xref: /third_party/curl/lib/parsedate.c (revision 13498266)
1/***************************************************************************
2 *                                  _   _ ____  _
3 *  Project                     ___| | | |  _ \| |
4 *                             / __| | | | |_) | |
5 *                            | (__| |_| |  _ <| |___
6 *                             \___|\___/|_| \_\_____|
7 *
8 * Copyright (C) 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.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 * SPDX-License-Identifier: curl
22 *
23 ***************************************************************************/
24/*
25  A brief summary of the date string formats this parser groks:
26
27  RFC 2616 3.3.1
28
29  Sun, 06 Nov 1994 08:49:37 GMT  ; RFC 822, updated by RFC 1123
30  Sunday, 06-Nov-94 08:49:37 GMT ; RFC 850, obsoleted by RFC 1036
31  Sun Nov  6 08:49:37 1994       ; ANSI C's asctime() format
32
33  we support dates without week day name:
34
35  06 Nov 1994 08:49:37 GMT
36  06-Nov-94 08:49:37 GMT
37  Nov  6 08:49:37 1994
38
39  without the time zone:
40
41  06 Nov 1994 08:49:37
42  06-Nov-94 08:49:37
43
44  weird order:
45
46  1994 Nov 6 08:49:37  (GNU date fails)
47  GMT 08:49:37 06-Nov-94 Sunday
48  94 6 Nov 08:49:37    (GNU date fails)
49
50  time left out:
51
52  1994 Nov 6
53  06-Nov-94
54  Sun Nov 6 94
55
56  unusual separators:
57
58  1994.Nov.6
59  Sun/Nov/6/94/GMT
60
61  commonly used time zone names:
62
63  Sun, 06 Nov 1994 08:49:37 CET
64  06 Nov 1994 08:49:37 EST
65
66  time zones specified using RFC822 style:
67
68  Sun, 12 Sep 2004 15:05:58 -0700
69  Sat, 11 Sep 2004 21:32:11 +0200
70
71  compact numerical date strings:
72
73  20040912 15:05:58 -0700
74  20040911 +0200
75
76*/
77
78#include "curl_setup.h"
79
80#include <limits.h>
81
82#include <curl/curl.h>
83#include "strcase.h"
84#include "warnless.h"
85#include "parsedate.h"
86
87/*
88 * parsedate()
89 *
90 * Returns:
91 *
92 * PARSEDATE_OK     - a fine conversion
93 * PARSEDATE_FAIL   - failed to convert
94 * PARSEDATE_LATER  - time overflow at the far end of time_t
95 * PARSEDATE_SOONER - time underflow at the low end of time_t
96 */
97
98static int parsedate(const char *date, time_t *output);
99
100#define PARSEDATE_OK     0
101#define PARSEDATE_FAIL   -1
102#define PARSEDATE_LATER  1
103#define PARSEDATE_SOONER 2
104
105#if !defined(CURL_DISABLE_PARSEDATE) || !defined(CURL_DISABLE_FTP) || \
106  !defined(CURL_DISABLE_FILE)
107/* These names are also used by FTP and FILE code */
108const char * const Curl_wkday[] =
109{"Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"};
110const char * const Curl_month[]=
111{ "Jan", "Feb", "Mar", "Apr", "May", "Jun",
112  "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" };
113#endif
114
115#ifndef CURL_DISABLE_PARSEDATE
116static const char * const weekday[] =
117{ "Monday", "Tuesday", "Wednesday", "Thursday",
118  "Friday", "Saturday", "Sunday" };
119
120struct tzinfo {
121  char name[5];
122  int offset; /* +/- in minutes */
123};
124
125/* Here's a bunch of frequently used time zone names. These were supported
126   by the old getdate parser. */
127#define tDAYZONE -60       /* offset for daylight savings time */
128static const struct tzinfo tz[]= {
129  {"GMT", 0},              /* Greenwich Mean */
130  {"UT",  0},              /* Universal Time */
131  {"UTC", 0},              /* Universal (Coordinated) */
132  {"WET", 0},              /* Western European */
133  {"BST", 0 tDAYZONE},     /* British Summer */
134  {"WAT", 60},             /* West Africa */
135  {"AST", 240},            /* Atlantic Standard */
136  {"ADT", 240 tDAYZONE},   /* Atlantic Daylight */
137  {"EST", 300},            /* Eastern Standard */
138  {"EDT", 300 tDAYZONE},   /* Eastern Daylight */
139  {"CST", 360},            /* Central Standard */
140  {"CDT", 360 tDAYZONE},   /* Central Daylight */
141  {"MST", 420},            /* Mountain Standard */
142  {"MDT", 420 tDAYZONE},   /* Mountain Daylight */
143  {"PST", 480},            /* Pacific Standard */
144  {"PDT", 480 tDAYZONE},   /* Pacific Daylight */
145  {"YST", 540},            /* Yukon Standard */
146  {"YDT", 540 tDAYZONE},   /* Yukon Daylight */
147  {"HST", 600},            /* Hawaii Standard */
148  {"HDT", 600 tDAYZONE},   /* Hawaii Daylight */
149  {"CAT", 600},            /* Central Alaska */
150  {"AHST", 600},           /* Alaska-Hawaii Standard */
151  {"NT",  660},            /* Nome */
152  {"IDLW", 720},           /* International Date Line West */
153  {"CET", -60},            /* Central European */
154  {"MET", -60},            /* Middle European */
155  {"MEWT", -60},           /* Middle European Winter */
156  {"MEST", -60 tDAYZONE},  /* Middle European Summer */
157  {"CEST", -60 tDAYZONE},  /* Central European Summer */
158  {"MESZ", -60 tDAYZONE},  /* Middle European Summer */
159  {"FWT", -60},            /* French Winter */
160  {"FST", -60 tDAYZONE},   /* French Summer */
161  {"EET", -120},           /* Eastern Europe, USSR Zone 1 */
162  {"WAST", -420},          /* West Australian Standard */
163  {"WADT", -420 tDAYZONE}, /* West Australian Daylight */
164  {"CCT", -480},           /* China Coast, USSR Zone 7 */
165  {"JST", -540},           /* Japan Standard, USSR Zone 8 */
166  {"EAST", -600},          /* Eastern Australian Standard */
167  {"EADT", -600 tDAYZONE}, /* Eastern Australian Daylight */
168  {"GST", -600},           /* Guam Standard, USSR Zone 9 */
169  {"NZT", -720},           /* New Zealand */
170  {"NZST", -720},          /* New Zealand Standard */
171  {"NZDT", -720 tDAYZONE}, /* New Zealand Daylight */
172  {"IDLE", -720},          /* International Date Line East */
173  /* Next up: Military timezone names. RFC822 allowed these, but (as noted in
174     RFC 1123) had their signs wrong. Here we use the correct signs to match
175     actual military usage.
176   */
177  {"A",  1 * 60},         /* Alpha */
178  {"B",  2 * 60},         /* Bravo */
179  {"C",  3 * 60},         /* Charlie */
180  {"D",  4 * 60},         /* Delta */
181  {"E",  5 * 60},         /* Echo */
182  {"F",  6 * 60},         /* Foxtrot */
183  {"G",  7 * 60},         /* Golf */
184  {"H",  8 * 60},         /* Hotel */
185  {"I",  9 * 60},         /* India */
186  /* "J", Juliet is not used as a timezone, to indicate the observer's local
187     time */
188  {"K", 10 * 60},         /* Kilo */
189  {"L", 11 * 60},         /* Lima */
190  {"M", 12 * 60},         /* Mike */
191  {"N",  -1 * 60},         /* November */
192  {"O",  -2 * 60},         /* Oscar */
193  {"P",  -3 * 60},         /* Papa */
194  {"Q",  -4 * 60},         /* Quebec */
195  {"R",  -5 * 60},         /* Romeo */
196  {"S",  -6 * 60},         /* Sierra */
197  {"T",  -7 * 60},         /* Tango */
198  {"U",  -8 * 60},         /* Uniform */
199  {"V",  -9 * 60},         /* Victor */
200  {"W", -10 * 60},         /* Whiskey */
201  {"X", -11 * 60},         /* X-ray */
202  {"Y", -12 * 60},         /* Yankee */
203  {"Z", 0},                /* Zulu, zero meridian, a.k.a. UTC */
204};
205
206/* returns:
207   -1 no day
208   0 monday - 6 sunday
209*/
210
211static int checkday(const char *check, size_t len)
212{
213  int i;
214  const char * const *what;
215  if(len > 3)
216    what = &weekday[0];
217  else if(len == 3)
218    what = &Curl_wkday[0];
219  else
220    return -1; /* too short */
221  for(i = 0; i<7; i++) {
222    size_t ilen = strlen(what[0]);
223    if((ilen == len) &&
224       strncasecompare(check, what[0], len))
225      return i;
226    what++;
227  }
228  return -1;
229}
230
231static int checkmonth(const char *check, size_t len)
232{
233  int i;
234  const char * const *what = &Curl_month[0];
235  if(len != 3)
236    return -1; /* not a month */
237
238  for(i = 0; i<12; i++) {
239    if(strncasecompare(check, what[0], 3))
240      return i;
241    what++;
242  }
243  return -1; /* return the offset or -1, no real offset is -1 */
244}
245
246/* return the time zone offset between GMT and the input one, in number
247   of seconds or -1 if the timezone wasn't found/legal */
248
249static int checktz(const char *check, size_t len)
250{
251  unsigned int i;
252  const struct tzinfo *what = tz;
253  if(len > 4) /* longer than any valid timezone */
254    return -1;
255
256  for(i = 0; i< sizeof(tz)/sizeof(tz[0]); i++) {
257    size_t ilen = strlen(what->name);
258    if((ilen == len) &&
259       strncasecompare(check, what->name, len))
260      return what->offset*60;
261    what++;
262  }
263  return -1;
264}
265
266static void skip(const char **date)
267{
268  /* skip everything that aren't letters or digits */
269  while(**date && !ISALNUM(**date))
270    (*date)++;
271}
272
273enum assume {
274  DATE_MDAY,
275  DATE_YEAR,
276  DATE_TIME
277};
278
279/*
280 * time2epoch: time stamp to seconds since epoch in GMT time zone.  Similar to
281 * mktime but for GMT only.
282 */
283static time_t time2epoch(int sec, int min, int hour,
284                         int mday, int mon, int year)
285{
286  static const int month_days_cumulative [12] =
287    { 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334 };
288  int leap_days = year - (mon <= 1);
289  leap_days = ((leap_days / 4) - (leap_days / 100) + (leap_days / 400)
290               - (1969 / 4) + (1969 / 100) - (1969 / 400));
291  return ((((time_t) (year - 1970) * 365
292            + leap_days + month_days_cumulative[mon] + mday - 1) * 24
293           + hour) * 60 + min) * 60 + sec;
294}
295
296/* Returns the value of a single-digit or two-digit decimal number, return
297   then pointer to after the number. The 'date' pointer is known to point to a
298   digit. */
299static int oneortwodigit(const char *date, const char **endp)
300{
301  int num = date[0] - '0';
302  if(ISDIGIT(date[1])) {
303    *endp = &date[2];
304    return num*10 + (date[1] - '0');
305  }
306  *endp = &date[1];
307  return num;
308}
309
310
311/* HH:MM:SS or HH:MM and accept single-digits too */
312static bool match_time(const char *date,
313                       int *h, int *m, int *s, char **endp)
314{
315  const char *p;
316  int hh, mm, ss = 0;
317  hh = oneortwodigit(date, &p);
318  if((hh < 24) && (*p == ':') && ISDIGIT(p[1])) {
319    mm = oneortwodigit(&p[1], &p);
320    if(mm < 60) {
321      if((*p == ':') && ISDIGIT(p[1])) {
322        ss = oneortwodigit(&p[1], &p);
323        if(ss <= 60) {
324          /* valid HH:MM:SS */
325          goto match;
326        }
327      }
328      else {
329        /* valid HH:MM */
330        goto match;
331      }
332    }
333  }
334  return FALSE; /* not a time string */
335match:
336  *h = hh;
337  *m = mm;
338  *s = ss;
339  *endp = (char *)p;
340  return TRUE;
341}
342
343/*
344 * parsedate()
345 *
346 * Returns:
347 *
348 * PARSEDATE_OK     - a fine conversion
349 * PARSEDATE_FAIL   - failed to convert
350 * PARSEDATE_LATER  - time overflow at the far end of time_t
351 * PARSEDATE_SOONER - time underflow at the low end of time_t
352 */
353
354/* Wednesday is the longest name this parser knows about */
355#define NAME_LEN 12
356
357static int parsedate(const char *date, time_t *output)
358{
359  time_t t = 0;
360  int wdaynum = -1;  /* day of the week number, 0-6 (mon-sun) */
361  int monnum = -1;   /* month of the year number, 0-11 */
362  int mdaynum = -1; /* day of month, 1 - 31 */
363  int hournum = -1;
364  int minnum = -1;
365  int secnum = -1;
366  int yearnum = -1;
367  int tzoff = -1;
368  enum assume dignext = DATE_MDAY;
369  const char *indate = date; /* save the original pointer */
370  int part = 0; /* max 6 parts */
371
372  while(*date && (part < 6)) {
373    bool found = FALSE;
374
375    skip(&date);
376
377    if(ISALPHA(*date)) {
378      /* a name coming up */
379      size_t len = 0;
380      const char *p = date;
381      while(ISALPHA(*p) && (len < NAME_LEN)) {
382        p++;
383        len++;
384      }
385
386      if(len != NAME_LEN) {
387        if(wdaynum == -1) {
388          wdaynum = checkday(date, len);
389          if(wdaynum != -1)
390            found = TRUE;
391        }
392        if(!found && (monnum == -1)) {
393          monnum = checkmonth(date, len);
394          if(monnum != -1)
395            found = TRUE;
396        }
397
398        if(!found && (tzoff == -1)) {
399          /* this just must be a time zone string */
400          tzoff = checktz(date, len);
401          if(tzoff != -1)
402            found = TRUE;
403        }
404      }
405      if(!found)
406        return PARSEDATE_FAIL; /* bad string */
407
408      date += len;
409    }
410    else if(ISDIGIT(*date)) {
411      /* a digit */
412      int val;
413      char *end;
414      if((secnum == -1) &&
415         match_time(date, &hournum, &minnum, &secnum, &end)) {
416        /* time stamp */
417        date = end;
418      }
419      else {
420        long lval;
421        int error;
422        int old_errno;
423
424        old_errno = errno;
425        errno = 0;
426        lval = strtol(date, &end, 10);
427        error = errno;
428        if(errno != old_errno)
429          errno = old_errno;
430
431        if(error)
432          return PARSEDATE_FAIL;
433
434#if LONG_MAX != INT_MAX
435        if((lval > (long)INT_MAX) || (lval < (long)INT_MIN))
436          return PARSEDATE_FAIL;
437#endif
438
439        val = curlx_sltosi(lval);
440
441        if((tzoff == -1) &&
442           ((end - date) == 4) &&
443           (val <= 1400) &&
444           (indate< date) &&
445           ((date[-1] == '+' || date[-1] == '-'))) {
446          /* four digits and a value less than or equal to 1400 (to take into
447             account all sorts of funny time zone diffs) and it is preceded
448             with a plus or minus. This is a time zone indication.  1400 is
449             picked since +1300 is frequently used and +1400 is mentioned as
450             an edge number in the document "ISO C 200X Proposal: Timezone
451             Functions" at http://david.tribble.com/text/c0xtimezone.html If
452             anyone has a more authoritative source for the exact maximum time
453             zone offsets, please speak up! */
454          found = TRUE;
455          tzoff = (val/100 * 60 + val%100)*60;
456
457          /* the + and - prefix indicates the local time compared to GMT,
458             this we need their reversed math to get what we want */
459          tzoff = date[-1]=='+'?-tzoff:tzoff;
460        }
461
462        if(((end - date) == 8) &&
463           (yearnum == -1) &&
464           (monnum == -1) &&
465           (mdaynum == -1)) {
466          /* 8 digits, no year, month or day yet. This is YYYYMMDD */
467          found = TRUE;
468          yearnum = val/10000;
469          monnum = (val%10000)/100-1; /* month is 0 - 11 */
470          mdaynum = val%100;
471        }
472
473        if(!found && (dignext == DATE_MDAY) && (mdaynum == -1)) {
474          if((val > 0) && (val<32)) {
475            mdaynum = val;
476            found = TRUE;
477          }
478          dignext = DATE_YEAR;
479        }
480
481        if(!found && (dignext == DATE_YEAR) && (yearnum == -1)) {
482          yearnum = val;
483          found = TRUE;
484          if(yearnum < 100) {
485            if(yearnum > 70)
486              yearnum += 1900;
487            else
488              yearnum += 2000;
489          }
490          if(mdaynum == -1)
491            dignext = DATE_MDAY;
492        }
493
494        if(!found)
495          return PARSEDATE_FAIL;
496
497        date = end;
498      }
499    }
500
501    part++;
502  }
503
504  if(-1 == secnum)
505    secnum = minnum = hournum = 0; /* no time, make it zero */
506
507  if((-1 == mdaynum) ||
508     (-1 == monnum) ||
509     (-1 == yearnum))
510    /* lacks vital info, fail */
511    return PARSEDATE_FAIL;
512
513#ifdef HAVE_TIME_T_UNSIGNED
514  if(yearnum < 1970) {
515    /* only positive numbers cannot return earlier */
516    *output = TIME_T_MIN;
517    return PARSEDATE_SOONER;
518  }
519#endif
520
521#if (SIZEOF_TIME_T < 5)
522
523#ifdef HAVE_TIME_T_UNSIGNED
524  /* an unsigned 32 bit time_t can only hold dates to 2106 */
525  if(yearnum > 2105) {
526    *output = TIME_T_MAX;
527    return PARSEDATE_LATER;
528  }
529#else
530  /* a signed 32 bit time_t can only hold dates to the beginning of 2038 */
531  if(yearnum > 2037) {
532    *output = TIME_T_MAX;
533    return PARSEDATE_LATER;
534  }
535  if(yearnum < 1903) {
536    *output = TIME_T_MIN;
537    return PARSEDATE_SOONER;
538  }
539#endif
540
541#else
542  /* The Gregorian calendar was introduced 1582 */
543  if(yearnum < 1583)
544    return PARSEDATE_FAIL;
545#endif
546
547  if((mdaynum > 31) || (monnum > 11) ||
548     (hournum > 23) || (minnum > 59) || (secnum > 60))
549    return PARSEDATE_FAIL; /* clearly an illegal date */
550
551  /* time2epoch() returns a time_t. time_t is often 32 bits, sometimes even on
552     architectures that feature 64 bit 'long' but ultimately time_t is the
553     correct data type to use.
554  */
555  t = time2epoch(secnum, minnum, hournum, mdaynum, monnum, yearnum);
556
557  /* Add the time zone diff between local time zone and GMT. */
558  if(tzoff == -1)
559    tzoff = 0;
560
561  if((tzoff > 0) && (t > TIME_T_MAX - tzoff)) {
562    *output = TIME_T_MAX;
563    return PARSEDATE_LATER; /* time_t overflow */
564  }
565
566  t += tzoff;
567
568  *output = t;
569
570  return PARSEDATE_OK;
571}
572#else
573/* disabled */
574static int parsedate(const char *date, time_t *output)
575{
576  (void)date;
577  *output = 0;
578  return PARSEDATE_OK; /* a lie */
579}
580#endif
581
582time_t curl_getdate(const char *p, const time_t *now)
583{
584  time_t parsed = -1;
585  int rc = parsedate(p, &parsed);
586  (void)now; /* legacy argument from the past that we ignore */
587
588  if(rc == PARSEDATE_OK) {
589    if(parsed == -1)
590      /* avoid returning -1 for a working scenario */
591      parsed++;
592    return parsed;
593  }
594  /* everything else is fail */
595  return -1;
596}
597
598/* Curl_getdate_capped() differs from curl_getdate() in that this will return
599   TIME_T_MAX in case the parsed time value was too big, instead of an
600   error. */
601
602time_t Curl_getdate_capped(const char *p)
603{
604  time_t parsed = -1;
605  int rc = parsedate(p, &parsed);
606
607  switch(rc) {
608  case PARSEDATE_OK:
609    if(parsed == -1)
610      /* avoid returning -1 for a working scenario */
611      parsed++;
612    return parsed;
613  case PARSEDATE_LATER:
614    /* this returns the maximum time value */
615    return parsed;
616  default:
617    return -1; /* everything else is fail */
618  }
619  /* UNREACHABLE */
620}
621
622/*
623 * Curl_gmtime() is a gmtime() replacement for portability. Do not use the
624 * gmtime_r() or gmtime() functions anywhere else but here.
625 *
626 */
627
628CURLcode Curl_gmtime(time_t intime, struct tm *store)
629{
630  const struct tm *tm;
631#ifdef HAVE_GMTIME_R
632  /* thread-safe version */
633  tm = (struct tm *)gmtime_r(&intime, store);
634#else
635  /* !checksrc! disable BANNEDFUNC 1 */
636  tm = gmtime(&intime);
637  if(tm)
638    *store = *tm; /* copy the pointed struct to the local copy */
639#endif
640
641  if(!tm)
642    return CURLE_BAD_FUNCTION_ARGUMENT;
643  return CURLE_OK;
644}
645