quoted, name.buf, name.len);
                        strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
                                                        -6, 1, max_length);
                        strbuf_release(&quoted);
                } else {
-                       strbuf_add_wrapped_bytes(sb, line, display_name_length,
-                                                       -6, 1, max_length);
+                       strbuf_add_wrapped_bytes(sb, name.buf, name.len,
+                                                -6, 1, max_length);
                }
-               if (namelen - display_name_length + last_line_length(sb) > max_length) {
+               if (namelen - name.len + last_line_length(sb) > max_length)
                        strbuf_addch(sb, '\n');
-                       if (!isspace(name_tail[0]))
-                               strbuf_addch(sb, ' ');
-               }
-               strbuf_add(sb, name_tail, namelen - display_name_length);
-               strbuf_addch(sb, '\n');
+
+               strbuf_addf(sb, " <%s>\n", mail.buf);
        } else {
-               strbuf_addf(sb, "%s: %.*s%.*s\n", what,
+               strbuf_addf(sb, "%s: %.*s%s <%s>\n", what,
                              (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0,
-                             "    ", namelen, line);
+                             "    ", name.buf, mail.buf);
        }
+
+       strbuf_release(&mail);
+       strbuf_release(&name);
+
        switch (pp->fmt) {
        case CMIT_FMT_MEDIUM:
                strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, pp->date_mode));
@@ -501,10 +534,11 @@ 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;
@@ -560,33 +594,94 @@ 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 || !*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(char *email, int email_len, char *name, int name_len)
+static int mailmap_name(const char **email, size_t *email_len,
+                       const char **name, size_t *name_len)
 {
        static struct string_list *mail_map;
        if (!mail_map) {
@@ -603,36 +698,26 @@ static size_t format_person_part(struct strbuf *sb, char part,
        const int placeholder_len = 2;
        int tz;
        unsigned long date = 0;
-       char person_name[1024];
-       char person_mail[1024];
        struct ident_split s;
-       const char *name_start, *name_end, *mail_start, *mail_end;
+       const char *name, *mail;
+       size_t maillen, namelen;
 
        if (split_ident_line(&s, msg, len) < 0)
                goto skip;
 
-       name_start = s.name_begin;
-       name_end = s.name_end;
-       mail_start = s.mail_begin;
-       mail_end = s.mail_end;
-
-       if (part == 'N' || part == 'E') { /* mailmap lookup */
-               snprintf(person_name, sizeof(person_name), "%.*s",
-                        (int)(name_end - name_start), name_start);
-               snprintf(person_mail, sizeof(person_mail), "%.*s",
-                        (int)(mail_end - mail_start), mail_start);
-               mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
-               name_start = person_name;
-               name_end = name_start + strlen(person_name);
-               mail_start = person_mail;
-               mail_end = mail_start +  strlen(person_mail);
-       }
+       name = s.name_begin;
+       namelen = s.name_end - s.name_begin;
+       mail = s.mail_begin;
+       maillen = s.mail_end - s.mail_begin;
+
+       if (part == 'N' || part == 'E') /* mailmap lookup */
+               mailmap_name(&mail, &maillen, &name, &namelen);
        if (part == 'n' || part == 'N') {       /* name */
-               strbuf_add(sb, name_start, name_end-name_start);
+               strbuf_add(sb, name, namelen);
                return placeholder_len;
        }
        if (part == 'e' || part == 'E') {       /* email */
-               strbuf_add(sb, mail_start, mail_end-mail_start);
+               strbuf_add(sb, mail, maillen);
                return placeholder_len;
        }
 
@@ -689,19 +774,14 @@ struct format_commit_context {
        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;
        char *message;
+       char *commit_encoding;
        size_t width, indent1, indent2;
 
        /* 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;
@@ -748,9 +828,6 @@ 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;
        }
@@ -877,59 +954,6 @@ 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,
@@ -960,12 +984,19 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
        switch (placeholder[0]) {
        case 'C':
                if (placeholder[1] == '(') {
-                       const char *end = strchr(placeholder + 2, ')');
+                       const char *begin = placeholder + 2;
+                       const char *end = strchr(begin, ')');
                        char color[COLOR_MAXLEN];
+
                        if (!end)
                                return 0;
-                       color_parse_mem(placeholder + 2,
-                                       end - (placeholder + 2),
+                       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;
@@ -1108,23 +1139,29 @@ 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;
@@ -1145,7 +1182,8 @@ 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 */
@@ -1256,17 +1294,17 @@ void format_commit_message(const struct commit *commit,
        context.commit = commit;
        context.pretty_ctx = pretty_ctx;
        context.wrap_start = sb->len;
-       context.message = logmsg_reencode(commit, output_enc);
-       if (!context.message)
-               context.message = commit->buffer;
+       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);
+       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,
@@ -1294,7 +1332,7 @@ 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;
@@ -1318,11 +1356,11 @@ 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);
@@ -1411,7 +1449,7 @@ 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;
@@ -1422,10 +1460,7 @@ 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;
@@ -1482,7 +1517,7 @@ 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,