convert-objects.con commit Add hash_sha1_file() (abdc3fc)
   1#define _XOPEN_SOURCE 500 /* glibc2 and AIX 5.3L need this */
   2#define _XOPEN_SOURCE_EXTENDED 1 /* AIX 5.3L needs this */
   3#define _GNU_SOURCE
   4#include <time.h>
   5#include "cache.h"
   6#include "blob.h"
   7#include "commit.h"
   8#include "tree.h"
   9
  10struct entry {
  11        unsigned char old_sha1[20];
  12        unsigned char new_sha1[20];
  13        int converted;
  14};
  15
  16#define MAXOBJECTS (1000000)
  17
  18static struct entry *convert[MAXOBJECTS];
  19static int nr_convert;
  20
  21static struct entry * convert_entry(unsigned char *sha1);
  22
  23static struct entry *insert_new(unsigned char *sha1, int pos)
  24{
  25        struct entry *new = xcalloc(1, sizeof(struct entry));
  26        hashcpy(new->old_sha1, sha1);
  27        memmove(convert + pos + 1, convert + pos, (nr_convert - pos) * sizeof(struct entry *));
  28        convert[pos] = new;
  29        nr_convert++;
  30        if (nr_convert == MAXOBJECTS)
  31                die("you're kidding me - hit maximum object limit");
  32        return new;
  33}
  34
  35static struct entry *lookup_entry(unsigned char *sha1)
  36{
  37        int low = 0, high = nr_convert;
  38
  39        while (low < high) {
  40                int next = (low + high) / 2;
  41                struct entry *n = convert[next];
  42                int cmp = hashcmp(sha1, n->old_sha1);
  43                if (!cmp)
  44                        return n;
  45                if (cmp < 0) {
  46                        high = next;
  47                        continue;
  48                }
  49                low = next+1;
  50        }
  51        return insert_new(sha1, low);
  52}
  53
  54static void convert_binary_sha1(void *buffer)
  55{
  56        struct entry *entry = convert_entry(buffer);
  57        hashcpy(buffer, entry->new_sha1);
  58}
  59
  60static void convert_ascii_sha1(void *buffer)
  61{
  62        unsigned char sha1[20];
  63        struct entry *entry;
  64
  65        if (get_sha1_hex(buffer, sha1))
  66                die("expected sha1, got '%s'", (char*) buffer);
  67        entry = convert_entry(sha1);
  68        memcpy(buffer, sha1_to_hex(entry->new_sha1), 40);
  69}
  70
  71static unsigned int convert_mode(unsigned int mode)
  72{
  73        unsigned int newmode;
  74
  75        newmode = mode & S_IFMT;
  76        if (S_ISREG(mode))
  77                newmode |= (mode & 0100) ? 0755 : 0644;
  78        return newmode;
  79}
  80
  81static int write_subdirectory(void *buffer, unsigned long size, const char *base, int baselen, unsigned char *result_sha1)
  82{
  83        char *new = xmalloc(size);
  84        unsigned long newlen = 0;
  85        unsigned long used;
  86
  87        used = 0;
  88        while (size) {
  89                int len = 21 + strlen(buffer);
  90                char *path = strchr(buffer, ' ');
  91                unsigned char *sha1;
  92                unsigned int mode;
  93                char *slash, *origpath;
  94
  95                if (!path || sscanf(buffer, "%o", &mode) != 1)
  96                        die("bad tree conversion");
  97                mode = convert_mode(mode);
  98                path++;
  99                if (memcmp(path, base, baselen))
 100                        break;
 101                origpath = path;
 102                path += baselen;
 103                slash = strchr(path, '/');
 104                if (!slash) {
 105                        newlen += sprintf(new + newlen, "%o %s", mode, path);
 106                        new[newlen++] = '\0';
 107                        hashcpy((unsigned char*)new + newlen, (unsigned char *) buffer + len - 20);
 108                        newlen += 20;
 109
 110                        used += len;
 111                        size -= len;
 112                        buffer = (char *) buffer + len;
 113                        continue;
 114                }
 115
 116                newlen += sprintf(new + newlen, "%o %.*s", S_IFDIR, (int)(slash - path), path);
 117                new[newlen++] = 0;
 118                sha1 = (unsigned char *)(new + newlen);
 119                newlen += 20;
 120
 121                len = write_subdirectory(buffer, size, origpath, slash-origpath+1, sha1);
 122
 123                used += len;
 124                size -= len;
 125                buffer = (char *) buffer + len;
 126        }
 127
 128        write_sha1_file(new, newlen, tree_type, result_sha1);
 129        free(new);
 130        return used;
 131}
 132
 133static void convert_tree(void *buffer, unsigned long size, unsigned char *result_sha1)
 134{
 135        void *orig_buffer = buffer;
 136        unsigned long orig_size = size;
 137
 138        while (size) {
 139                int len = 1+strlen(buffer);
 140
 141                convert_binary_sha1((char *) buffer + len);
 142
 143                len += 20;
 144                if (len > size)
 145                        die("corrupt tree object");
 146                size -= len;
 147                buffer = (char *) buffer + len;
 148        }
 149
 150        write_subdirectory(orig_buffer, orig_size, "", 0, result_sha1);
 151}
 152
 153static unsigned long parse_oldstyle_date(const char *buf)
 154{
 155        char c, *p;
 156        char buffer[100];
 157        struct tm tm;
 158        const char *formats[] = {
 159                "%c",
 160                "%a %b %d %T",
 161                "%Z",
 162                "%Y",
 163                " %Y",
 164                NULL
 165        };
 166        /* We only ever did two timezones in the bad old format .. */
 167        const char *timezones[] = {
 168                "PDT", "PST", "CEST", NULL
 169        };
 170        const char **fmt = formats;
 171
 172        p = buffer;
 173        while (isspace(c = *buf))
 174                buf++;
 175        while ((c = *buf++) != '\n')
 176                *p++ = c;
 177        *p++ = 0;
 178        buf = buffer;
 179        memset(&tm, 0, sizeof(tm));
 180        do {
 181                const char *next = strptime(buf, *fmt, &tm);
 182                if (next) {
 183                        if (!*next)
 184                                return mktime(&tm);
 185                        buf = next;
 186                } else {
 187                        const char **p = timezones;
 188                        while (isspace(*buf))
 189                                buf++;
 190                        while (*p) {
 191                                if (!memcmp(buf, *p, strlen(*p))) {
 192                                        buf += strlen(*p);
 193                                        break;
 194                                }
 195                                p++;
 196                        }
 197                }
 198                fmt++;
 199        } while (*buf && *fmt);
 200        printf("left: %s\n", buf);
 201        return mktime(&tm);                             
 202}
 203
 204static int convert_date_line(char *dst, void **buf, unsigned long *sp)
 205{
 206        unsigned long size = *sp;
 207        char *line = *buf;
 208        char *next = strchr(line, '\n');
 209        char *date = strchr(line, '>');
 210        int len;
 211
 212        if (!next || !date)
 213                die("missing or bad author/committer line %s", line);
 214        next++; date += 2;
 215
 216        *buf = next;
 217        *sp = size - (next - line);
 218
 219        len = date - line;
 220        memcpy(dst, line, len);
 221        dst += len;
 222
 223        /* Is it already in new format? */
 224        if (isdigit(*date)) {
 225                int datelen = next - date;
 226                memcpy(dst, date, datelen);
 227                return len + datelen;
 228        }
 229
 230        /*
 231         * Hacky hacky: one of the sparse old-style commits does not have
 232         * any date at all, but we can fake it by using the committer date.
 233         */
 234        if (*date == '\n' && strchr(next, '>'))
 235                date = strchr(next, '>')+2;
 236
 237        return len + sprintf(dst, "%lu -0700\n", parse_oldstyle_date(date));
 238}
 239
 240static void convert_date(void *buffer, unsigned long size, unsigned char *result_sha1)
 241{
 242        char *new = xmalloc(size + 100);
 243        unsigned long newlen = 0;
 244
 245        /* "tree <sha1>\n" */
 246        memcpy(new + newlen, buffer, 46);
 247        newlen += 46;
 248        buffer = (char *) buffer + 46;
 249        size -= 46;
 250
 251        /* "parent <sha1>\n" */
 252        while (!memcmp(buffer, "parent ", 7)) {
 253                memcpy(new + newlen, buffer, 48);
 254                newlen += 48;
 255                buffer = (char *) buffer + 48;
 256                size -= 48;
 257        }
 258
 259        /* "author xyz <xyz> date" */
 260        newlen += convert_date_line(new + newlen, &buffer, &size);
 261        /* "committer xyz <xyz> date" */
 262        newlen += convert_date_line(new + newlen, &buffer, &size);
 263
 264        /* Rest */
 265        memcpy(new + newlen, buffer, size);
 266        newlen += size;
 267
 268        write_sha1_file(new, newlen, commit_type, result_sha1);
 269        free(new);
 270}
 271
 272static void convert_commit(void *buffer, unsigned long size, unsigned char *result_sha1)
 273{
 274        void *orig_buffer = buffer;
 275        unsigned long orig_size = size;
 276
 277        if (memcmp(buffer, "tree ", 5))
 278                die("Bad commit '%s'", (char*) buffer);
 279        convert_ascii_sha1((char *) buffer + 5);
 280        buffer = (char *) buffer + 46;    /* "tree " + "hex sha1" + "\n" */
 281        while (!memcmp(buffer, "parent ", 7)) {
 282                convert_ascii_sha1((char *) buffer + 7);
 283                buffer = (char *) buffer + 48;
 284        }
 285        convert_date(orig_buffer, orig_size, result_sha1);
 286}
 287
 288static struct entry * convert_entry(unsigned char *sha1)
 289{
 290        struct entry *entry = lookup_entry(sha1);
 291        char type[20];
 292        void *buffer, *data;
 293        unsigned long size;
 294
 295        if (entry->converted)
 296                return entry;
 297        data = read_sha1_file(sha1, type, &size);
 298        if (!data)
 299                die("unable to read object %s", sha1_to_hex(sha1));
 300
 301        buffer = xmalloc(size);
 302        memcpy(buffer, data, size);
 303
 304        if (!strcmp(type, blob_type)) {
 305                write_sha1_file(buffer, size, blob_type, entry->new_sha1);
 306        } else if (!strcmp(type, tree_type))
 307                convert_tree(buffer, size, entry->new_sha1);
 308        else if (!strcmp(type, commit_type))
 309                convert_commit(buffer, size, entry->new_sha1);
 310        else
 311                die("unknown object type '%s' in %s", type, sha1_to_hex(sha1));
 312        entry->converted = 1;
 313        free(buffer);
 314        free(data);
 315        return entry;
 316}
 317
 318int main(int argc, char **argv)
 319{
 320        unsigned char sha1[20];
 321        struct entry *entry;
 322
 323        setup_git_directory();
 324
 325        if (argc != 2)
 326                usage("git-convert-objects <sha1>");
 327        if (get_sha1(argv[1], sha1))
 328                die("Not a valid object name %s", argv[1]);
 329
 330        entry = convert_entry(sha1);
 331        printf("new sha1: %s\n", sha1_to_hex(entry->new_sha1));
 332        return 0;
 333}