fast-import: Introduce --import-marks-if-exists
[gitweb.git] / sha1_name.c
index 7013c53ca6f083b6887f2d8567c72c5a63054c1c..709ff2eee64cf106191ad274bede82a95d00e2a3 100644 (file)
@@ -5,6 +5,9 @@
 #include "blob.h"
 #include "tree-walk.h"
 #include "refs.h"
+#include "remote.h"
+
+static int get_sha1_oneline(const char *, unsigned char *, struct commit_list *);
 
 static int find_short_object_filename(int len, const char *name, unsigned char *sha1)
 {
@@ -205,7 +208,9 @@ const char *find_unique_abbrev(const unsigned char *sha1, int len)
                if (exists
                    ? !status
                    : status == SHORT_NAME_NOT_FOUND) {
-                       hex[len] = 0;
+                       int cut_at = len + unique_abbrev_extra_length;
+                       cut_at = (cut_at < 40) ? cut_at : 40;
+                       hex[cut_at] = 0;
                        return hex;
                }
                len++;
@@ -240,7 +245,8 @@ static int ambiguous_path(const char *path, int len)
 
 /*
  * *string and *len will only be substituted, and *string returned (for
- * later free()ing) if the string passed in is of the form @{-<n>}.
+ * later free()ing) if the string passed in is a magic short-hand form
+ * to name a branch.
  */
 static char *substitute_branch_name(const char **string, int *len)
 {
@@ -322,11 +328,25 @@ int dwim_log(const char *str, int len, unsigned char *sha1, char **log)
        return logs_found;
 }
 
+static inline int upstream_mark(const char *string, int len)
+{
+       const char *suffix[] = { "@{upstream}", "@{u}" };
+       int i;
+
+       for (i = 0; i < ARRAY_SIZE(suffix); i++) {
+               int suffix_len = strlen(suffix[i]);
+               if (suffix_len <= len
+                   && !memcmp(string, suffix[i], suffix_len))
+                       return suffix_len;
+       }
+       return 0;
+}
+
 static int get_sha1_1(const char *name, int len, unsigned char *sha1);
 
 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
 {
-       static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
+       static const char *warn_msg = "refname '%.*s' is ambiguous.";
        char *real_ref = NULL;
        int refs_found = 0;
        int at, reflog_len;
@@ -339,8 +359,10 @@ static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
        if (len && str[len-1] == '}') {
                for (at = len-2; at >= 0; at--) {
                        if (str[at] == '@' && str[at+1] == '{') {
-                               reflog_len = (len-1) - (at+2);
-                               len = at;
+                               if (!upstream_mark(str + at, len - at)) {
+                                       reflog_len = (len-1) - (at+2);
+                                       len = at;
+                               }
                                break;
                        }
                }
@@ -372,7 +394,7 @@ static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
                return -1;
 
        if (warn_ambiguous_refs && refs_found > 1)
-               fprintf(stderr, warning, len, str);
+               warning(warn_msg, len, str);
 
        if (reflog_len) {
                int nth, i;
@@ -380,6 +402,10 @@ static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
                unsigned long co_time;
                int co_tz, co_cnt;
 
+               /* a @{-N} placed anywhere except the start is an error */
+               if (str[at+2] == '-')
+                       return -1;
+
                /* Is it asking for N-th entry, or approxidate? */
                for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
                        char ch = str[at+2+i];
@@ -404,14 +430,14 @@ static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
                if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
                                &co_time, &co_tz, &co_cnt)) {
                        if (at_time)
-                               fprintf(stderr,
-                                       "warning: Log for '%.*s' only goes "
-                                       "back to %s.\n", len, str,
+                               warning("Log for '%.*s' only goes "
+                                       "back to %s.", len, str,
                                        show_date(co_time, co_tz, DATE_RFC2822));
-                       else
-                               fprintf(stderr,
-                                       "warning: Log for '%.*s' only has "
-                                       "%d entries.\n", len, str, co_cnt);
+                       else {
+                               free(real_ref);
+                               die("Log for '%.*s' only has %d entries.",
+                                   len, str, co_cnt);
+                       }
                }
        }
 
@@ -538,6 +564,8 @@ static int peel_onion(const char *name, int len, unsigned char *sha1)
                expected_type = OBJ_BLOB;
        else if (sp[0] == '}')
                expected_type = OBJ_NONE;
+       else if (sp[0] == '/')
+               expected_type = OBJ_COMMIT;
        else
                return -1;
 
@@ -552,19 +580,37 @@ static int peel_onion(const char *name, int len, unsigned char *sha1)
                if (!o || (!o->parsed && !parse_object(o->sha1)))
                        return -1;
                hashcpy(sha1, o->sha1);
+               return 0;
        }
-       else {
+
+       /*
+        * At this point, the syntax look correct, so
+        * if we do not get the needed object, we should
+        * barf.
+        */
+       o = peel_to_type(name, len, o, expected_type);
+       if (!o)
+               return -1;
+
+       hashcpy(sha1, o->sha1);
+       if (sp[0] == '/') {
+               /* "$commit^{/foo}" */
+               char *prefix;
+               int ret;
+               struct commit_list *list = NULL;
+
                /*
-                * At this point, the syntax look correct, so
-                * if we do not get the needed object, we should
-                * barf.
+                * $commit^{/}. Some regex implementation may reject.
+                * We don't need regex anyway. '' pattern always matches.
                 */
-               o = peel_to_type(name, len, o, expected_type);
-               if (o) {
-                       hashcpy(sha1, o->sha1);
+               if (sp[1] == '}')
                        return 0;
-               }
-               return -1;
+
+               prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
+               commit_list_insert((struct commit *)o, &list);
+               ret = get_sha1_oneline(prefix, sha1, list);
+               free(prefix);
+               return ret;
        }
        return 0;
 }
@@ -637,6 +683,16 @@ static int get_sha1_1(const char *name, int len, unsigned char *sha1)
        return get_short_sha1(name, len, sha1, 0);
 }
 
+/*
+ * This interprets names like ':/Initial revision of "git"' by searching
+ * through history and returning the first commit whose message starts
+ * the given regular expression.
+ *
+ * For future extension, ':/!' is reserved. If you want to match a message
+ * beginning with a '!', you have to repeat the exclamation mark.
+ */
+#define ONELINE_SEEN (1u<<20)
+
 static int handle_one_ref(const char *path,
                const unsigned char *sha1, int flag, void *cb_data)
 {
@@ -651,65 +707,65 @@ static int handle_one_ref(const char *path,
        }
        if (object->type != OBJ_COMMIT)
                return 0;
-       insert_by_date((struct commit *)object, list);
+       commit_list_insert_by_date((struct commit *)object, list);
        return 0;
 }
 
-/*
- * This interprets names like ':/Initial revision of "git"' by searching
- * through history and returning the first commit whose message starts
- * with the given string.
- *
- * For future extension, ':/!' is reserved. If you want to match a message
- * beginning with a '!', you have to repeat the exclamation mark.
- */
-
-#define ONELINE_SEEN (1u<<20)
-static int get_sha1_oneline(const char *prefix, unsigned char *sha1)
+static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
+                           struct commit_list *list)
 {
-       struct commit_list *list = NULL, *backup = NULL, *l;
-       int retval = -1;
-       char *temp_commit_buffer = NULL;
+       struct commit_list *backup = NULL, *l;
+       int found = 0;
+       regex_t regex;
 
        if (prefix[0] == '!') {
                if (prefix[1] != '!')
                        die ("Invalid search pattern: %s", prefix);
                prefix++;
        }
-       for_each_ref(handle_one_ref, &list);
-       for (l = list; l; l = l->next)
+
+       if (regcomp(&regex, prefix, REG_EXTENDED))
+               die("Invalid search pattern: %s", prefix);
+
+       for (l = list; l; l = l->next) {
+               l->item->object.flags |= ONELINE_SEEN;
                commit_list_insert(l->item, &backup);
+       }
        while (list) {
-               char *p;
+               char *p, *to_free = NULL;
                struct commit *commit;
                enum object_type type;
                unsigned long size;
+               int matches;
 
                commit = pop_most_recent_commit(&list, ONELINE_SEEN);
                if (!parse_object(commit->object.sha1))
                        continue;
-               free(temp_commit_buffer);
                if (commit->buffer)
                        p = commit->buffer;
                else {
                        p = read_sha1_file(commit->object.sha1, &type, &size);
                        if (!p)
                                continue;
-                       temp_commit_buffer = p;
+                       to_free = p;
                }
-               if (!(p = strstr(p, "\n\n")))
-                       continue;
-               if (!prefixcmp(p + 2, prefix)) {
+
+               p = strstr(p, "\n\n");
+               matches = p && !regexec(&regex, p + 2, 0, NULL, 0);
+               free(to_free);
+
+               if (matches) {
                        hashcpy(sha1, commit->object.sha1);
-                       retval = 0;
+                       found = 1;
                        break;
                }
        }
-       free(temp_commit_buffer);
+       regfree(&regex);
        free_commit_list(list);
        for (l = backup; l; l = l->next)
                clear_commit_marks(l->item, ONELINE_SEEN);
-       return retval;
+       free_commit_list(backup);
+       return found ? 0 : -1;
 }
 
 struct grab_nth_branch_switch_cbdata {
@@ -742,17 +798,10 @@ static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
 }
 
 /*
- * This reads "@{-N}" syntax, finds the name of the Nth previous
- * branch we were on, and places the name of the branch in the given
- * buf and returns the number of characters parsed if successful.
- *
- * If the input is not of the accepted format, it returns a negative
- * number to signal an error.
- *
- * If the input was ok but there are not N branch switches in the
- * reflog, it returns 0.
+ * Parse @{-N} syntax, return the number of characters parsed
+ * if successful; otherwise signal an error with negative value.
  */
-int interpret_branch_name(const char *name, struct strbuf *buf)
+static int interpret_nth_prior_checkout(const char *name, struct strbuf *buf)
 {
        long nth;
        int i, retval;
@@ -796,36 +845,298 @@ int interpret_branch_name(const char *name, struct strbuf *buf)
        return retval;
 }
 
+int get_sha1_mb(const char *name, unsigned char *sha1)
+{
+       struct commit *one, *two;
+       struct commit_list *mbs;
+       unsigned char sha1_tmp[20];
+       const char *dots;
+       int st;
+
+       dots = strstr(name, "...");
+       if (!dots)
+               return get_sha1(name, sha1);
+       if (dots == name)
+               st = get_sha1("HEAD", sha1_tmp);
+       else {
+               struct strbuf sb;
+               strbuf_init(&sb, dots - name);
+               strbuf_add(&sb, name, dots - name);
+               st = get_sha1(sb.buf, sha1_tmp);
+               strbuf_release(&sb);
+       }
+       if (st)
+               return st;
+       one = lookup_commit_reference_gently(sha1_tmp, 0);
+       if (!one)
+               return -1;
+
+       if (get_sha1(dots[3] ? (dots + 3) : "HEAD", sha1_tmp))
+               return -1;
+       two = lookup_commit_reference_gently(sha1_tmp, 0);
+       if (!two)
+               return -1;
+       mbs = get_merge_bases(one, two, 1);
+       if (!mbs || mbs->next)
+               st = -1;
+       else {
+               st = 0;
+               hashcpy(sha1, mbs->item->object.sha1);
+       }
+       free_commit_list(mbs);
+       return st;
+}
+
+/*
+ * This reads short-hand syntax that not only evaluates to a commit
+ * object name, but also can act as if the end user spelled the name
+ * of the branch from the command line.
+ *
+ * - "@{-N}" finds the name of the Nth previous branch we were on, and
+ *   places the name of the branch in the given buf and returns the
+ *   number of characters parsed if successful.
+ *
+ * - "<branch>@{upstream}" finds the name of the other ref that
+ *   <branch> is configured to merge with (missing <branch> defaults
+ *   to the current branch), and places the name of the branch in the
+ *   given buf and returns the number of characters parsed if
+ *   successful.
+ *
+ * If the input is not of the accepted format, it returns a negative
+ * number to signal an error.
+ *
+ * If the input was ok but there are not N branch switches in the
+ * reflog, it returns 0.
+ */
+int interpret_branch_name(const char *name, struct strbuf *buf)
+{
+       char *cp;
+       struct branch *upstream;
+       int namelen = strlen(name);
+       int len = interpret_nth_prior_checkout(name, buf);
+       int tmp_len;
+
+       if (!len)
+               return len; /* syntax Ok, not enough switches */
+       if (0 < len && len == namelen)
+               return len; /* consumed all */
+       else if (0 < len) {
+               /* we have extra data, which might need further processing */
+               struct strbuf tmp = STRBUF_INIT;
+               int used = buf->len;
+               int ret;
+
+               strbuf_add(buf, name + len, namelen - len);
+               ret = interpret_branch_name(buf->buf, &tmp);
+               /* that data was not interpreted, remove our cruft */
+               if (ret < 0) {
+                       strbuf_setlen(buf, used);
+                       return len;
+               }
+               strbuf_reset(buf);
+               strbuf_addbuf(buf, &tmp);
+               strbuf_release(&tmp);
+               /* tweak for size of {-N} versus expanded ref name */
+               return ret - used + len;
+       }
+
+       cp = strchr(name, '@');
+       if (!cp)
+               return -1;
+       tmp_len = upstream_mark(cp, namelen - (cp - name));
+       if (!tmp_len)
+               return -1;
+       len = cp + tmp_len - name;
+       cp = xstrndup(name, cp - name);
+       upstream = branch_get(*cp ? cp : NULL);
+       if (!upstream
+           || !upstream->merge
+           || !upstream->merge[0]->dst)
+               return error("No upstream branch found for '%s'", cp);
+       free(cp);
+       cp = shorten_unambiguous_ref(upstream->merge[0]->dst, 0);
+       strbuf_reset(buf);
+       strbuf_addstr(buf, cp);
+       free(cp);
+       return len;
+}
+
+int strbuf_branchname(struct strbuf *sb, const char *name)
+{
+       int len = strlen(name);
+       if (interpret_branch_name(name, sb) == len)
+               return 0;
+       strbuf_add(sb, name, len);
+       return len;
+}
+
+int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
+{
+       strbuf_branchname(sb, name);
+       if (name[0] == '-')
+               return CHECK_REF_FORMAT_ERROR;
+       strbuf_splice(sb, 0, 0, "refs/heads/", 11);
+       return check_ref_format(sb->buf);
+}
+
 /*
  * This is like "get_sha1_basic()", except it allows "sha1 expressions",
  * notably "xyz^" for "parent of xyz"
  */
 int get_sha1(const char *name, unsigned char *sha1)
 {
-       unsigned unused;
-       return get_sha1_with_mode(name, sha1, &unused);
+       struct object_context unused;
+       return get_sha1_with_context(name, sha1, &unused);
 }
 
-int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
+/* Must be called only when object_name:filename doesn't exist. */
+static void diagnose_invalid_sha1_path(const char *prefix,
+                                      const char *filename,
+                                      const unsigned char *tree_sha1,
+                                      const char *object_name)
+{
+       struct stat st;
+       unsigned char sha1[20];
+       unsigned mode;
+
+       if (!prefix)
+               prefix = "";
+
+       if (!lstat(filename, &st))
+               die("Path '%s' exists on disk, but not in '%s'.",
+                   filename, object_name);
+       if (errno == ENOENT || errno == ENOTDIR) {
+               char *fullname = xmalloc(strlen(filename)
+                                            + strlen(prefix) + 1);
+               strcpy(fullname, prefix);
+               strcat(fullname, filename);
+
+               if (!get_tree_entry(tree_sha1, fullname,
+                                   sha1, &mode)) {
+                       die("Path '%s' exists, but not '%s'.\n"
+                           "Did you mean '%s:%s'?",
+                           fullname,
+                           filename,
+                           object_name,
+                           fullname);
+               }
+               die("Path '%s' does not exist in '%s'",
+                   filename, object_name);
+       }
+}
+
+/* Must be called only when :stage:filename doesn't exist. */
+static void diagnose_invalid_index_path(int stage,
+                                       const char *prefix,
+                                       const char *filename)
+{
+       struct stat st;
+       struct cache_entry *ce;
+       int pos;
+       unsigned namelen = strlen(filename);
+       unsigned fullnamelen;
+       char *fullname;
+
+       if (!prefix)
+               prefix = "";
+
+       /* Wrong stage number? */
+       pos = cache_name_pos(filename, namelen);
+       if (pos < 0)
+               pos = -pos - 1;
+       if (pos < active_nr) {
+               ce = active_cache[pos];
+               if (ce_namelen(ce) == namelen &&
+                   !memcmp(ce->name, filename, namelen))
+                       die("Path '%s' is in the index, but not at stage %d.\n"
+                           "Did you mean ':%d:%s'?",
+                           filename, stage,
+                           ce_stage(ce), filename);
+       }
+
+       /* Confusion between relative and absolute filenames? */
+       fullnamelen = namelen + strlen(prefix);
+       fullname = xmalloc(fullnamelen + 1);
+       strcpy(fullname, prefix);
+       strcat(fullname, filename);
+       pos = cache_name_pos(fullname, fullnamelen);
+       if (pos < 0)
+               pos = -pos - 1;
+       if (pos < active_nr) {
+               ce = active_cache[pos];
+               if (ce_namelen(ce) == fullnamelen &&
+                   !memcmp(ce->name, fullname, fullnamelen))
+                       die("Path '%s' is in the index, but not '%s'.\n"
+                           "Did you mean ':%d:%s'?",
+                           fullname, filename,
+                           ce_stage(ce), fullname);
+       }
+
+       if (!lstat(filename, &st))
+               die("Path '%s' exists on disk, but not in the index.", filename);
+       if (errno == ENOENT || errno == ENOTDIR)
+               die("Path '%s' does not exist (neither on disk nor in the index).",
+                   filename);
+
+       free(fullname);
+}
+
+
+int get_sha1_with_mode_1(const char *name, unsigned char *sha1, unsigned *mode, int gently, const char *prefix)
+{
+       struct object_context oc;
+       int ret;
+       ret = get_sha1_with_context_1(name, sha1, &oc, gently, prefix);
+       *mode = oc.mode;
+       return ret;
+}
+
+static char *resolve_relative_path(const char *rel)
+{
+       if (prefixcmp(rel, "./") && prefixcmp(rel, "../"))
+               return NULL;
+
+       if (!startup_info)
+               die("BUG: startup_info struct is not initialized.");
+
+       if (!is_inside_work_tree())
+               die("relative path syntax can't be used outside working tree.");
+
+       /* die() inside prefix_path() if resolved path is outside worktree */
+       return prefix_path(startup_info->prefix,
+                          startup_info->prefix ? strlen(startup_info->prefix) : 0,
+                          rel);
+}
+
+int get_sha1_with_context_1(const char *name, unsigned char *sha1,
+                           struct object_context *oc,
+                           int gently, const char *prefix)
 {
        int ret, bracket_depth;
        int namelen = strlen(name);
        const char *cp;
 
-       *mode = S_IFINVALID;
+       memset(oc, 0, sizeof(*oc));
+       oc->mode = S_IFINVALID;
        ret = get_sha1_1(name, namelen, sha1);
        if (!ret)
                return ret;
        /* sha1:path --> object name of path in ent sha1
-        * :path -> object name of path in index
+        * :path -> object name of absolute path in index
+        * :./path -> object name of path relative to cwd in index
         * :[0-3]:path -> object name of path in index at stage
+        * :/foo -> recent commit matching foo
         */
        if (name[0] == ':') {
                int stage = 0;
                struct cache_entry *ce;
+               char *new_path = NULL;
                int pos;
-               if (namelen > 2 && name[1] == '/')
-                       return get_sha1_oneline(name + 2, sha1);
+               if (namelen > 2 && name[1] == '/') {
+                       struct commit_list *list = NULL;
+                       for_each_ref(handle_one_ref, &list);
+                       return get_sha1_oneline(name + 2, sha1, list);
+               }
                if (namelen < 3 ||
                    name[2] != ':' ||
                    name[1] < '0' || '3' < name[1])
@@ -834,7 +1145,18 @@ int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
                        stage = name[1] - '0';
                        cp = name + 3;
                }
-               namelen = namelen - (cp - name);
+               new_path = resolve_relative_path(cp);
+               if (!new_path) {
+                       namelen = namelen - (cp - name);
+               } else {
+                       cp = new_path;
+                       namelen = strlen(cp);
+               }
+
+               strncpy(oc->path, cp,
+                       sizeof(oc->path));
+               oc->path[sizeof(oc->path)-1] = '\0';
+
                if (!active_cache)
                        read_cache();
                pos = cache_name_pos(cp, namelen);
@@ -847,11 +1169,15 @@ int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
                                break;
                        if (ce_stage(ce) == stage) {
                                hashcpy(sha1, ce->sha1);
-                               *mode = ce->ce_mode;
+                               oc->mode = ce->ce_mode;
+                               free(new_path);
                                return 0;
                        }
                        pos++;
                }
+               if (!gently)
+                       diagnose_invalid_index_path(stage, prefix, cp);
+               free(new_path);
                return -1;
        }
        for (cp = name, bracket_depth = 0; *cp; cp++) {
@@ -864,9 +1190,36 @@ int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
        }
        if (*cp == ':') {
                unsigned char tree_sha1[20];
-               if (!get_sha1_1(name, cp-name, tree_sha1))
-                       return get_tree_entry(tree_sha1, cp+1, sha1,
-                                             mode);
+               char *object_name = NULL;
+               if (!gently) {
+                       object_name = xmalloc(cp-name+1);
+                       strncpy(object_name, name, cp-name);
+                       object_name[cp-name] = '\0';
+               }
+               if (!get_sha1_1(name, cp-name, tree_sha1)) {
+                       const char *filename = cp+1;
+                       char *new_filename = NULL;
+
+                       new_filename = resolve_relative_path(filename);
+                       if (new_filename)
+                               filename = new_filename;
+                       ret = get_tree_entry(tree_sha1, filename, sha1, &oc->mode);
+                       if (!gently) {
+                               diagnose_invalid_sha1_path(prefix, filename,
+                                                          tree_sha1, object_name);
+                               free(object_name);
+                       }
+                       hashcpy(oc->tree, tree_sha1);
+                       strncpy(oc->path, filename,
+                               sizeof(oc->path));
+                       oc->path[sizeof(oc->path)-1] = '\0';
+
+                       free(new_filename);
+                       return ret;
+               } else {
+                       if (!gently)
+                               die("Invalid object name '%s'.", object_name);
+               }
        }
        return ret;
 }