Merge branch 'pt/pull-ff-vs-merge-ff'
[gitweb.git] / sha1_file.c
index dc940e63c453199dd9a7285533fbf2355bab03d1..ccc6dac54b570d7174175242ba2d535d4006ab1c 100644 (file)
@@ -8,6 +8,7 @@
  */
 #include "cache.h"
 #include "string-list.h"
+#include "lockfile.h"
 #include "delta.h"
 #include "pack.h"
 #include "blob.h"
@@ -404,7 +405,7 @@ void add_to_alternates_file(const char *reference)
 {
        struct lock_file *lock = xcalloc(1, sizeof(struct lock_file));
        int fd = hold_lock_file_for_append(lock, git_path("objects/info/alternates"), LOCK_DIE_ON_ERROR);
-       char *alt = mkpath("%s\n", reference);
+       const char *alt = mkpath("%s\n", reference);
        write_or_die(fd, alt, strlen(alt));
        if (commit_lock_file(lock))
                die("could not close alternates file");
@@ -412,14 +413,18 @@ void add_to_alternates_file(const char *reference)
                link_alt_odb_entries(alt, strlen(alt), '\n', NULL, 0);
 }
 
-void foreach_alt_odb(alt_odb_fn fn, void *cb)
+int foreach_alt_odb(alt_odb_fn fn, void *cb)
 {
        struct alternate_object_database *ent;
+       int r = 0;
 
        prepare_alt_odb();
-       for (ent = alt_odb_list; ent; ent = ent->next)
-               if (fn(ent, cb))
-                       return;
+       for (ent = alt_odb_list; ent; ent = ent->next) {
+               r = fn(ent, cb);
+               if (r)
+                       break;
+       }
+       return r;
 }
 
 void prepare_alt_odb(void)
@@ -438,27 +443,53 @@ void prepare_alt_odb(void)
        read_info_alternates(get_object_directory(), 0);
 }
 
-static int has_loose_object_local(const unsigned char *sha1)
+static int freshen_file(const char *fn)
 {
-       return !access(sha1_file_name(sha1), F_OK);
+       struct utimbuf t;
+       t.actime = t.modtime = time(NULL);
+       return !utime(fn, &t);
 }
 
-int has_loose_object_nonlocal(const unsigned char *sha1)
+static int check_and_freshen_file(const char *fn, int freshen)
+{
+       if (access(fn, F_OK))
+               return 0;
+       if (freshen && freshen_file(fn))
+               return 0;
+       return 1;
+}
+
+static int check_and_freshen_local(const unsigned char *sha1, int freshen)
+{
+       return check_and_freshen_file(sha1_file_name(sha1), freshen);
+}
+
+static int check_and_freshen_nonlocal(const unsigned char *sha1, int freshen)
 {
        struct alternate_object_database *alt;
        prepare_alt_odb();
        for (alt = alt_odb_list; alt; alt = alt->next) {
                fill_sha1_path(alt->name, sha1);
-               if (!access(alt->base, F_OK))
+               if (check_and_freshen_file(alt->base, freshen))
                        return 1;
        }
        return 0;
 }
 
+static int check_and_freshen(const unsigned char *sha1, int freshen)
+{
+       return check_and_freshen_local(sha1, freshen) ||
+              check_and_freshen_nonlocal(sha1, freshen);
+}
+
+int has_loose_object_nonlocal(const unsigned char *sha1)
+{
+       return check_and_freshen_nonlocal(sha1, 0);
+}
+
 static int has_loose_object(const unsigned char *sha1)
 {
-       return has_loose_object_local(sha1) ||
-              has_loose_object_nonlocal(sha1);
+       return check_and_freshen(sha1, 0);
 }
 
 static unsigned int pack_used_ctr;
@@ -663,10 +694,26 @@ void release_pack_memory(size_t need)
                ; /* nothing */
 }
 
+static void mmap_limit_check(size_t length)
+{
+       static size_t limit = 0;
+       if (!limit) {
+               limit = git_env_ulong("GIT_MMAP_LIMIT", 0);
+               if (!limit)
+                       limit = SIZE_MAX;
+       }
+       if (length > limit)
+               die("attempting to mmap %"PRIuMAX" over limit %"PRIuMAX,
+                   (uintmax_t)length, (uintmax_t)limit);
+}
+
 void *xmmap(void *start, size_t length,
        int prot, int flags, int fd, off_t offset)
 {
-       void *ret = mmap(start, length, prot, flags, fd, offset);
+       void *ret;
+
+       mmap_limit_check(length);
+       ret = mmap(start, length, prot, flags, fd, offset);
        if (ret == MAP_FAILED) {
                if (!length)
                        return NULL;
@@ -1151,7 +1198,7 @@ static void report_pack_garbage(struct string_list *list)
        if (!report_garbage)
                return;
 
-       sort_string_list(list);
+       string_list_sort(list);
 
        for (i = 0; i < list->nr; i++) {
                const char *path = list->items[i].string;
@@ -1517,6 +1564,40 @@ int unpack_sha1_header(git_zstream *stream, unsigned char *map, unsigned long ma
        return git_inflate(stream, 0);
 }
 
+static int unpack_sha1_header_to_strbuf(git_zstream *stream, unsigned char *map,
+                                       unsigned long mapsize, void *buffer,
+                                       unsigned long bufsiz, struct strbuf *header)
+{
+       int status;
+
+       status = unpack_sha1_header(stream, map, mapsize, buffer, bufsiz);
+
+       /*
+        * Check if entire header is unpacked in the first iteration.
+        */
+       if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
+               return 0;
+
+       /*
+        * buffer[0..bufsiz] was not large enough.  Copy the partial
+        * result out to header, and then append the result of further
+        * reading the stream.
+        */
+       strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
+       stream->next_out = buffer;
+       stream->avail_out = bufsiz;
+
+       do {
+               status = git_inflate(stream, 0);
+               strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
+               if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
+                       return 0;
+               stream->next_out = buffer;
+               stream->avail_out = bufsiz;
+       } while (status != Z_STREAM_END);
+       return -1;
+}
+
 static void *unpack_sha1_rest(git_zstream *stream, void *buffer, unsigned long size, const unsigned char *sha1)
 {
        int bytes = strlen(buffer) + 1;
@@ -1567,27 +1648,38 @@ static void *unpack_sha1_rest(git_zstream *stream, void *buffer, unsigned long s
  * too permissive for what we want to check. So do an anal
  * object header parse by hand.
  */
-int parse_sha1_header(const char *hdr, unsigned long *sizep)
+static int parse_sha1_header_extended(const char *hdr, struct object_info *oi,
+                              unsigned int flags)
 {
-       char type[10];
-       int i;
+       const char *type_buf = hdr;
        unsigned long size;
+       int type, type_len = 0;
 
        /*
-        * The type can be at most ten bytes (including the
-        * terminating '\0' that we add), and is followed by
+        * The type can be of any size but is followed by
         * a space.
         */
-       i = 0;
        for (;;) {
                char c = *hdr++;
                if (c == ' ')
                        break;
-               type[i++] = c;
-               if (i >= sizeof(type))
-                       return -1;
+               type_len++;
        }
-       type[i] = 0;
+
+       type = type_from_string_gently(type_buf, type_len, 1);
+       if (oi->typename)
+               strbuf_add(oi->typename, type_buf, type_len);
+       /*
+        * Set type to 0 if its an unknown object and
+        * we're obtaining the type using '--allow-unkown-type'
+        * option.
+        */
+       if ((flags & LOOKUP_UNKNOWN_OBJECT) && (type < 0))
+               type = 0;
+       else if (type < 0)
+               die("invalid object type");
+       if (oi->typep)
+               *oi->typep = type;
 
        /*
         * The length must follow immediately, and be in canonical
@@ -1605,12 +1697,24 @@ int parse_sha1_header(const char *hdr, unsigned long *sizep)
                        size = size * 10 + c;
                }
        }
-       *sizep = size;
+
+       if (oi->sizep)
+               *oi->sizep = size;
 
        /*
         * The length must be followed by a zero byte
         */
-       return *hdr ? -1 : type_from_string(type);
+       return *hdr ? -1 : type;
+}
+
+int parse_sha1_header(const char *hdr, unsigned long *sizep)
+{
+       struct object_info oi;
+
+       oi.sizep = sizep;
+       oi.typename = NULL;
+       oi.typep = NULL;
+       return parse_sha1_header_extended(hdr, &oi, LOOKUP_REPLACE_OBJECT);
 }
 
 static void *unpack_sha1_file(void *map, unsigned long mapsize, enum object_type *type, unsigned long *size, const unsigned char *sha1)
@@ -2426,10 +2530,8 @@ static int fill_pack_entry(const unsigned char *sha1,
         * answer, as it may have been deleted since the index was
         * loaded!
         */
-       if (!is_pack_valid(p)) {
-               warning("packfile %s cannot be accessed", p->pack_name);
+       if (!is_pack_valid(p))
                return 0;
-       }
        e->offset = offset;
        e->p = p;
        hashcpy(e->sha1, sha1);
@@ -2477,13 +2579,15 @@ struct packed_git *find_sha1_pack(const unsigned char *sha1,
 }
 
 static int sha1_loose_object_info(const unsigned char *sha1,
-                                 struct object_info *oi)
+                                 struct object_info *oi,
+                                 int flags)
 {
-       int status;
-       unsigned long mapsize, size;
+       int status = 0;
+       unsigned long mapsize;
        void *map;
        git_zstream stream;
        char hdr[32];
+       struct strbuf hdrbuf = STRBUF_INIT;
 
        if (oi->delta_base_sha1)
                hashclr(oi->delta_base_sha1);
@@ -2496,7 +2600,7 @@ static int sha1_loose_object_info(const unsigned char *sha1,
         * return value implicitly indicates whether the
         * object even exists.
         */
-       if (!oi->typep && !oi->sizep) {
+       if (!oi->typep && !oi->typename && !oi->sizep) {
                struct stat st;
                if (stat_sha1_file(sha1, &st) < 0)
                        return -1;
@@ -2510,17 +2614,26 @@ static int sha1_loose_object_info(const unsigned char *sha1,
                return -1;
        if (oi->disk_sizep)
                *oi->disk_sizep = mapsize;
-       if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
+       if ((flags & LOOKUP_UNKNOWN_OBJECT)) {
+               if (unpack_sha1_header_to_strbuf(&stream, map, mapsize, hdr, sizeof(hdr), &hdrbuf) < 0)
+                       status = error("unable to unpack %s header with --allow-unknown-type",
+                                      sha1_to_hex(sha1));
+       } else if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
                status = error("unable to unpack %s header",
                               sha1_to_hex(sha1));
-       else if ((status = parse_sha1_header(hdr, &size)) < 0)
+       if (status < 0)
+               ; /* Do nothing */
+       else if (hdrbuf.len) {
+               if ((status = parse_sha1_header_extended(hdrbuf.buf, oi, flags)) < 0)
+                       status = error("unable to parse %s header with --allow-unknown-type",
+                                      sha1_to_hex(sha1));
+       } else if ((status = parse_sha1_header_extended(hdr, oi, flags)) < 0)
                status = error("unable to parse %s header", sha1_to_hex(sha1));
-       else if (oi->sizep)
-               *oi->sizep = size;
        git_inflate_end(&stream);
        munmap(map, mapsize);
-       if (oi->typep)
+       if (status && oi->typep)
                *oi->typep = status;
+       strbuf_release(&hdrbuf);
        return 0;
 }
 
@@ -2529,6 +2642,7 @@ int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi,
        struct cached_object *co;
        struct pack_entry e;
        int rtype;
+       enum object_type real_type;
        const unsigned char *real = lookup_replace_object_extended(sha1, flags);
 
        co = find_cached_object(real);
@@ -2541,13 +2655,15 @@ int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi,
                        *(oi->disk_sizep) = 0;
                if (oi->delta_base_sha1)
                        hashclr(oi->delta_base_sha1);
+               if (oi->typename)
+                       strbuf_addstr(oi->typename, typename(co->type));
                oi->whence = OI_CACHED;
                return 0;
        }
 
        if (!find_pack_entry(real, &e)) {
                /* Most likely it's a loose object. */
-               if (!sha1_loose_object_info(real, oi)) {
+               if (!sha1_loose_object_info(real, oi, flags)) {
                        oi->whence = OI_LOOSE;
                        return 0;
                }
@@ -2558,9 +2674,18 @@ int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi,
                        return -1;
        }
 
+       /*
+        * packed_object_info() does not follow the delta chain to
+        * find out the real type, unless it is given oi->typep.
+        */
+       if (oi->typename && !oi->typep)
+               oi->typep = &real_type;
+
        rtype = packed_object_info(e.p, e.offset, oi);
        if (rtype < 0) {
                mark_bad_packed_object(e.p, real);
+               if (oi->typep == &real_type)
+                       oi->typep = NULL;
                return sha1_object_info_extended(real, oi, 0);
        } else if (in_delta_base_cache(e.p, e.offset)) {
                oi->whence = OI_DBCACHED;
@@ -2571,6 +2696,10 @@ int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi,
                oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA ||
                                         rtype == OBJ_OFS_DELTA);
        }
+       if (oi->typename)
+               strbuf_addstr(oi->typename, typename(*oi->typep));
+       if (oi->typep == &real_type)
+               oi->typep = NULL;
 
        return 0;
 }
@@ -2896,7 +3025,6 @@ static int write_loose_object(const unsigned char *sha1, char *hdr, int hdrlen,
        }
 
        /* Set it up */
-       memset(&stream, 0, sizeof(stream));
        git_deflate_init(&stream, zlib_compression_level);
        stream.next_out = compressed;
        stream.avail_out = sizeof(compressed);
@@ -2945,9 +3073,26 @@ static int write_loose_object(const unsigned char *sha1, char *hdr, int hdrlen,
        return move_temp_to_file(tmp_file, filename);
 }
 
-int write_sha1_file(const void *buf, unsigned long len, const char *type, unsigned char *returnsha1)
+static int freshen_loose_object(const unsigned char *sha1)
+{
+       return check_and_freshen(sha1, 1);
+}
+
+static int freshen_packed_object(const unsigned char *sha1)
+{
+       struct pack_entry e;
+       if (!find_pack_entry(sha1, &e))
+               return 0;
+       if (e.p->freshened)
+               return 1;
+       if (!freshen_file(e.p->pack_name))
+               return 0;
+       e.p->freshened = 1;
+       return 1;
+}
+
+int write_sha1_file(const void *buf, unsigned long len, const char *type, unsigned char *sha1)
 {
-       unsigned char sha1[20];
        char hdr[32];
        int hdrlen;
 
@@ -2955,9 +3100,7 @@ int write_sha1_file(const void *buf, unsigned long len, const char *type, unsign
         * it out into .git/objects/??/?{38} file.
         */
        write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
-       if (returnsha1)
-               hashcpy(returnsha1, sha1);
-       if (has_sha1_file(sha1))
+       if (freshen_packed_object(sha1) || freshen_loose_object(sha1))
                return 0;
        return write_loose_object(sha1, hdr, hdrlen, buf, len, 0);
 }
@@ -2974,7 +3117,7 @@ int hash_sha1_file_literally(const void *buf, unsigned long len, const char *typ
 
        if (!(flags & HASH_WRITE_OBJECT))
                goto cleanup;
-       if (has_sha1_file(sha1))
+       if (freshen_packed_object(sha1) || freshen_loose_object(sha1))
                goto cleanup;
        status = write_loose_object(sha1, header, hdrlen, buf, len, 0);
 
@@ -3097,6 +3240,29 @@ static int index_mem(unsigned char *sha1, void *buf, size_t size,
        return ret;
 }
 
+static int index_stream_convert_blob(unsigned char *sha1, int fd,
+                                    const char *path, unsigned flags)
+{
+       int ret;
+       const int write_object = flags & HASH_WRITE_OBJECT;
+       struct strbuf sbuf = STRBUF_INIT;
+
+       assert(path);
+       assert(would_convert_to_git_filter_fd(path));
+
+       convert_to_git_filter_fd(path, fd, &sbuf,
+                                write_object ? safe_crlf : SAFE_CRLF_FALSE);
+
+       if (write_object)
+               ret = write_sha1_file(sbuf.buf, sbuf.len, typename(OBJ_BLOB),
+                                     sha1);
+       else
+               ret = hash_sha1_file(sbuf.buf, sbuf.len, typename(OBJ_BLOB),
+                                    sha1);
+       strbuf_release(&sbuf);
+       return ret;
+}
+
 static int index_pipe(unsigned char *sha1, int fd, enum object_type type,
                      const char *path, unsigned flags)
 {
@@ -3162,15 +3328,22 @@ int index_fd(unsigned char *sha1, int fd, struct stat *st,
             enum object_type type, const char *path, unsigned flags)
 {
        int ret;
-       size_t size = xsize_t(st->st_size);
 
-       if (!S_ISREG(st->st_mode))
+       /*
+        * Call xsize_t() only when needed to avoid potentially unnecessary
+        * die() for large files.
+        */
+       if (type == OBJ_BLOB && path && would_convert_to_git_filter_fd(path))
+               ret = index_stream_convert_blob(sha1, fd, path, flags);
+       else if (!S_ISREG(st->st_mode))
                ret = index_pipe(sha1, fd, type, path, flags);
-       else if (size <= big_file_threshold || type != OBJ_BLOB ||
-                (path && would_convert_to_git(path, NULL, 0, 0)))
-               ret = index_core(sha1, fd, size, type, path, flags);
+       else if (st->st_size <= big_file_threshold || type != OBJ_BLOB ||
+                (path && would_convert_to_git(path)))
+               ret = index_core(sha1, fd, xsize_t(st->st_size), type, path,
+                                flags);
        else
-               ret = index_stream(sha1, fd, size, type, path, flags);
+               ret = index_stream(sha1, fd, xsize_t(st->st_size), type, path,
+                                  flags);
        close(fd);
        return ret;
 }
@@ -3235,3 +3408,172 @@ void assert_sha1_type(const unsigned char *sha1, enum object_type expect)
                die("%s is not a valid '%s' object", sha1_to_hex(sha1),
                    typename(expect));
 }
+
+static int for_each_file_in_obj_subdir(int subdir_nr,
+                                      struct strbuf *path,
+                                      each_loose_object_fn obj_cb,
+                                      each_loose_cruft_fn cruft_cb,
+                                      each_loose_subdir_fn subdir_cb,
+                                      void *data)
+{
+       size_t baselen = path->len;
+       DIR *dir = opendir(path->buf);
+       struct dirent *de;
+       int r = 0;
+
+       if (!dir) {
+               if (errno == ENOENT)
+                       return 0;
+               return error("unable to open %s: %s", path->buf, strerror(errno));
+       }
+
+       while ((de = readdir(dir))) {
+               if (is_dot_or_dotdot(de->d_name))
+                       continue;
+
+               strbuf_setlen(path, baselen);
+               strbuf_addf(path, "/%s", de->d_name);
+
+               if (strlen(de->d_name) == 38)  {
+                       char hex[41];
+                       unsigned char sha1[20];
+
+                       snprintf(hex, sizeof(hex), "%02x%s",
+                                subdir_nr, de->d_name);
+                       if (!get_sha1_hex(hex, sha1)) {
+                               if (obj_cb) {
+                                       r = obj_cb(sha1, path->buf, data);
+                                       if (r)
+                                               break;
+                               }
+                               continue;
+                       }
+               }
+
+               if (cruft_cb) {
+                       r = cruft_cb(de->d_name, path->buf, data);
+                       if (r)
+                               break;
+               }
+       }
+       strbuf_setlen(path, baselen);
+
+       if (!r && subdir_cb)
+               r = subdir_cb(subdir_nr, path->buf, data);
+
+       closedir(dir);
+       return r;
+}
+
+int for_each_loose_file_in_objdir_buf(struct strbuf *path,
+                           each_loose_object_fn obj_cb,
+                           each_loose_cruft_fn cruft_cb,
+                           each_loose_subdir_fn subdir_cb,
+                           void *data)
+{
+       size_t baselen = path->len;
+       int r = 0;
+       int i;
+
+       for (i = 0; i < 256; i++) {
+               strbuf_addf(path, "/%02x", i);
+               r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb,
+                                               subdir_cb, data);
+               strbuf_setlen(path, baselen);
+               if (r)
+                       break;
+       }
+
+       return r;
+}
+
+int for_each_loose_file_in_objdir(const char *path,
+                                 each_loose_object_fn obj_cb,
+                                 each_loose_cruft_fn cruft_cb,
+                                 each_loose_subdir_fn subdir_cb,
+                                 void *data)
+{
+       struct strbuf buf = STRBUF_INIT;
+       int r;
+
+       strbuf_addstr(&buf, path);
+       r = for_each_loose_file_in_objdir_buf(&buf, obj_cb, cruft_cb,
+                                             subdir_cb, data);
+       strbuf_release(&buf);
+
+       return r;
+}
+
+struct loose_alt_odb_data {
+       each_loose_object_fn *cb;
+       void *data;
+};
+
+static int loose_from_alt_odb(struct alternate_object_database *alt,
+                             void *vdata)
+{
+       struct loose_alt_odb_data *data = vdata;
+       struct strbuf buf = STRBUF_INIT;
+       int r;
+
+       /* copy base not including trailing '/' */
+       strbuf_add(&buf, alt->base, alt->name - alt->base - 1);
+       r = for_each_loose_file_in_objdir_buf(&buf,
+                                             data->cb, NULL, NULL,
+                                             data->data);
+       strbuf_release(&buf);
+       return r;
+}
+
+int for_each_loose_object(each_loose_object_fn cb, void *data, unsigned flags)
+{
+       struct loose_alt_odb_data alt;
+       int r;
+
+       r = for_each_loose_file_in_objdir(get_object_directory(),
+                                         cb, NULL, NULL, data);
+       if (r)
+               return r;
+
+       if (flags & FOR_EACH_OBJECT_LOCAL_ONLY)
+               return 0;
+
+       alt.cb = cb;
+       alt.data = data;
+       return foreach_alt_odb(loose_from_alt_odb, &alt);
+}
+
+static int for_each_object_in_pack(struct packed_git *p, each_packed_object_fn cb, void *data)
+{
+       uint32_t i;
+       int r = 0;
+
+       for (i = 0; i < p->num_objects; i++) {
+               const unsigned char *sha1 = nth_packed_object_sha1(p, i);
+
+               if (!sha1)
+                       return error("unable to get sha1 of object %u in %s",
+                                    i, p->pack_name);
+
+               r = cb(sha1, p, i, data);
+               if (r)
+                       break;
+       }
+       return r;
+}
+
+int for_each_packed_object(each_packed_object_fn cb, void *data, unsigned flags)
+{
+       struct packed_git *p;
+       int r = 0;
+
+       prepare_packed_git();
+       for (p = packed_git; p; p = p->next) {
+               if ((flags & FOR_EACH_OBJECT_LOCAL_ONLY) && !p->pack_local)
+                       continue;
+               r = for_each_object_in_pack(p, cb, data);
+               if (r)
+                       break;
+       }
+       return r;
+}