Code

Revert "t0081 (line-buffer): add buffering tests"
[git.git] / fast-import.c
index 77549ebd6fbfbea051901d4ab474eebc8e2dbdbd..e1268b8cb4ae2ac482928b23e6a3a04bf78c7fcb 100644 (file)
@@ -24,10 +24,12 @@ Format of STDIN stream:
     commit_msg
     ('from' sp committish lf)?
     ('merge' sp committish lf)*
-    file_change*
+    (file_change | ls)*
     lf?;
   commit_msg ::= data;
 
+  ls ::= 'ls' sp '"' quoted(path) '"' lf;
+
   file_change ::= file_clr
     | file_del
     | file_rnm
@@ -132,14 +134,19 @@ Format of STDIN stream:
   ts    ::= # time since the epoch in seconds, ascii base10 notation;
   tz    ::= # GIT style timezone;
 
-     # note: comments may appear anywhere in the input, except
-     # within a data command.  Any form of the data command
-     # always escapes the related input from comment processing.
+     # note: comments, ls and cat requests may appear anywhere
+     # in the input, except within a data command.  Any form
+     # of the data command always escapes the related input
+     # from comment processing.
      #
      # In case it is not clear, the '#' that starts the comment
      # must be the first character on that line (an lf
      # preceded it).
      #
+
+  cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
+  ls_tree  ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
+
   comment ::= '#' not_lf* lf;
   not_lf  ::= # Any byte that is not ASCII newline (LF);
 */
@@ -156,6 +163,7 @@ Format of STDIN stream:
 #include "csum-file.h"
 #include "quote.h"
 #include "exec_cmd.h"
+#include "dir.h"
 
 #define PACK_ID_BITS 16
 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
@@ -325,6 +333,7 @@ static struct mark_set *marks;
 static const char *export_marks_file;
 static const char *import_marks_file;
 static int import_marks_file_from_stream;
+static int import_marks_file_ignore_missing;
 static int relative_marks_paths;
 
 /* Our last blob */
@@ -361,7 +370,15 @@ static uintmax_t next_mark;
 static struct strbuf new_data = STRBUF_INIT;
 static int seen_data_command;
 
+/* Signal handling */
+static volatile sig_atomic_t checkpoint_requested;
+
+/* Where to write output of cat-blob commands */
+static int cat_blob_fd = STDOUT_FILENO;
+
 static void parse_argv(void);
+static void parse_cat_blob(void);
+static void parse_ls(struct branch *b);
 
 static void write_branch_report(FILE *rpt, struct branch *b)
 {
@@ -500,6 +517,32 @@ static NORETURN void die_nicely(const char *err, va_list params)
        exit(128);
 }
 
+#ifndef SIGUSR1        /* Windows, for example */
+
+static void set_checkpoint_signal(void)
+{
+}
+
+#else
+
+static void checkpoint_signal(int signo)
+{
+       checkpoint_requested = 1;
+}
+
+static void set_checkpoint_signal(void)
+{
+       struct sigaction sa;
+
+       memset(&sa, 0, sizeof(sa));
+       sa.sa_handler = checkpoint_signal;
+       sigemptyset(&sa.sa_mask);
+       sa.sa_flags = SA_RESTART;
+       sigaction(SIGUSR1, &sa, NULL);
+}
+
+#endif
+
 static void alloc_objects(unsigned int cnt)
 {
        struct object_entry_pool *b;
@@ -539,22 +582,17 @@ static struct object_entry *insert_object(unsigned char *sha1)
 {
        unsigned int h = sha1[0] << 8 | sha1[1];
        struct object_entry *e = object_table[h];
-       struct object_entry *p = NULL;
 
        while (e) {
                if (!hashcmp(sha1, e->idx.sha1))
                        return e;
-               p = e;
                e = e->next;
        }
 
        e = new_object(sha1);
-       e->next = NULL;
+       e->next = object_table[h];
        e->idx.offset = 0;
-       if (p)
-               p->next = e;
-       else
-               object_table[h] = e;
+       object_table[h] = e;
        return e;
 }
 
@@ -1437,6 +1475,20 @@ static void store_tree(struct tree_entry *root)
        t->entry_count -= del;
 }
 
+static void tree_content_replace(
+       struct tree_entry *root,
+       const unsigned char *sha1,
+       const uint16_t mode,
+       struct tree_content *newtree)
+{
+       if (!S_ISDIR(mode))
+               die("Root cannot be a non-directory");
+       hashcpy(root->versions[1].sha1, sha1);
+       if (root->tree)
+               release_tree_content_recursive(root->tree);
+       root->tree = newtree;
+}
+
 static int tree_content_set(
        struct tree_entry *root,
        const char *p,
@@ -1444,7 +1496,7 @@ static int tree_content_set(
        const uint16_t mode,
        struct tree_content *subtree)
 {
-       struct tree_content *t = root->tree;
+       struct tree_content *t;
        const char *slash1;
        unsigned int i, n;
        struct tree_entry *e;
@@ -1454,23 +1506,17 @@ static int tree_content_set(
                n = slash1 - p;
        else
                n = strlen(p);
-       if (!slash1 && !n) {
-               if (!S_ISDIR(mode))
-                       die("Root cannot be a non-directory");
-               hashcpy(root->versions[1].sha1, sha1);
-               if (root->tree)
-                       release_tree_content_recursive(root->tree);
-               root->tree = subtree;
-               return 1;
-       }
        if (!n)
                die("Empty path component found in input");
        if (!slash1 && !S_ISDIR(mode) && subtree)
                die("Non-directories cannot have subtrees");
 
+       if (!root->tree)
+               load_tree(root);
+       t = root->tree;
        for (i = 0; i < t->entry_count; i++) {
                e = t->entries[i];
-               if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
+               if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
                        if (!slash1) {
                                if (!S_ISDIR(mode)
                                                && e->versions[1].mode == mode
@@ -1523,7 +1569,7 @@ static int tree_content_remove(
        const char *p,
        struct tree_entry *backup_leaf)
 {
-       struct tree_content *t = root->tree;
+       struct tree_content *t;
        const char *slash1;
        unsigned int i, n;
        struct tree_entry *e;
@@ -1534,9 +1580,12 @@ static int tree_content_remove(
        else
                n = strlen(p);
 
+       if (!root->tree)
+               load_tree(root);
+       t = root->tree;
        for (i = 0; i < t->entry_count; i++) {
                e = t->entries[i];
-               if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
+               if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
                        if (slash1 && !S_ISDIR(e->versions[1].mode))
                                /*
                                 * If p names a file in some subdirectory, and a
@@ -1581,7 +1630,7 @@ static int tree_content_get(
        const char *p,
        struct tree_entry *leaf)
 {
-       struct tree_content *t = root->tree;
+       struct tree_content *t;
        const char *slash1;
        unsigned int i, n;
        struct tree_entry *e;
@@ -1592,9 +1641,12 @@ static int tree_content_get(
        else
                n = strlen(p);
 
+       if (!root->tree)
+               load_tree(root);
+       t = root->tree;
        for (i = 0; i < t->entry_count; i++) {
                e = t->entries[i];
-               if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
+               if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
                        if (!slash1) {
                                memcpy(leaf, e, sizeof(*leaf));
                                if (e->tree && is_null_sha1(e->versions[1].sha1))
@@ -1749,7 +1801,11 @@ static void read_marks(void)
 {
        char line[512];
        FILE *f = fopen(import_marks_file, "r");
-       if (!f)
+       if (f)
+               ;
+       else if (import_marks_file_ignore_missing && errno == ENOENT)
+               return; /* Marks file does not exist */
+       else
                die_errno("cannot read '%s'", import_marks_file);
        while (fgets(line, sizeof(line), f)) {
                uintmax_t mark;
@@ -1790,7 +1846,7 @@ static int read_next_command(void)
                return EOF;
        }
 
-       do {
+       for (;;) {
                if (unread_command_buf) {
                        unread_command_buf = 0;
                } else {
@@ -1823,9 +1879,14 @@ static int read_next_command(void)
                        rc->prev->next = rc;
                        cmd_tail = rc;
                }
-       } while (command_buf.buf[0] == '#');
-
-       return 0;
+               if (!prefixcmp(command_buf.buf, "cat-blob ")) {
+                       parse_cat_blob();
+                       continue;
+               }
+               if (command_buf.buf[0] == '#')
+                       continue;
+               return 0;
+       }
 }
 
 static void skip_optional_lf(void)
@@ -2180,6 +2241,12 @@ static void file_change_m(struct branch *b)
                p = uq.buf;
        }
 
+       /* Git does not track empty, non-toplevel directories. */
+       if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
+               tree_content_remove(&b->branch_tree, p, NULL);
+               return;
+       }
+
        if (S_ISGITLINK(mode)) {
                if (inline_data)
                        die("Git links cannot be specified 'inline': %s",
@@ -2218,6 +2285,10 @@ static void file_change_m(struct branch *b)
                                command_buf.buf);
        }
 
+       if (!*p) {
+               tree_content_replace(&b->branch_tree, sha1, mode, NULL);
+               return;
+       }
        tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
 }
 
@@ -2276,6 +2347,13 @@ static void file_change_cr(struct branch *b, int rename)
                tree_content_get(&b->branch_tree, s, &leaf);
        if (!leaf.versions[1].mode)
                die("Path %s not in branch", s);
+       if (!*d) {      /* C "path/to/subdir" "" */
+               tree_content_replace(&b->branch_tree,
+                       leaf.versions[1].sha1,
+                       leaf.versions[1].mode,
+                       leaf.tree);
+               return;
+       }
        tree_content_set(&b->branch_tree, d,
                leaf.versions[1].sha1,
                leaf.versions[1].mode,
@@ -2541,6 +2619,8 @@ static void parse_new_commit(void)
                        note_change_n(b, prev_fanout);
                else if (!strcmp("deleteall", command_buf.buf))
                        file_change_deleteall(b);
+               else if (!prefixcmp(command_buf.buf, "ls "))
+                       parse_ls(b);
                else {
                        unread_command_buf = 1;
                        break;
@@ -2689,14 +2769,242 @@ static void parse_reset_branch(void)
                unread_command_buf = 1;
 }
 
-static void parse_checkpoint(void)
+static void cat_blob_write(const char *buf, unsigned long size)
+{
+       if (write_in_full(cat_blob_fd, buf, size) != size)
+               die_errno("Write to frontend failed");
+}
+
+static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
+{
+       struct strbuf line = STRBUF_INIT;
+       unsigned long size;
+       enum object_type type = 0;
+       char *buf;
+
+       if (!oe || oe->pack_id == MAX_PACK_ID) {
+               buf = read_sha1_file(sha1, &type, &size);
+       } else {
+               type = oe->type;
+               buf = gfi_unpack_entry(oe, &size);
+       }
+
+       /*
+        * Output based on batch_one_object() from cat-file.c.
+        */
+       if (type <= 0) {
+               strbuf_reset(&line);
+               strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
+               cat_blob_write(line.buf, line.len);
+               strbuf_release(&line);
+               free(buf);
+               return;
+       }
+       if (!buf)
+               die("Can't read object %s", sha1_to_hex(sha1));
+       if (type != OBJ_BLOB)
+               die("Object %s is a %s but a blob was expected.",
+                   sha1_to_hex(sha1), typename(type));
+       strbuf_reset(&line);
+       strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
+                                               typename(type), size);
+       cat_blob_write(line.buf, line.len);
+       strbuf_release(&line);
+       cat_blob_write(buf, size);
+       cat_blob_write("\n", 1);
+       free(buf);
+}
+
+static void parse_cat_blob(void)
+{
+       const char *p;
+       struct object_entry *oe = oe;
+       unsigned char sha1[20];
+
+       /* cat-blob SP <object> LF */
+       p = command_buf.buf + strlen("cat-blob ");
+       if (*p == ':') {
+               char *x;
+               oe = find_mark(strtoumax(p + 1, &x, 10));
+               if (x == p + 1)
+                       die("Invalid mark: %s", command_buf.buf);
+               if (!oe)
+                       die("Unknown mark: %s", command_buf.buf);
+               if (*x)
+                       die("Garbage after mark: %s", command_buf.buf);
+               hashcpy(sha1, oe->idx.sha1);
+       } else {
+               if (get_sha1_hex(p, sha1))
+                       die("Invalid SHA1: %s", command_buf.buf);
+               if (p[40])
+                       die("Garbage after SHA1: %s", command_buf.buf);
+               oe = find_object(sha1);
+       }
+
+       cat_blob(oe, sha1);
+}
+
+static struct object_entry *dereference(struct object_entry *oe,
+                                       unsigned char sha1[20])
+{
+       unsigned long size;
+       char *buf = NULL;
+       if (!oe) {
+               enum object_type type = sha1_object_info(sha1, NULL);
+               if (type < 0)
+                       die("object not found: %s", sha1_to_hex(sha1));
+               /* cache it! */
+               oe = insert_object(sha1);
+               oe->type = type;
+               oe->pack_id = MAX_PACK_ID;
+               oe->idx.offset = 1;
+       }
+       switch (oe->type) {
+       case OBJ_TREE:  /* easy case. */
+               return oe;
+       case OBJ_COMMIT:
+       case OBJ_TAG:
+               break;
+       default:
+               die("Not a treeish: %s", command_buf.buf);
+       }
+
+       if (oe->pack_id != MAX_PACK_ID) {       /* in a pack being written */
+               buf = gfi_unpack_entry(oe, &size);
+       } else {
+               enum object_type unused;
+               buf = read_sha1_file(sha1, &unused, &size);
+       }
+       if (!buf)
+               die("Can't load object %s", sha1_to_hex(sha1));
+
+       /* Peel one layer. */
+       switch (oe->type) {
+       case OBJ_TAG:
+               if (size < 40 + strlen("object ") ||
+                   get_sha1_hex(buf + strlen("object "), sha1))
+                       die("Invalid SHA1 in tag: %s", command_buf.buf);
+               break;
+       case OBJ_COMMIT:
+               if (size < 40 + strlen("tree ") ||
+                   get_sha1_hex(buf + strlen("tree "), sha1))
+                       die("Invalid SHA1 in commit: %s", command_buf.buf);
+       }
+
+       free(buf);
+       return find_object(sha1);
+}
+
+static struct object_entry *parse_treeish_dataref(const char **p)
+{
+       unsigned char sha1[20];
+       struct object_entry *e;
+
+       if (**p == ':') {       /* <mark> */
+               char *endptr;
+               e = find_mark(strtoumax(*p + 1, &endptr, 10));
+               if (endptr == *p + 1)
+                       die("Invalid mark: %s", command_buf.buf);
+               if (!e)
+                       die("Unknown mark: %s", command_buf.buf);
+               *p = endptr;
+               hashcpy(sha1, e->idx.sha1);
+       } else {        /* <sha1> */
+               if (get_sha1_hex(*p, sha1))
+                       die("Invalid SHA1: %s", command_buf.buf);
+               e = find_object(sha1);
+               *p += 40;
+       }
+
+       while (!e || e->type != OBJ_TREE)
+               e = dereference(e, sha1);
+       return e;
+}
+
+static void print_ls(int mode, const unsigned char *sha1, const char *path)
+{
+       static struct strbuf line = STRBUF_INIT;
+
+       /* See show_tree(). */
+       const char *type =
+               S_ISGITLINK(mode) ? commit_type :
+               S_ISDIR(mode) ? tree_type :
+               blob_type;
+
+       if (!mode) {
+               /* missing SP path LF */
+               strbuf_reset(&line);
+               strbuf_addstr(&line, "missing ");
+               quote_c_style(path, &line, NULL, 0);
+               strbuf_addch(&line, '\n');
+       } else {
+               /* mode SP type SP object_name TAB path LF */
+               strbuf_reset(&line);
+               strbuf_addf(&line, "%06o %s %s\t",
+                               mode, type, sha1_to_hex(sha1));
+               quote_c_style(path, &line, NULL, 0);
+               strbuf_addch(&line, '\n');
+       }
+       cat_blob_write(line.buf, line.len);
+}
+
+static void parse_ls(struct branch *b)
+{
+       const char *p;
+       struct tree_entry *root = NULL;
+       struct tree_entry leaf = {0};
+
+       /* ls SP (<treeish> SP)? <path> */
+       p = command_buf.buf + strlen("ls ");
+       if (*p == '"') {
+               if (!b)
+                       die("Not in a commit: %s", command_buf.buf);
+               root = &b->branch_tree;
+       } else {
+               struct object_entry *e = parse_treeish_dataref(&p);
+               root = new_tree_entry();
+               hashcpy(root->versions[1].sha1, e->idx.sha1);
+               load_tree(root);
+               if (*p++ != ' ')
+                       die("Missing space after tree-ish: %s", command_buf.buf);
+       }
+       if (*p == '"') {
+               static struct strbuf uq = STRBUF_INIT;
+               const char *endp;
+               strbuf_reset(&uq);
+               if (unquote_c_style(&uq, p, &endp))
+                       die("Invalid path: %s", command_buf.buf);
+               if (*endp)
+                       die("Garbage after path in: %s", command_buf.buf);
+               p = uq.buf;
+       }
+       tree_content_get(root, p, &leaf);
+       /*
+        * A directory in preparation would have a sha1 of zero
+        * until it is saved.  Save, for simplicity.
+        */
+       if (S_ISDIR(leaf.versions[1].mode))
+               store_tree(&leaf);
+
+       print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
+       if (!b || root != &b->branch_tree)
+               release_tree_entry(root);
+}
+
+static void checkpoint(void)
 {
+       checkpoint_requested = 0;
        if (object_count) {
                cycle_packfile();
                dump_branches();
                dump_tags();
                dump_marks();
        }
+}
+
+static void parse_checkpoint(void)
+{
+       checkpoint_requested = 1;
        skip_optional_lf();
 }
 
@@ -2718,7 +3026,8 @@ static char* make_fast_import_path(const char *path)
        return strbuf_detach(&abs_path, NULL);
 }
 
-static void option_import_marks(const char *marks, int from_stream)
+static void option_import_marks(const char *marks,
+                                       int from_stream, int ignore_missing)
 {
        if (import_marks_file) {
                if (from_stream)
@@ -2732,6 +3041,7 @@ static void option_import_marks(const char *marks, int from_stream)
        import_marks_file = make_fast_import_path(marks);
        safe_create_leading_directories_const(import_marks_file);
        import_marks_file_from_stream = from_stream;
+       import_marks_file_ignore_missing = ignore_missing;
 }
 
 static void option_date_format(const char *fmt)
@@ -2746,16 +3056,25 @@ static void option_date_format(const char *fmt)
                die("unknown --date-format argument %s", fmt);
 }
 
+static unsigned long ulong_arg(const char *option, const char *arg)
+{
+       char *endptr;
+       unsigned long rv = strtoul(arg, &endptr, 0);
+       if (strchr(arg, '-') || endptr == arg || *endptr)
+               die("%s: argument must be a non-negative integer", option);
+       return rv;
+}
+
 static void option_depth(const char *depth)
 {
-       max_depth = strtoul(depth, NULL, 0);
+       max_depth = ulong_arg("--depth", depth);
        if (max_depth > MAX_DEPTH)
                die("--depth cannot exceed %u", MAX_DEPTH);
 }
 
 static void option_active_branches(const char *branches)
 {
-       max_active_branches = strtoul(branches, NULL, 0);
+       max_active_branches = ulong_arg("--active-branches", branches);
 }
 
 static void option_export_marks(const char *marks)
@@ -2764,6 +3083,14 @@ static void option_export_marks(const char *marks)
        safe_create_leading_directories_const(export_marks_file);
 }
 
+static void option_cat_blob_fd(const char *fd)
+{
+       unsigned long n = ulong_arg("--cat-blob-fd", fd);
+       if (n > (unsigned long) INT_MAX)
+               die("--cat-blob-fd cannot exceed %d", INT_MAX);
+       cat_blob_fd = (int) n;
+}
+
 static void option_export_pack_edges(const char *edges)
 {
        if (pack_edges)
@@ -2814,15 +3141,22 @@ static int parse_one_feature(const char *feature, int from_stream)
        if (!prefixcmp(feature, "date-format=")) {
                option_date_format(feature + 12);
        } else if (!prefixcmp(feature, "import-marks=")) {
-               option_import_marks(feature + 13, from_stream);
+               option_import_marks(feature + 13, from_stream, 0);
+       } else if (!prefixcmp(feature, "import-marks-if-exists=")) {
+               option_import_marks(feature + strlen("import-marks-if-exists="),
+                                       from_stream, 1);
        } else if (!prefixcmp(feature, "export-marks=")) {
                option_export_marks(feature + 13);
+       } else if (!strcmp(feature, "cat-blob")) {
+               ; /* Don't die - this feature is supported */
        } else if (!prefixcmp(feature, "relative-marks")) {
                relative_marks_paths = 1;
        } else if (!prefixcmp(feature, "no-relative-marks")) {
                relative_marks_paths = 0;
        } else if (!prefixcmp(feature, "force")) {
                force_update = 1;
+       } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
+               ; /* do nothing; we have the feature */
        } else {
                return 0;
        }
@@ -2911,6 +3245,11 @@ static void parse_argv(void)
                if (parse_one_feature(a + 2, 0))
                        continue;
 
+               if (!prefixcmp(a + 2, "cat-blob-fd=")) {
+                       option_cat_blob_fd(a + 2 + strlen("cat-blob-fd="));
+                       continue;
+               }
+
                die("unknown option %s", a);
        }
        if (i != global_argc)
@@ -2953,9 +3292,12 @@ int main(int argc, const char **argv)
        prepare_packed_git();
        start_packfile();
        set_die_routine(die_nicely);
+       set_checkpoint_signal();
        while (read_next_command() != EOF) {
                if (!strcmp("blob", command_buf.buf))
                        parse_new_blob();
+               else if (!prefixcmp(command_buf.buf, "ls "))
+                       parse_ls(NULL);
                else if (!prefixcmp(command_buf.buf, "commit "))
                        parse_new_commit();
                else if (!prefixcmp(command_buf.buf, "tag "))
@@ -2974,6 +3316,9 @@ int main(int argc, const char **argv)
                        /* ignore non-git options*/;
                else
                        die("Unsupported command: %s", command_buf.buf);
+
+               if (checkpoint_requested)
+                       checkpoint();
        }
 
        /* argv hasn't been parsed yet, do so */