sha1_file.c: cleanup hdr usage
[gitweb.git] / commit.c
index 946615d2ad5c364fe63b201aa9a9574737804c38..8d279b0b635f58482b5a45c6a8a0086744b0735d 100644 (file)
--- a/commit.c
+++ b/commit.c
@@ -1,21 +1,23 @@
 #include "cache.h"
 #include "tag.h"
 #include "commit.h"
+#include "pkt-line.h"
+#include "utf8.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.
-         */
+        * 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.
-         */
+        * reference to original list item that we will re-use
+        * on output.
+        */
        struct commit_list * list_item;
 
 };
@@ -45,7 +47,8 @@ enum cmit_fmt get_commit_format(const char *arg)
        if (*arg == '=')
                arg++;
        for (i = 0; i < ARRAY_SIZE(cmt_fmts); i++) {
-               if (!strncmp(arg, cmt_fmts[i].n, cmt_fmts[i].cmp_len))
+               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;
        }
 
@@ -56,7 +59,7 @@ static struct commit *check_commit(struct object *obj,
                                   const unsigned char *sha1,
                                   int quiet)
 {
-       if (obj->type != TYPE_COMMIT) {
+       if (obj->type != OBJ_COMMIT) {
                if (!quiet)
                        error("Object %s is a %s, not a commit",
                              sha1_to_hex(sha1), typename(obj->type));
@@ -86,11 +89,11 @@ struct commit *lookup_commit(const unsigned char *sha1)
        if (!obj) {
                struct commit *ret = alloc_commit_node();
                created_object(sha1, &ret->object);
-               ret->object.type = TYPE_COMMIT;
+               ret->object.type = OBJ_COMMIT;
                return ret;
        }
        if (!obj->type)
-               obj->type = TYPE_COMMIT;
+               obj->type = OBJ_COMMIT;
        return check_commit(obj, sha1, 0);
 }
 
@@ -123,7 +126,7 @@ static int commit_graft_pos(const unsigned char *sha1)
        while (lo < hi) {
                int mi = (lo + hi) / 2;
                struct commit_graft *graft = commit_graft[mi];
-               int cmp = memcmp(sha1, graft->sha1, 20);
+               int cmp = hashcmp(sha1, graft->sha1);
                if (!cmp)
                        return mi;
                if (cmp < 0)
@@ -221,6 +224,8 @@ static void prepare_commit_graft(void)
                return;
        graft_file = get_graft_file();
        read_graft_file(graft_file);
+       /* make sure shallows are read */
+       is_repository_shallow();
        commit_graft_prepared = 1;
 }
 
@@ -234,8 +239,42 @@ static struct commit_graft *lookup_commit_graft(const unsigned char *sha1)
        return commit_graft[pos];
 }
 
+int write_shallow_commits(int fd, int use_pack_protocol)
+{
+       int i, count = 0;
+       for (i = 0; i < commit_graft_nr; i++)
+               if (commit_graft[i]->nr_parent < 0) {
+                       const char *hex =
+                               sha1_to_hex(commit_graft[i]->sha1);
+                       count++;
+                       if (use_pack_protocol)
+                               packet_write(fd, "shallow %s", hex);
+                       else {
+                               if (write_in_full(fd, hex,  40) != 40)
+                                       break;
+                               if (write_in_full(fd, "\n", 1) != 1)
+                                       break;
+                       }
+               }
+       return count;
+}
+
+int unregister_shallow(const unsigned char *sha1)
+{
+       int pos = commit_graft_pos(sha1);
+       if (pos < 0)
+               return -1;
+       if (pos + 1 < commit_graft_nr)
+               memcpy(commit_graft + pos, commit_graft + pos + 1,
+                               sizeof(struct commit_graft *)
+                               * (commit_graft_nr - pos - 1));
+       commit_graft_nr--;
+       return 0;
+}
+
 int parse_commit_buffer(struct commit *item, void *buffer, unsigned long size)
 {
+       char *tail = buffer;
        char *bufptr = buffer;
        unsigned char parent[20];
        struct commit_list **pptr;
@@ -245,9 +284,10 @@ int parse_commit_buffer(struct commit *item, void *buffer, unsigned long size)
        if (item->object.parsed)
                return 0;
        item->object.parsed = 1;
-       if (memcmp(bufptr, "tree ", 5))
+       tail += size;
+       if (tail <= bufptr + 5 || memcmp(bufptr, "tree ", 5))
                return error("bogus commit object %s", sha1_to_hex(item->object.sha1));
-       if (get_sha1_hex(bufptr + 5, parent) < 0)
+       if (tail <= bufptr + 45 || 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);
@@ -257,10 +297,12 @@ int parse_commit_buffer(struct commit *item, void *buffer, unsigned long size)
        pptr = &item->parents;
 
        graft = lookup_commit_graft(item->object.sha1);
-       while (!memcmp(bufptr, "parent ", 7)) {
+       while (bufptr + 48 < tail && !memcmp(bufptr, "parent ", 7)) {
                struct commit *new_parent;
 
-               if (get_sha1_hex(bufptr + 7, parent) || bufptr[47] != '\n')
+               if (tail <= bufptr + 48 ||
+                   get_sha1_hex(bufptr + 7, parent) ||
+                   bufptr[47] != '\n')
                        return error("bad parents in commit %s", sha1_to_hex(item->object.sha1));
                bufptr += 48;
                if (graft)
@@ -393,12 +435,13 @@ void clear_commit_marks(struct commit *commit, unsigned int mark)
 {
        struct commit_list *parents;
 
-       parents = commit->parents;
        commit->object.flags &= ~mark;
+       parents = commit->parents;
        while (parents) {
                struct commit *parent = parents->item;
-               if (parent && parent->object.parsed &&
-                   (parent->object.flags & mark))
+
+               /* Have we already cleared this? */
+               if (mark & parent->object.flags)
                        clear_commit_marks(parent, mark);
                parents = parents->next;
        }
@@ -422,20 +465,29 @@ static int get_one_line(const char *msg, unsigned long len)
        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 ((ch & 0x80) || (ch == '=') || (ch == '?') || (ch == '_'));
+       return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 }
 
-static int add_rfc2047(char *buf, const char *line, int len)
+static int add_rfc2047(char *buf, const char *line, int len,
+                      const char *encoding)
 {
        char *bp = buf;
        int i, needquote;
-       static const char q_utf8[] = "=?utf-8?q?";
+       char q_encoding[128];
+       const char *q_encoding_fmt = "=?%s?q?";
 
        for (i = needquote = 0; !needquote && i < len; i++) {
-               unsigned ch = line[i];
-               if (ch & 0x80)
+               int ch = line[i];
+               if (non_ascii(ch))
                        needquote++;
                if ((i + 1 < len) &&
                    (ch == '=' && line[i+1] == '?'))
@@ -444,8 +496,11 @@ static int add_rfc2047(char *buf, const char *line, int len)
        if (!needquote)
                return sprintf(buf, "%.*s", len, line);
 
-       memcpy(bp, q_utf8, sizeof(q_utf8)-1);
-       bp += sizeof(q_utf8)-1;
+       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;
                if (is_rfc2047_special(ch)) {
@@ -462,7 +517,9 @@ static int add_rfc2047(char *buf, const char *line, int len)
        return bp - buf;
 }
 
-static int add_user_info(const char *what, enum cmit_fmt fmt, char *buf, const char *line)
+static int add_user_info(const char *what, enum cmit_fmt fmt, char *buf,
+                        const char *line, int relative_date,
+                        const char *encoding)
 {
        char *date;
        int namelen;
@@ -490,7 +547,8 @@ static int add_user_info(const char *what, enum cmit_fmt fmt, char *buf, const c
                filler = "";
                strcpy(buf, "From: ");
                ret = strlen(buf);
-               ret += add_rfc2047(buf + ret, line, display_name_length);
+               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';
@@ -502,14 +560,16 @@ static int add_user_info(const char *what, enum cmit_fmt fmt, char *buf, const c
        }
        switch (fmt) {
        case CMIT_FMT_MEDIUM:
-               ret += sprintf(buf + ret, "Date:   %s\n", show_date(time, tz));
+               ret += sprintf(buf + ret, "Date:   %s\n",
+                              show_date(time, tz, relative_date));
                break;
        case CMIT_FMT_EMAIL:
                ret += sprintf(buf + ret, "Date: %s\n",
                               show_rfc2822_date(time, tz));
                break;
        case CMIT_FMT_FULLER:
-               ret += sprintf(buf + ret, "%sDate: %s\n", what, show_date(time, tz));
+               ret += sprintf(buf + ret, "%sDate: %s\n", what,
+                              show_date(time, tz, relative_date));
                break;
        default:
                /* notin' */
@@ -540,10 +600,13 @@ static int add_merge_info(enum cmit_fmt fmt, char *buf, const struct commit *com
 
        while (parent) {
                struct commit *p = parent->item;
-               const char *hex = abbrev
-                       ? find_unique_abbrev(p->object.sha1, abbrev)
-                       : sha1_to_hex(p->object.sha1);
-               char *dots = (abbrev && strlen(hex) != 40) ? "..." : "";
+               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);
@@ -552,14 +615,126 @@ static int add_merge_info(enum cmit_fmt fmt, char *buf, const struct commit *com
        return offset;
 }
 
-unsigned long pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit, unsigned long len, char *buf, unsigned long space, int abbrev, const char *subject, const char *after_subject)
+static char *get_header(const struct commit *commit, const char *key)
 {
-       int hdr = 1, body = 0;
+       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 (!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, char *encoding)
+{
+       char *encoding_header = strstr(buf, "\nencoding ");
+       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 (!encoding_header)
+               return buf; /* should not happen but be defensive */
+       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,
+                            char *output_encoding)
+{
+       char *encoding;
+       char *out;
+       char *utf8 = "utf-8";
+
+       if (!*output_encoding)
+               return NULL;
+       encoding = get_header(commit, "encoding");
+       if (!encoding)
+               encoding = utf8;
+       if (!strcmp(encoding, output_encoding))
+               out = strdup(commit->buffer);
+       else
+               out = reencode_string(commit->buffer,
+                                     output_encoding, encoding);
+       if (out)
+               out = replace_encoding_header(out, output_encoding);
+
+       if (encoding != utf8)
+               free(encoding);
+       if (!out)
+               return NULL;
+       return out;
+}
+
+unsigned long pretty_print_commit(enum cmit_fmt fmt,
+                                 const struct commit *commit,
+                                 unsigned long len,
+                                 char *buf, unsigned long space,
+                                 int abbrev, const char *subject,
+                                 const char *after_subject,
+                                 int relative_date)
+{
+       int hdr = 1, body = 0, seen_title = 0;
        unsigned long offset = 0;
        int indent = 4;
        int parents_shown = 0;
        const char *msg = commit->buffer;
        int plain_non_ascii = 0;
+       char *reencoded;
+       char *encoding;
+
+       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;
 
        if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
                indent = 0;
@@ -576,14 +751,14 @@ unsigned long pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit
                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 so; skip over the
+                                * 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 (ch & 0x80) {
+                       else if (non_ascii(ch)) {
                                plain_non_ascii = 1;
                                break;
                        }
@@ -641,31 +816,40 @@ unsigned long pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit
                        if (!memcmp(line, "author ", 7))
                                offset += add_user_info("Author", fmt,
                                                        buf + offset,
-                                                       line + 7);
+                                                       line + 7,
+                                                       relative_date,
+                                                       encoding);
                        if (!memcmp(line, "committer ", 10) &&
                            (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER))
                                offset += add_user_info("Commit", fmt,
                                                        buf + offset,
-                                                       line + 10);
+                                                       line + 10,
+                                                       relative_date,
+                                                       encoding);
                        continue;
                }
 
+               if (!subject)
+                       body = 1;
+
                if (is_empty_line(line, &linelen)) {
+                       if (!seen_title)
+                               continue;
                        if (!body)
                                continue;
                        if (subject)
                                continue;
                        if (fmt == CMIT_FMT_SHORT)
                                break;
-               } else {
-                       body = 1;
                }
 
+               seen_title = 1;
                if (subject) {
                        int slen = strlen(subject);
                        memcpy(buf + offset, subject, slen);
                        offset += slen;
-                       offset += add_rfc2047(buf + offset, line, linelen);
+                       offset += add_rfc2047(buf + offset, line, linelen,
+                                             encoding);
                }
                else {
                        memset(buf + offset, ' ', indent);
@@ -676,11 +860,17 @@ unsigned long pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit
                if (fmt == CMIT_FMT_ONELINE)
                        break;
                if (subject && plain_non_ascii) {
-                       static const char header[] =
-                               "Content-Type: text/plain; charset=UTF-8\n"
+                       int sz;
+                       char header[512];
+                       const char *header_fmt =
+                               "Content-Type: text/plain; charset=%s\n"
                                "Content-Transfer-Encoding: 8bit\n";
-                       memcpy(buf + offset, header, sizeof(header)-1);
-                       offset += sizeof(header)-1;
+                       sz = snprintf(header, sizeof(header), header_fmt,
+                                     encoding);
+                       if (sizeof(header) < sz)
+                               die("Encoding name %s too long", encoding);
+                       memcpy(buf + offset, header, sz);
+                       offset += sz;
                }
                if (after_subject) {
                        int slen = strlen(after_subject);
@@ -697,7 +887,15 @@ unsigned long pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit
        /* Make sure there is an EOLN for the non-oneline case */
        if (fmt != CMIT_FMT_ONELINE)
                buf[offset++] = '\n';
+       /*
+        * make sure there is another EOLN to separate the headers from whatever
+        * body the caller appends if we haven't already written a body
+        */
+       if (fmt == CMIT_FMT_EMAIL && !body)
+               buf[offset++] = '\n';
        buf[offset] = '\0';
+
+       free(reencoded);
        return offset;
 }
 
@@ -715,10 +913,10 @@ struct commit *pop_commit(struct commit_list **stack)
 
 int count_parents(struct commit * commit)
 {
-        int count = 0;
+        int count;
         struct commit_list * parents = commit->parents;
-        for (count=0;parents; parents=parents->next,count++)
-          ;
+        for (count = 0; parents; parents = parents->next,count++)
+               ;
         return count;
 }
 
@@ -842,3 +1040,169 @@ void sort_in_topological_order_fn(struct commit_list ** list, int lifo,
        }
        free(nodes);
 }
+
+/* merge-base stuff */
+
+/* bits #0..15 in revision.h */
+#define PARENT1                (1u<<16)
+#define PARENT2                (1u<<17)
+#define STALE          (1u<<18)
+#define RESULT         (1u<<19)
+
+static const unsigned all_flags = (PARENT1 | PARENT2 | STALE | RESULT);
+
+static struct commit *interesting(struct commit_list *list)
+{
+       while (list) {
+               struct commit *commit = list->item;
+               list = list->next;
+               if (commit->object.flags & STALE)
+                       continue;
+               return commit;
+       }
+       return NULL;
+}
+
+static struct commit_list *merge_bases(struct commit *one, struct commit *two)
+{
+       struct commit_list *list = NULL;
+       struct commit_list *result = NULL;
+
+       if (one == two)
+               /* We do not mark this even with RESULT so we do not
+                * have to clean it up.
+                */
+               return commit_list_insert(one, &result);
+
+       parse_commit(one);
+       parse_commit(two);
+
+       one->object.flags |= PARENT1;
+       two->object.flags |= PARENT2;
+       insert_by_date(one, &list);
+       insert_by_date(two, &list);
+
+       while (interesting(list)) {
+               struct commit *commit;
+               struct commit_list *parents;
+               struct commit_list *n;
+               int flags;
+
+               commit = list->item;
+               n = list->next;
+               free(list);
+               list = n;
+
+               flags = commit->object.flags & (PARENT1 | PARENT2 | STALE);
+               if (flags == (PARENT1 | PARENT2)) {
+                       if (!(commit->object.flags & RESULT)) {
+                               commit->object.flags |= RESULT;
+                               insert_by_date(commit, &result);
+                       }
+                       /* Mark parents of a found merge stale */
+                       flags |= STALE;
+               }
+               parents = commit->parents;
+               while (parents) {
+                       struct commit *p = parents->item;
+                       parents = parents->next;
+                       if ((p->object.flags & flags) == flags)
+                               continue;
+                       parse_commit(p);
+                       p->object.flags |= flags;
+                       insert_by_date(p, &list);
+               }
+       }
+
+       /* Clean up the result to remove stale ones */
+       free_commit_list(list);
+       list = result; result = NULL;
+       while (list) {
+               struct commit_list *n = list->next;
+               if (!(list->item->object.flags & STALE))
+                       insert_by_date(list->item, &result);
+               free(list);
+               list = n;
+       }
+       return result;
+}
+
+struct commit_list *get_merge_bases(struct commit *one,
+                                   struct commit *two,
+                                    int cleanup)
+{
+       struct commit_list *list;
+       struct commit **rslt;
+       struct commit_list *result;
+       int cnt, i, j;
+
+       result = merge_bases(one, two);
+       if (one == two)
+               return result;
+       if (!result || !result->next) {
+               if (cleanup) {
+                       clear_commit_marks(one, all_flags);
+                       clear_commit_marks(two, all_flags);
+               }
+               return result;
+       }
+
+       /* There are more than one */
+       cnt = 0;
+       list = result;
+       while (list) {
+               list = list->next;
+               cnt++;
+       }
+       rslt = xcalloc(cnt, sizeof(*rslt));
+       for (list = result, i = 0; list; list = list->next)
+               rslt[i++] = list->item;
+       free_commit_list(result);
+
+       clear_commit_marks(one, all_flags);
+       clear_commit_marks(two, all_flags);
+       for (i = 0; i < cnt - 1; i++) {
+               for (j = i+1; j < cnt; j++) {
+                       if (!rslt[i] || !rslt[j])
+                               continue;
+                       result = merge_bases(rslt[i], rslt[j]);
+                       clear_commit_marks(rslt[i], all_flags);
+                       clear_commit_marks(rslt[j], all_flags);
+                       for (list = result; list; list = list->next) {
+                               if (rslt[i] == list->item)
+                                       rslt[i] = NULL;
+                               if (rslt[j] == list->item)
+                                       rslt[j] = NULL;
+                       }
+               }
+       }
+
+       /* Surviving ones in rslt[] are the independent results */
+       result = NULL;
+       for (i = 0; i < cnt; i++) {
+               if (rslt[i])
+                       insert_by_date(rslt[i], &result);
+       }
+       free(rslt);
+       return result;
+}
+
+int in_merge_bases(struct commit *commit, struct commit **reference, int num)
+{
+       struct commit_list *bases, *b;
+       int ret = 0;
+
+       if (num == 1)
+               bases = get_merge_bases(commit, *reference, 1);
+       else
+               die("not yet");
+       for (b = bases; b; b = b->next) {
+               if (!hashcmp(commit->object.sha1, b->item->object.sha1)) {
+                       ret = 1;
+                       break;
+               }
+       }
+
+       free_commit_list(bases);
+       return ret;
+}