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