1 /***************************************************************************
2 * _ _ ____ _
3 * Project ___| | | | _ \| |
4 * / __| | | | |_) | |
5 * | (__| |_| | _ <| |___
6 * \___|\___/|_| \_\_____|
7 *
8 * Copyright (C) 1998 - 2018, Daniel Stenberg, <daniel@haxx.se>, et al.
9 *
10 * This software is licensed as described in the file COPYING, which
11 * you should have received as part of this distribution. The terms
12 * are also available at https://curl.haxx.se/docs/copyright.html.
13 *
14 * You may opt to use, copy, modify, merge, publish, distribute and/or sell
15 * copies of the Software, and permit persons to whom the Software is
16 * furnished to do so, under the terms of the COPYING file.
17 *
18 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
19 * KIND, either express or implied.
20 *
21 ***************************************************************************/
22 /*
23 A brief summary of the date string formats this parser groks:
24
25 RFC 2616 3.3.1
26
27 Sun, 06 Nov 1994 08:49:37 GMT ; RFC 822, updated by RFC 1123
28 Sunday, 06-Nov-94 08:49:37 GMT ; RFC 850, obsoleted by RFC 1036
29 Sun Nov 6 08:49:37 1994 ; ANSI C's asctime() format
30
31 we support dates without week day name:
32
33 06 Nov 1994 08:49:37 GMT
34 06-Nov-94 08:49:37 GMT
35 Nov 6 08:49:37 1994
36
37 without the time zone:
38
39 06 Nov 1994 08:49:37
40 06-Nov-94 08:49:37
41
42 weird order:
43
44 1994 Nov 6 08:49:37 (GNU date fails)
45 GMT 08:49:37 06-Nov-94 Sunday
46 94 6 Nov 08:49:37 (GNU date fails)
47
48 time left out:
49
50 1994 Nov 6
51 06-Nov-94
52 Sun Nov 6 94
53
54 unusual separators:
55
56 1994.Nov.6
57 Sun/Nov/6/94/GMT
58
59 commonly used time zone names:
60
61 Sun, 06 Nov 1994 08:49:37 CET
62 06 Nov 1994 08:49:37 EST
63
64 time zones specified using RFC822 style:
65
66 Sun, 12 Sep 2004 15:05:58 -0700
67 Sat, 11 Sep 2004 21:32:11 +0200
68
69 compact numerical date strings:
70
71 20040912 15:05:58 -0700
72 20040911 +0200
73
74 */
75
76 #include "curl_setup.h"
77
78 #include <limits.h>
79
80 #include <curl/curl.h>
81 #include "strcase.h"
82 #include "warnless.h"
83 #include "parsedate.h"
84
85 const char * const Curl_wkday[] =
86 {"Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"};
87 static const char * const weekday[] =
88 { "Monday", "Tuesday", "Wednesday", "Thursday",
89 "Friday", "Saturday", "Sunday" };
90 const char * const Curl_month[]=
91 { "Jan", "Feb", "Mar", "Apr", "May", "Jun",
92 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" };
93
94 struct tzinfo {
95 char name[5];
96 int offset; /* +/- in minutes */
97 };
98
99 /*
100 * parsedate()
101 *
102 * Returns:
103 *
104 * PARSEDATE_OK - a fine conversion
105 * PARSEDATE_FAIL - failed to convert
106 * PARSEDATE_LATER - time overflow at the far end of time_t
107 * PARSEDATE_SOONER - time underflow at the low end of time_t
108 */
109
110 static int parsedate(const char *date, time_t *output);
111
112 #define PARSEDATE_OK 0
113 #define PARSEDATE_FAIL -1
114 #define PARSEDATE_LATER 1
115 #define PARSEDATE_SOONER 2
116
117 /* Here's a bunch of frequently used time zone names. These were supported
118 by the old getdate parser. */
119 #define tDAYZONE -60 /* offset for daylight savings time */
120 static const struct tzinfo tz[]= {
121 {"GMT", 0}, /* Greenwich Mean */
122 {"UT", 0}, /* Universal Time */
123 {"UTC", 0}, /* Universal (Coordinated) */
124 {"WET", 0}, /* Western European */
125 {"BST", 0 tDAYZONE}, /* British Summer */
126 {"WAT", 60}, /* West Africa */
127 {"AST", 240}, /* Atlantic Standard */
128 {"ADT", 240 tDAYZONE}, /* Atlantic Daylight */
129 {"EST", 300}, /* Eastern Standard */
130 {"EDT", 300 tDAYZONE}, /* Eastern Daylight */
131 {"CST", 360}, /* Central Standard */
132 {"CDT", 360 tDAYZONE}, /* Central Daylight */
133 {"MST", 420}, /* Mountain Standard */
134 {"MDT", 420 tDAYZONE}, /* Mountain Daylight */
135 {"PST", 480}, /* Pacific Standard */
136 {"PDT", 480 tDAYZONE}, /* Pacific Daylight */
137 {"YST", 540}, /* Yukon Standard */
138 {"YDT", 540 tDAYZONE}, /* Yukon Daylight */
139 {"HST", 600}, /* Hawaii Standard */
140 {"HDT", 600 tDAYZONE}, /* Hawaii Daylight */
141 {"CAT", 600}, /* Central Alaska */
142 {"AHST", 600}, /* Alaska-Hawaii Standard */
143 {"NT", 660}, /* Nome */
144 {"IDLW", 720}, /* International Date Line West */
145 {"CET", -60}, /* Central European */
146 {"MET", -60}, /* Middle European */
147 {"MEWT", -60}, /* Middle European Winter */
148 {"MEST", -60 tDAYZONE}, /* Middle European Summer */
149 {"CEST", -60 tDAYZONE}, /* Central European Summer */
150 {"MESZ", -60 tDAYZONE}, /* Middle European Summer */
151 {"FWT", -60}, /* French Winter */
152 {"FST", -60 tDAYZONE}, /* French Summer */
153 {"EET", -120}, /* Eastern Europe, USSR Zone 1 */
154 {"WAST", -420}, /* West Australian Standard */
155 {"WADT", -420 tDAYZONE}, /* West Australian Daylight */
156 {"CCT", -480}, /* China Coast, USSR Zone 7 */
157 {"JST", -540}, /* Japan Standard, USSR Zone 8 */
158 {"EAST", -600}, /* Eastern Australian Standard */
159 {"EADT", -600 tDAYZONE}, /* Eastern Australian Daylight */
160 {"GST", -600}, /* Guam Standard, USSR Zone 9 */
161 {"NZT", -720}, /* New Zealand */
162 {"NZST", -720}, /* New Zealand Standard */
163 {"NZDT", -720 tDAYZONE}, /* New Zealand Daylight */
164 {"IDLE", -720}, /* International Date Line East */
165 /* Next up: Military timezone names. RFC822 allowed these, but (as noted in
166 RFC 1123) had their signs wrong. Here we use the correct signs to match
167 actual military usage.
168 */
169 {"A", 1 * 60}, /* Alpha */
170 {"B", 2 * 60}, /* Bravo */
171 {"C", 3 * 60}, /* Charlie */
172 {"D", 4 * 60}, /* Delta */
173 {"E", 5 * 60}, /* Echo */
174 {"F", 6 * 60}, /* Foxtrot */
175 {"G", 7 * 60}, /* Golf */
176 {"H", 8 * 60}, /* Hotel */
177 {"I", 9 * 60}, /* India */
178 /* "J", Juliet is not used as a timezone, to indicate the observer's local
179 time */
180 {"K", 10 * 60}, /* Kilo */
181 {"L", 11 * 60}, /* Lima */
182 {"M", 12 * 60}, /* Mike */
183 {"N", -1 * 60}, /* November */
184 {"O", -2 * 60}, /* Oscar */
185 {"P", -3 * 60}, /* Papa */
186 {"Q", -4 * 60}, /* Quebec */
187 {"R", -5 * 60}, /* Romeo */
188 {"S", -6 * 60}, /* Sierra */
189 {"T", -7 * 60}, /* Tango */
190 {"U", -8 * 60}, /* Uniform */
191 {"V", -9 * 60}, /* Victor */
192 {"W", -10 * 60}, /* Whiskey */
193 {"X", -11 * 60}, /* X-ray */
194 {"Y", -12 * 60}, /* Yankee */
195 {"Z", 0}, /* Zulu, zero meridian, a.k.a. UTC */
196 };
197
198 /* returns:
199 -1 no day
200 0 monday - 6 sunday
201 */
202
checkday(const char * check,size_t len)203 static int checkday(const char *check, size_t len)
204 {
205 int i;
206 const char * const *what;
207 bool found = FALSE;
208 if(len > 3)
209 what = &weekday[0];
210 else
211 what = &Curl_wkday[0];
212 for(i = 0; i<7; i++) {
213 if(strcasecompare(check, what[0])) {
214 found = TRUE;
215 break;
216 }
217 what++;
218 }
219 return found?i:-1;
220 }
221
checkmonth(const char * check)222 static int checkmonth(const char *check)
223 {
224 int i;
225 const char * const *what;
226 bool found = FALSE;
227
228 what = &Curl_month[0];
229 for(i = 0; i<12; i++) {
230 if(strcasecompare(check, what[0])) {
231 found = TRUE;
232 break;
233 }
234 what++;
235 }
236 return found?i:-1; /* return the offset or -1, no real offset is -1 */
237 }
238
239 /* return the time zone offset between GMT and the input one, in number
240 of seconds or -1 if the timezone wasn't found/legal */
241
checktz(const char * check)242 static int checktz(const char *check)
243 {
244 unsigned int i;
245 const struct tzinfo *what;
246 bool found = FALSE;
247
248 what = tz;
249 for(i = 0; i< sizeof(tz)/sizeof(tz[0]); i++) {
250 if(strcasecompare(check, what->name)) {
251 found = TRUE;
252 break;
253 }
254 what++;
255 }
256 return found?what->offset*60:-1;
257 }
258
skip(const char ** date)259 static void skip(const char **date)
260 {
261 /* skip everything that aren't letters or digits */
262 while(**date && !ISALNUM(**date))
263 (*date)++;
264 }
265
266 enum assume {
267 DATE_MDAY,
268 DATE_YEAR,
269 DATE_TIME
270 };
271
272 /* this is a clone of 'struct tm' but with all fields we don't need or use
273 cut out */
274 struct my_tm {
275 int tm_sec;
276 int tm_min;
277 int tm_hour;
278 int tm_mday;
279 int tm_mon;
280 int tm_year; /* full year */
281 };
282
283 /* struct tm to time since epoch in GMT time zone.
284 * This is similar to the standard mktime function but for GMT only, and
285 * doesn't suffer from the various bugs and portability problems that
286 * some systems' implementations have.
287 *
288 * Returns 0 on success, otherwise non-zero.
289 */
my_timegm(struct my_tm * tm,time_t * t)290 static void my_timegm(struct my_tm *tm, time_t *t)
291 {
292 static const int month_days_cumulative [12] =
293 { 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334 };
294 int month, year, leap_days;
295
296 year = tm->tm_year;
297 month = tm->tm_mon;
298 if(month < 0) {
299 year += (11 - month) / 12;
300 month = 11 - (11 - month) % 12;
301 }
302 else if(month >= 12) {
303 year -= month / 12;
304 month = month % 12;
305 }
306
307 leap_days = year - (tm->tm_mon <= 1);
308 leap_days = ((leap_days / 4) - (leap_days / 100) + (leap_days / 400)
309 - (1969 / 4) + (1969 / 100) - (1969 / 400));
310
311 *t = ((((time_t) (year - 1970) * 365
312 + leap_days + month_days_cumulative[month] + tm->tm_mday - 1) * 24
313 + tm->tm_hour) * 60 + tm->tm_min) * 60 + tm->tm_sec;
314 }
315
316 /*
317 * parsedate()
318 *
319 * Returns:
320 *
321 * PARSEDATE_OK - a fine conversion
322 * PARSEDATE_FAIL - failed to convert
323 * PARSEDATE_LATER - time overflow at the far end of time_t
324 * PARSEDATE_SOONER - time underflow at the low end of time_t
325 */
326
parsedate(const char * date,time_t * output)327 static int parsedate(const char *date, time_t *output)
328 {
329 time_t t = 0;
330 int wdaynum = -1; /* day of the week number, 0-6 (mon-sun) */
331 int monnum = -1; /* month of the year number, 0-11 */
332 int mdaynum = -1; /* day of month, 1 - 31 */
333 int hournum = -1;
334 int minnum = -1;
335 int secnum = -1;
336 int yearnum = -1;
337 int tzoff = -1;
338 struct my_tm tm;
339 enum assume dignext = DATE_MDAY;
340 const char *indate = date; /* save the original pointer */
341 int part = 0; /* max 6 parts */
342
343 while(*date && (part < 6)) {
344 bool found = FALSE;
345
346 skip(&date);
347
348 if(ISALPHA(*date)) {
349 /* a name coming up */
350 char buf[32]="";
351 size_t len;
352 if(sscanf(date, "%31[ABCDEFGHIJKLMNOPQRSTUVWXYZ"
353 "abcdefghijklmnopqrstuvwxyz]", buf))
354 len = strlen(buf);
355 else
356 len = 0;
357
358 if(wdaynum == -1) {
359 wdaynum = checkday(buf, len);
360 if(wdaynum != -1)
361 found = TRUE;
362 }
363 if(!found && (monnum == -1)) {
364 monnum = checkmonth(buf);
365 if(monnum != -1)
366 found = TRUE;
367 }
368
369 if(!found && (tzoff == -1)) {
370 /* this just must be a time zone string */
371 tzoff = checktz(buf);
372 if(tzoff != -1)
373 found = TRUE;
374 }
375
376 if(!found)
377 return PARSEDATE_FAIL; /* bad string */
378
379 date += len;
380 }
381 else if(ISDIGIT(*date)) {
382 /* a digit */
383 int val;
384 char *end;
385 int len = 0;
386 if((secnum == -1) &&
387 (3 == sscanf(date, "%02d:%02d:%02d%n",
388 &hournum, &minnum, &secnum, &len))) {
389 /* time stamp! */
390 date += len;
391 }
392 else if((secnum == -1) &&
393 (2 == sscanf(date, "%02d:%02d%n", &hournum, &minnum, &len))) {
394 /* time stamp without seconds */
395 date += len;
396 secnum = 0;
397 }
398 else {
399 long lval;
400 int error;
401 int old_errno;
402
403 old_errno = errno;
404 errno = 0;
405 lval = strtol(date, &end, 10);
406 error = errno;
407 if(errno != old_errno)
408 errno = old_errno;
409
410 if(error)
411 return PARSEDATE_FAIL;
412
413 #if LONG_MAX != INT_MAX
414 if((lval > (long)INT_MAX) || (lval < (long)INT_MIN))
415 return PARSEDATE_FAIL;
416 #endif
417
418 val = curlx_sltosi(lval);
419
420 if((tzoff == -1) &&
421 ((end - date) == 4) &&
422 (val <= 1400) &&
423 (indate< date) &&
424 ((date[-1] == '+' || date[-1] == '-'))) {
425 /* four digits and a value less than or equal to 1400 (to take into
426 account all sorts of funny time zone diffs) and it is preceded
427 with a plus or minus. This is a time zone indication. 1400 is
428 picked since +1300 is frequently used and +1400 is mentioned as
429 an edge number in the document "ISO C 200X Proposal: Timezone
430 Functions" at http://david.tribble.com/text/c0xtimezone.html If
431 anyone has a more authoritative source for the exact maximum time
432 zone offsets, please speak up! */
433 found = TRUE;
434 tzoff = (val/100 * 60 + val%100)*60;
435
436 /* the + and - prefix indicates the local time compared to GMT,
437 this we need their reversed math to get what we want */
438 tzoff = date[-1]=='+'?-tzoff:tzoff;
439 }
440
441 if(((end - date) == 8) &&
442 (yearnum == -1) &&
443 (monnum == -1) &&
444 (mdaynum == -1)) {
445 /* 8 digits, no year, month or day yet. This is YYYYMMDD */
446 found = TRUE;
447 yearnum = val/10000;
448 monnum = (val%10000)/100-1; /* month is 0 - 11 */
449 mdaynum = val%100;
450 }
451
452 if(!found && (dignext == DATE_MDAY) && (mdaynum == -1)) {
453 if((val > 0) && (val<32)) {
454 mdaynum = val;
455 found = TRUE;
456 }
457 dignext = DATE_YEAR;
458 }
459
460 if(!found && (dignext == DATE_YEAR) && (yearnum == -1)) {
461 yearnum = val;
462 found = TRUE;
463 if(yearnum < 100) {
464 if(yearnum > 70)
465 yearnum += 1900;
466 else
467 yearnum += 2000;
468 }
469 if(mdaynum == -1)
470 dignext = DATE_MDAY;
471 }
472
473 if(!found)
474 return PARSEDATE_FAIL;
475
476 date = end;
477 }
478 }
479
480 part++;
481 }
482
483 if(-1 == secnum)
484 secnum = minnum = hournum = 0; /* no time, make it zero */
485
486 if((-1 == mdaynum) ||
487 (-1 == monnum) ||
488 (-1 == yearnum))
489 /* lacks vital info, fail */
490 return PARSEDATE_FAIL;
491
492 #ifdef HAVE_TIME_T_UNSIGNED
493 if(yearnum < 1970) {
494 /* only positive numbers cannot return earlier */
495 *output = TIME_T_MIN;
496 return PARSEDATE_SOONER;
497 }
498 #endif
499
500 #if (SIZEOF_TIME_T < 5)
501
502 #ifdef HAVE_TIME_T_UNSIGNED
503 /* an unsigned 32 bit time_t can only hold dates to 2106 */
504 if(yearnum > 2105) {
505 *output = TIME_T_MAX;
506 return PARSEDATE_LATER;
507 }
508 #else
509 /* a signed 32 bit time_t can only hold dates to the beginning of 2038 */
510 if(yearnum > 2037) {
511 *output = TIME_T_MAX;
512 return PARSEDATE_LATER;
513 }
514 if(yearnum < 1903) {
515 *output = TIME_T_MIN;
516 return PARSEDATE_SOONER;
517 }
518 #endif
519
520 #else
521 /* The Gregorian calendar was introduced 1582 */
522 if(yearnum < 1583)
523 return PARSEDATE_FAIL;
524 #endif
525
526 if((mdaynum > 31) || (monnum > 11) ||
527 (hournum > 23) || (minnum > 59) || (secnum > 60))
528 return PARSEDATE_FAIL; /* clearly an illegal date */
529
530 tm.tm_sec = secnum;
531 tm.tm_min = minnum;
532 tm.tm_hour = hournum;
533 tm.tm_mday = mdaynum;
534 tm.tm_mon = monnum;
535 tm.tm_year = yearnum;
536
537 /* my_timegm() returns a time_t. time_t is often 32 bits, sometimes even on
538 architectures that feature 64 bit 'long' but ultimately time_t is the
539 correct data type to use.
540 */
541 my_timegm(&tm, &t);
542
543 /* Add the time zone diff between local time zone and GMT. */
544 if(tzoff == -1)
545 tzoff = 0;
546
547 if((tzoff > 0) && (t > TIME_T_MAX - tzoff)) {
548 *output = TIME_T_MAX;
549 return PARSEDATE_LATER; /* time_t overflow */
550 }
551
552 t += tzoff;
553
554 *output = t;
555
556 return PARSEDATE_OK;
557 }
558
curl_getdate(const char * p,const time_t * now)559 time_t curl_getdate(const char *p, const time_t *now)
560 {
561 time_t parsed = -1;
562 int rc = parsedate(p, &parsed);
563 (void)now; /* legacy argument from the past that we ignore */
564
565 if(rc == PARSEDATE_OK) {
566 if(parsed == -1)
567 /* avoid returning -1 for a working scenario */
568 parsed++;
569 return parsed;
570 }
571 /* everything else is fail */
572 return -1;
573 }
574
575 /*
576 * Curl_gmtime() is a gmtime() replacement for portability. Do not use the
577 * gmtime_r() or gmtime() functions anywhere else but here.
578 *
579 */
580
Curl_gmtime(time_t intime,struct tm * store)581 CURLcode Curl_gmtime(time_t intime, struct tm *store)
582 {
583 const struct tm *tm;
584 #ifdef HAVE_GMTIME_R
585 /* thread-safe version */
586 tm = (struct tm *)gmtime_r(&intime, store);
587 #else
588 tm = gmtime(&intime);
589 if(tm)
590 *store = *tm; /* copy the pointed struct to the local copy */
591 #endif
592
593 if(!tm)
594 return CURLE_BAD_FUNCTION_ARGUMENT;
595 return CURLE_OK;
596 }
597