1/***************************************************************************
2 *                                  _   _ ____  _
3 *  Project                     ___| | | |  _ \| |
4 *                             / __| | | | |_) | |
5 *                            | (__| |_| |  _ <| |___
6 *                             \___|\___/|_| \_\_____|
7 *
8 * Copyright (C) 1998 - 2014, 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 http://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#ifdef HAVE_LIMITS_H
79#include <limits.h>
80#endif
81
82#include <curl/curl.h>
83#include "rawstr.h"
84#include "warnless.h"
85#include "parsedate.h"
86
87const char * const Curl_wkday[] =
88{"Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"};
89static const char * const weekday[] =
90{ "Monday", "Tuesday", "Wednesday", "Thursday",
91  "Friday", "Saturday", "Sunday" };
92const char * const Curl_month[]=
93{ "Jan", "Feb", "Mar", "Apr", "May", "Jun",
94  "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" };
95
96struct tzinfo {
97  char name[5];
98  int offset; /* +/- in minutes */
99};
100
101/*
102 * parsedate()
103 *
104 * Returns:
105 *
106 * PARSEDATE_OK     - a fine conversion
107 * PARSEDATE_FAIL   - failed to convert
108 * PARSEDATE_LATER  - time overflow at the far end of time_t
109 * PARSEDATE_SOONER - time underflow at the low end of time_t
110 */
111
112static int parsedate(const char *date, time_t *output);
113
114#define PARSEDATE_OK     0
115#define PARSEDATE_FAIL   -1
116#define PARSEDATE_LATER  1
117#define PARSEDATE_SOONER 2
118
119/* Here's a bunch of frequently used time zone names. These were supported
120   by the old getdate parser. */
121#define tDAYZONE -60       /* offset for daylight savings time */
122static const struct tzinfo tz[]= {
123  {"GMT", 0},              /* Greenwich Mean */
124  {"UTC", 0},              /* Universal (Coordinated) */
125  {"WET", 0},              /* Western European */
126  {"BST", 0 tDAYZONE},     /* British Summer */
127  {"WAT", 60},             /* West Africa */
128  {"AST", 240},            /* Atlantic Standard */
129  {"ADT", 240 tDAYZONE},   /* Atlantic Daylight */
130  {"EST", 300},            /* Eastern Standard */
131  {"EDT", 300 tDAYZONE},   /* Eastern Daylight */
132  {"CST", 360},            /* Central Standard */
133  {"CDT", 360 tDAYZONE},   /* Central Daylight */
134  {"MST", 420},            /* Mountain Standard */
135  {"MDT", 420 tDAYZONE},   /* Mountain Daylight */
136  {"PST", 480},            /* Pacific Standard */
137  {"PDT", 480 tDAYZONE},   /* Pacific Daylight */
138  {"YST", 540},            /* Yukon Standard */
139  {"YDT", 540 tDAYZONE},   /* Yukon Daylight */
140  {"HST", 600},            /* Hawaii Standard */
141  {"HDT", 600 tDAYZONE},   /* Hawaii Daylight */
142  {"CAT", 600},            /* Central Alaska */
143  {"AHST", 600},           /* Alaska-Hawaii Standard */
144  {"NT",  660},            /* Nome */
145  {"IDLW", 720},           /* International Date Line West */
146  {"CET", -60},            /* Central European */
147  {"MET", -60},            /* Middle European */
148  {"MEWT", -60},           /* Middle European Winter */
149  {"MEST", -60 tDAYZONE},  /* Middle European Summer */
150  {"CEST", -60 tDAYZONE},  /* Central European Summer */
151  {"MESZ", -60 tDAYZONE},  /* Middle European Summer */
152  {"FWT", -60},            /* French Winter */
153  {"FST", -60 tDAYZONE},   /* French Summer */
154  {"EET", -120},           /* Eastern Europe, USSR Zone 1 */
155  {"WAST", -420},          /* West Australian Standard */
156  {"WADT", -420 tDAYZONE}, /* West Australian Daylight */
157  {"CCT", -480},           /* China Coast, USSR Zone 7 */
158  {"JST", -540},           /* Japan Standard, USSR Zone 8 */
159  {"EAST", -600},          /* Eastern Australian Standard */
160  {"EADT", -600 tDAYZONE}, /* Eastern Australian Daylight */
161  {"GST", -600},           /* Guam Standard, USSR Zone 9 */
162  {"NZT", -720},           /* New Zealand */
163  {"NZST", -720},          /* New Zealand Standard */
164  {"NZDT", -720 tDAYZONE}, /* New Zealand Daylight */
165  {"IDLE", -720},          /* International Date Line East */
166  /* Next up: Military timezone names. RFC822 allowed these, but (as noted in
167     RFC 1123) had their signs wrong. Here we use the correct signs to match
168     actual military usage.
169   */
170  {"A",  +1 * 60},         /* Alpha */
171  {"B",  +2 * 60},         /* Bravo */
172  {"C",  +3 * 60},         /* Charlie */
173  {"D",  +4 * 60},         /* Delta */
174  {"E",  +5 * 60},         /* Echo */
175  {"F",  +6 * 60},         /* Foxtrot */
176  {"G",  +7 * 60},         /* Golf */
177  {"H",  +8 * 60},         /* Hotel */
178  {"I",  +9 * 60},         /* India */
179  /* "J", Juliet is not used as a timezone, to indicate the observer's local
180     time */
181  {"K", +10 * 60},         /* Kilo */
182  {"L", +11 * 60},         /* Lima */
183  {"M", +12 * 60},         /* Mike */
184  {"N",  -1 * 60},         /* November */
185  {"O",  -2 * 60},         /* Oscar */
186  {"P",  -3 * 60},         /* Papa */
187  {"Q",  -4 * 60},         /* Quebec */
188  {"R",  -5 * 60},         /* Romeo */
189  {"S",  -6 * 60},         /* Sierra */
190  {"T",  -7 * 60},         /* Tango */
191  {"U",  -8 * 60},         /* Uniform */
192  {"V",  -9 * 60},         /* Victor */
193  {"W", -10 * 60},         /* Whiskey */
194  {"X", -11 * 60},         /* X-ray */
195  {"Y", -12 * 60},         /* Yankee */
196  {"Z", 0},                /* Zulu, zero meridian, a.k.a. UTC */
197};
198
199/* returns:
200   -1 no day
201   0 monday - 6 sunday
202*/
203
204static int checkday(const char *check, size_t len)
205{
206  int i;
207  const char * const *what;
208  bool found= FALSE;
209  if(len > 3)
210    what = &weekday[0];
211  else
212    what = &Curl_wkday[0];
213  for(i=0; i<7; i++) {
214    if(Curl_raw_equal(check, what[0])) {
215      found=TRUE;
216      break;
217    }
218    what++;
219  }
220  return found?i:-1;
221}
222
223static int checkmonth(const char *check)
224{
225  int i;
226  const char * const *what;
227  bool found= FALSE;
228
229  what = &Curl_month[0];
230  for(i=0; i<12; i++) {
231    if(Curl_raw_equal(check, what[0])) {
232      found=TRUE;
233      break;
234    }
235    what++;
236  }
237  return found?i:-1; /* return the offset or -1, no real offset is -1 */
238}
239
240/* return the time zone offset between GMT and the input one, in number
241   of seconds or -1 if the timezone wasn't found/legal */
242
243static int checktz(const char *check)
244{
245  unsigned int i;
246  const struct tzinfo *what;
247  bool found= FALSE;
248
249  what = tz;
250  for(i=0; i< sizeof(tz)/sizeof(tz[0]); i++) {
251    if(Curl_raw_equal(check, what->name)) {
252      found=TRUE;
253      break;
254    }
255    what++;
256  }
257  return found?what->offset*60:-1;
258}
259
260static void skip(const char **date)
261{
262  /* skip everything that aren't letters or digits */
263  while(**date && !ISALNUM(**date))
264    (*date)++;
265}
266
267enum assume {
268  DATE_MDAY,
269  DATE_YEAR,
270  DATE_TIME
271};
272
273/* this is a clone of 'struct tm' but with all fields we don't need or use
274   cut out */
275struct my_tm {
276  int tm_sec;
277  int tm_min;
278  int tm_hour;
279  int tm_mday;
280  int tm_mon;
281  int tm_year;
282};
283
284/* struct tm to time since epoch in GMT time zone.
285 * This is similar to the standard mktime function but for GMT only, and
286 * doesn't suffer from the various bugs and portability problems that
287 * some systems' implementations have.
288 */
289static time_t my_timegm(struct my_tm *tm)
290{
291  static const int month_days_cumulative [12] =
292    { 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334 };
293  int month, year, leap_days;
294
295  if(tm->tm_year < 70)
296    /* we don't support years before 1970 as they will cause this function
297       to return a negative value */
298    return -1;
299
300  year = tm->tm_year + 1900;
301  month = tm->tm_mon;
302  if(month < 0) {
303    year += (11 - month) / 12;
304    month = 11 - (11 - month) % 12;
305  }
306  else if(month >= 12) {
307    year -= month / 12;
308    month = month % 12;
309  }
310
311  leap_days = year - (tm->tm_mon <= 1);
312  leap_days = ((leap_days / 4) - (leap_days / 100) + (leap_days / 400)
313               - (1969 / 4) + (1969 / 100) - (1969 / 400));
314
315  return ((((time_t) (year - 1970) * 365
316            + leap_days + month_days_cumulative [month] + tm->tm_mday - 1) * 24
317           + tm->tm_hour) * 60 + tm->tm_min) * 60 + tm->tm_sec;
318}
319
320/*
321 * parsedate()
322 *
323 * Returns:
324 *
325 * PARSEDATE_OK     - a fine conversion
326 * PARSEDATE_FAIL   - failed to convert
327 * PARSEDATE_LATER  - time overflow at the far end of time_t
328 * PARSEDATE_SOONER - time underflow at the low end of time_t
329 */
330
331static int parsedate(const char *date, time_t *output)
332{
333  time_t t = 0;
334  int wdaynum=-1;  /* day of the week number, 0-6 (mon-sun) */
335  int monnum=-1;   /* month of the year number, 0-11 */
336  int mdaynum=-1; /* day of month, 1 - 31 */
337  int hournum=-1;
338  int minnum=-1;
339  int secnum=-1;
340  int yearnum=-1;
341  int tzoff=-1;
342  struct my_tm tm;
343  enum assume dignext = DATE_MDAY;
344  const char *indate = date; /* save the original pointer */
345  int part = 0; /* max 6 parts */
346
347  while(*date && (part < 6)) {
348    bool found=FALSE;
349
350    skip(&date);
351
352    if(ISALPHA(*date)) {
353      /* a name coming up */
354      char buf[32]="";
355      size_t len;
356      if(sscanf(date, "%31[ABCDEFGHIJKLMNOPQRSTUVWXYZ"
357                          "abcdefghijklmnopqrstuvwxyz]", buf))
358        len = strlen(buf);
359      else
360        len = 0;
361
362      if(wdaynum == -1) {
363        wdaynum = checkday(buf, len);
364        if(wdaynum != -1)
365          found = TRUE;
366      }
367      if(!found && (monnum == -1)) {
368        monnum = checkmonth(buf);
369        if(monnum != -1)
370          found = TRUE;
371      }
372
373      if(!found && (tzoff == -1)) {
374        /* this just must be a time zone string */
375        tzoff = checktz(buf);
376        if(tzoff != -1)
377          found = TRUE;
378      }
379
380      if(!found)
381        return PARSEDATE_FAIL; /* bad string */
382
383      date += len;
384    }
385    else if(ISDIGIT(*date)) {
386      /* a digit */
387      int val;
388      char *end;
389      if((secnum == -1) &&
390         (3 == sscanf(date, "%02d:%02d:%02d", &hournum, &minnum, &secnum))) {
391        /* time stamp! */
392        date += 8;
393      }
394      else if((secnum == -1) &&
395              (2 == sscanf(date, "%02d:%02d", &hournum, &minnum))) {
396        /* time stamp without seconds */
397        date += 5;
398        secnum = 0;
399      }
400      else {
401        long lval;
402        int error;
403        int old_errno;
404
405        old_errno = ERRNO;
406        SET_ERRNO(0);
407        lval = strtol(date, &end, 10);
408        error = ERRNO;
409        if(error != old_errno)
410          SET_ERRNO(old_errno);
411
412        if(error)
413          return PARSEDATE_FAIL;
414
415#if LONG_MAX != INT_MAX
416        if((lval > (long)INT_MAX) || (lval < (long)INT_MIN))
417          return PARSEDATE_FAIL;
418#endif
419
420        val = curlx_sltosi(lval);
421
422        if((tzoff == -1) &&
423           ((end - date) == 4) &&
424           (val <= 1400) &&
425           (indate< date) &&
426           ((date[-1] == '+' || date[-1] == '-'))) {
427          /* four digits and a value less than or equal to 1400 (to take into
428             account all sorts of funny time zone diffs) and it is preceded
429             with a plus or minus. This is a time zone indication.  1400 is
430             picked since +1300 is frequently used and +1400 is mentioned as
431             an edge number in the document "ISO C 200X Proposal: Timezone
432             Functions" at http://david.tribble.com/text/c0xtimezone.html If
433             anyone has a more authoritative source for the exact maximum time
434             zone offsets, please speak up! */
435          found = TRUE;
436          tzoff = (val/100 * 60 + val%100)*60;
437
438          /* the + and - prefix indicates the local time compared to GMT,
439             this we need ther reversed math to get what we want */
440          tzoff = date[-1]=='+'?-tzoff:tzoff;
441        }
442
443        if(((end - date) == 8) &&
444           (yearnum == -1) &&
445           (monnum == -1) &&
446           (mdaynum == -1)) {
447          /* 8 digits, no year, month or day yet. This is YYYYMMDD */
448          found = TRUE;
449          yearnum = val/10000;
450          monnum = (val%10000)/100-1; /* month is 0 - 11 */
451          mdaynum = val%100;
452        }
453
454        if(!found && (dignext == DATE_MDAY) && (mdaynum == -1)) {
455          if((val > 0) && (val<32)) {
456            mdaynum = val;
457            found = TRUE;
458          }
459          dignext = DATE_YEAR;
460        }
461
462        if(!found && (dignext == DATE_YEAR) && (yearnum == -1)) {
463          yearnum = val;
464          found = TRUE;
465          if(yearnum < 1900) {
466            if(yearnum > 70)
467              yearnum += 1900;
468            else
469              yearnum += 2000;
470          }
471          if(mdaynum == -1)
472            dignext = DATE_MDAY;
473        }
474
475        if(!found)
476          return PARSEDATE_FAIL;
477
478        date = end;
479      }
480    }
481
482    part++;
483  }
484
485  if(-1 == secnum)
486    secnum = minnum = hournum = 0; /* no time, make it zero */
487
488  if((-1 == mdaynum) ||
489     (-1 == monnum) ||
490     (-1 == yearnum))
491    /* lacks vital info, fail */
492    return PARSEDATE_FAIL;
493
494  /* 32 bit time_t can only hold dates to the beginning of 2038 */
495  if(sizeof(time_t) < 5 && yearnum > 2037) {
496    *output = 0x7fffffff;
497    return PARSEDATE_LATER;
498  }
499
500  if(yearnum < 1970) {
501    *output = 0;
502    return PARSEDATE_SOONER;
503  }
504
505  if((mdaynum > 31) || (monnum > 11) ||
506     (hournum > 23) || (minnum > 59) || (secnum > 60))
507    return PARSEDATE_FAIL; /* clearly an illegal date */
508
509  tm.tm_sec = secnum;
510  tm.tm_min = minnum;
511  tm.tm_hour = hournum;
512  tm.tm_mday = mdaynum;
513  tm.tm_mon = monnum;
514  tm.tm_year = yearnum - 1900;
515
516  /* my_timegm() returns a time_t. time_t is often 32 bits, even on many
517     architectures that feature 64 bit 'long'.
518
519     Some systems have 64 bit time_t and deal with years beyond 2038. However,
520     even on some of the systems with 64 bit time_t mktime() returns -1 for
521     dates beyond 03:14:07 UTC, January 19, 2038. (Such as AIX 5100-06)
522  */
523  t = my_timegm(&tm);
524
525  /* time zone adjust (cast t to int to compare to negative one) */
526  if(-1 != (int)t) {
527
528    /* Add the time zone diff between local time zone and GMT. */
529    long delta = (long)(tzoff!=-1?tzoff:0);
530
531    if((delta>0) && (t > LONG_MAX - delta)) {
532      *output = 0x7fffffff;
533      return PARSEDATE_LATER; /* time_t overflow */
534    }
535
536    t += delta;
537  }
538
539  *output = t;
540
541  return PARSEDATE_OK;
542}
543
544time_t curl_getdate(const char *p, const time_t *now)
545{
546  time_t parsed = -1;
547  int rc = parsedate(p, &parsed);
548  (void)now; /* legacy argument from the past that we ignore */
549
550  switch(rc) {
551  case PARSEDATE_OK:
552  case PARSEDATE_LATER:
553  case PARSEDATE_SOONER:
554    return parsed;
555  }
556  /* everything else is fail */
557  return -1;
558}
559
560/*
561 * Curl_gmtime() is a gmtime() replacement for portability. Do not use the
562 * gmtime_r() or gmtime() functions anywhere else but here.
563 *
564 */
565
566CURLcode Curl_gmtime(time_t intime, struct tm *store)
567{
568  const struct tm *tm;
569#ifdef HAVE_GMTIME_R
570  /* thread-safe version */
571  tm = (struct tm *)gmtime_r(&intime, store);
572#else
573  tm = gmtime(&intime);
574  if(tm)
575    *store = *tm; /* copy the pointed struct to the local copy */
576#endif
577
578  if(!tm)
579    return CURLE_BAD_FUNCTION_ARGUMENT;
580  return CURLE_OK;
581}
582