xref: /freebsd/contrib/tzcode/strftime.c (revision 9768746b)
1 /* Convert a broken-down timestamp to a string.  */
2 
3 /* Copyright 1989 The Regents of the University of California.
4    All rights reserved.
5 
6    Redistribution and use in source and binary forms, with or without
7    modification, are permitted provided that the following conditions
8    are met:
9    1. Redistributions of source code must retain the above copyright
10       notice, this list of conditions and the following disclaimer.
11    2. Redistributions in binary form must reproduce the above copyright
12       notice, this list of conditions and the following disclaimer in the
13       documentation and/or other materials provided with the distribution.
14    3. Neither the name of the University nor the names of its contributors
15       may be used to endorse or promote products derived from this software
16       without specific prior written permission.
17 
18    THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS" AND
19    ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20    IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21    ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22    FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23    DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24    OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25    HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26    LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27    OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28    SUCH DAMAGE.  */
29 
30 /*
31 ** Based on the UCB version with the copyright notice appearing above.
32 **
33 ** This is ANSIish only when "multibyte character == plain character".
34 */
35 
36 #include "private.h"
37 
38 #include <fcntl.h>
39 #include <locale.h>
40 #include <stdio.h>
41 
42 #ifndef DEPRECATE_TWO_DIGIT_YEARS
43 # define DEPRECATE_TWO_DIGIT_YEARS false
44 #endif
45 
46 struct lc_time_T {
47 	const char *	mon[MONSPERYEAR];
48 	const char *	month[MONSPERYEAR];
49 	const char *	wday[DAYSPERWEEK];
50 	const char *	weekday[DAYSPERWEEK];
51 	const char *	X_fmt;
52 	const char *	x_fmt;
53 	const char *	c_fmt;
54 	const char *	am;
55 	const char *	pm;
56 	const char *	date_fmt;
57 };
58 
59 static const struct lc_time_T	C_time_locale = {
60 	{
61 		"Jan", "Feb", "Mar", "Apr", "May", "Jun",
62 		"Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
63 	}, {
64 		"January", "February", "March", "April", "May", "June",
65 		"July", "August", "September", "October", "November", "December"
66 	}, {
67 		"Sun", "Mon", "Tue", "Wed",
68 		"Thu", "Fri", "Sat"
69 	}, {
70 		"Sunday", "Monday", "Tuesday", "Wednesday",
71 		"Thursday", "Friday", "Saturday"
72 	},
73 
74 	/* X_fmt */
75 	"%H:%M:%S",
76 
77 	/*
78 	** x_fmt
79 	** C99 and later require this format.
80 	** Using just numbers (as here) makes Quakers happier;
81 	** it's also compatible with SVR4.
82 	*/
83 	"%m/%d/%y",
84 
85 	/*
86 	** c_fmt
87 	** C99 and later require this format.
88 	** Previously this code used "%D %X", but we now conform to C99.
89 	** Note that
90 	**	"%a %b %d %H:%M:%S %Y"
91 	** is used by Solaris 2.3.
92 	*/
93 	"%a %b %e %T %Y",
94 
95 	/* am */
96 	"AM",
97 
98 	/* pm */
99 	"PM",
100 
101 	/* date_fmt */
102 	"%a %b %e %H:%M:%S %Z %Y"
103 };
104 
105 enum warn { IN_NONE, IN_SOME, IN_THIS, IN_ALL };
106 
107 static char *	_add(const char *, char *, const char *);
108 static char *	_conv(int, const char *, char *, const char *);
109 static char *	_fmt(const char *, const struct tm *, char *, const char *,
110 		     enum warn *);
111 static char *	_yconv(int, int, bool, bool, char *, char const *);
112 
113 #ifndef YEAR_2000_NAME
114 # define YEAR_2000_NAME "CHECK_STRFTIME_FORMATS_FOR_TWO_DIGIT_YEARS"
115 #endif /* !defined YEAR_2000_NAME */
116 
117 #if HAVE_STRFTIME_L
118 size_t
119 strftime_l(char *s, size_t maxsize, char const *format, struct tm const *t,
120 	   ATTRIBUTE_MAYBE_UNUSED locale_t locale)
121 {
122   /* Just call strftime, as only the C locale is supported.  */
123   return strftime(s, maxsize, format, t);
124 }
125 #endif
126 
127 size_t
128 strftime(char *s, size_t maxsize, const char *format, const struct tm *t)
129 {
130 	char *	p;
131 	int saved_errno = errno;
132 	enum warn warn = IN_NONE;
133 
134 	tzset();
135 	p = _fmt(format, t, s, s + maxsize, &warn);
136 	if (!p) {
137 	  errno = EOVERFLOW;
138 	  return 0;
139 	}
140 	if (DEPRECATE_TWO_DIGIT_YEARS
141 	    && warn != IN_NONE && getenv(YEAR_2000_NAME)) {
142 		fprintf(stderr, "\n");
143 		fprintf(stderr, "strftime format \"%s\" ", format);
144 		fprintf(stderr, "yields only two digits of years in ");
145 		if (warn == IN_SOME)
146 			fprintf(stderr, "some locales");
147 		else if (warn == IN_THIS)
148 			fprintf(stderr, "the current locale");
149 		else	fprintf(stderr, "all locales");
150 		fprintf(stderr, "\n");
151 	}
152 	if (p == s + maxsize) {
153 		errno = ERANGE;
154 		return 0;
155 	}
156 	*p = '\0';
157 	errno = saved_errno;
158 	return p - s;
159 }
160 
161 static char *
162 _fmt(const char *format, const struct tm *t, char *pt,
163      const char *ptlim, enum warn *warnp)
164 {
165 	struct lc_time_T const *Locale = &C_time_locale;
166 
167 	for ( ; *format; ++format) {
168 		if (*format == '%') {
169 label:
170 			switch (*++format) {
171 			case '\0':
172 				--format;
173 				break;
174 			case 'A':
175 				pt = _add((t->tm_wday < 0 ||
176 					t->tm_wday >= DAYSPERWEEK) ?
177 					"?" : Locale->weekday[t->tm_wday],
178 					pt, ptlim);
179 				continue;
180 			case 'a':
181 				pt = _add((t->tm_wday < 0 ||
182 					t->tm_wday >= DAYSPERWEEK) ?
183 					"?" : Locale->wday[t->tm_wday],
184 					pt, ptlim);
185 				continue;
186 			case 'B':
187 				pt = _add((t->tm_mon < 0 ||
188 					t->tm_mon >= MONSPERYEAR) ?
189 					"?" : Locale->month[t->tm_mon],
190 					pt, ptlim);
191 				continue;
192 			case 'b':
193 			case 'h':
194 				pt = _add((t->tm_mon < 0 ||
195 					t->tm_mon >= MONSPERYEAR) ?
196 					"?" : Locale->mon[t->tm_mon],
197 					pt, ptlim);
198 				continue;
199 			case 'C':
200 				/*
201 				** %C used to do a...
202 				**	_fmt("%a %b %e %X %Y", t);
203 				** ...whereas now POSIX 1003.2 calls for
204 				** something completely different.
205 				** (ado, 1993-05-24)
206 				*/
207 				pt = _yconv(t->tm_year, TM_YEAR_BASE,
208 					    true, false, pt, ptlim);
209 				continue;
210 			case 'c':
211 				{
212 				enum warn warn2 = IN_SOME;
213 
214 				pt = _fmt(Locale->c_fmt, t, pt, ptlim, &warn2);
215 				if (warn2 == IN_ALL)
216 					warn2 = IN_THIS;
217 				if (warn2 > *warnp)
218 					*warnp = warn2;
219 				}
220 				continue;
221 			case 'D':
222 				pt = _fmt("%m/%d/%y", t, pt, ptlim, warnp);
223 				continue;
224 			case 'd':
225 				pt = _conv(t->tm_mday, "%02d", pt, ptlim);
226 				continue;
227 			case 'E':
228 			case 'O':
229 				/*
230 				** Locale modifiers of C99 and later.
231 				** The sequences
232 				**	%Ec %EC %Ex %EX %Ey %EY
233 				**	%Od %oe %OH %OI %Om %OM
234 				**	%OS %Ou %OU %OV %Ow %OW %Oy
235 				** are supposed to provide alternative
236 				** representations.
237 				*/
238 				goto label;
239 			case 'e':
240 				pt = _conv(t->tm_mday, "%2d", pt, ptlim);
241 				continue;
242 			case 'F':
243 				pt = _fmt("%Y-%m-%d", t, pt, ptlim, warnp);
244 				continue;
245 			case 'H':
246 				pt = _conv(t->tm_hour, "%02d", pt, ptlim);
247 				continue;
248 			case 'I':
249 				pt = _conv((t->tm_hour % 12) ?
250 					(t->tm_hour % 12) : 12,
251 					"%02d", pt, ptlim);
252 				continue;
253 			case 'j':
254 				pt = _conv(t->tm_yday + 1, "%03d", pt, ptlim);
255 				continue;
256 			case 'k':
257 				/*
258 				** This used to be...
259 				**	_conv(t->tm_hour % 12 ?
260 				**		t->tm_hour % 12 : 12, 2, ' ');
261 				** ...and has been changed to the below to
262 				** match SunOS 4.1.1 and Arnold Robbins'
263 				** strftime version 3.0. That is, "%k" and
264 				** "%l" have been swapped.
265 				** (ado, 1993-05-24)
266 				*/
267 				pt = _conv(t->tm_hour, "%2d", pt, ptlim);
268 				continue;
269 #ifdef KITCHEN_SINK
270 			case 'K':
271 				/*
272 				** After all this time, still unclaimed!
273 				*/
274 				pt = _add("kitchen sink", pt, ptlim);
275 				continue;
276 #endif /* defined KITCHEN_SINK */
277 			case 'l':
278 				/*
279 				** This used to be...
280 				**	_conv(t->tm_hour, 2, ' ');
281 				** ...and has been changed to the below to
282 				** match SunOS 4.1.1 and Arnold Robbin's
283 				** strftime version 3.0. That is, "%k" and
284 				** "%l" have been swapped.
285 				** (ado, 1993-05-24)
286 				*/
287 				pt = _conv((t->tm_hour % 12) ?
288 					(t->tm_hour % 12) : 12,
289 					"%2d", pt, ptlim);
290 				continue;
291 			case 'M':
292 				pt = _conv(t->tm_min, "%02d", pt, ptlim);
293 				continue;
294 			case 'm':
295 				pt = _conv(t->tm_mon + 1, "%02d", pt, ptlim);
296 				continue;
297 			case 'n':
298 				pt = _add("\n", pt, ptlim);
299 				continue;
300 			case 'p':
301 				pt = _add((t->tm_hour >= (HOURSPERDAY / 2)) ?
302 					Locale->pm :
303 					Locale->am,
304 					pt, ptlim);
305 				continue;
306 			case 'R':
307 				pt = _fmt("%H:%M", t, pt, ptlim, warnp);
308 				continue;
309 			case 'r':
310 				pt = _fmt("%I:%M:%S %p", t, pt, ptlim, warnp);
311 				continue;
312 			case 'S':
313 				pt = _conv(t->tm_sec, "%02d", pt, ptlim);
314 				continue;
315 			case 's':
316 				{
317 					struct tm	tm;
318 					char		buf[INT_STRLEN_MAXIMUM(
319 								time_t) + 1];
320 					time_t		mkt;
321 
322 					tm.tm_sec = t->tm_sec;
323 					tm.tm_min = t->tm_min;
324 					tm.tm_hour = t->tm_hour;
325 					tm.tm_mday = t->tm_mday;
326 					tm.tm_mon = t->tm_mon;
327 					tm.tm_year = t->tm_year;
328 					tm.tm_isdst = t->tm_isdst;
329 #if defined TM_GMTOFF && ! UNINIT_TRAP
330 					tm.TM_GMTOFF = t->TM_GMTOFF;
331 #endif
332 					mkt = mktime(&tm);
333 					/* If mktime fails, %s expands to the
334 					   value of (time_t) -1 as a failure
335 					   marker; this is better in practice
336 					   than strftime failing.  */
337 					if (TYPE_SIGNED(time_t)) {
338 					  intmax_t n = mkt;
339 					  sprintf(buf, "%"PRIdMAX, n);
340 					} else {
341 					  uintmax_t n = mkt;
342 					  sprintf(buf, "%"PRIuMAX, n);
343 					}
344 					pt = _add(buf, pt, ptlim);
345 				}
346 				continue;
347 			case 'T':
348 				pt = _fmt("%H:%M:%S", t, pt, ptlim, warnp);
349 				continue;
350 			case 't':
351 				pt = _add("\t", pt, ptlim);
352 				continue;
353 			case 'U':
354 				pt = _conv((t->tm_yday + DAYSPERWEEK -
355 					t->tm_wday) / DAYSPERWEEK,
356 					"%02d", pt, ptlim);
357 				continue;
358 			case 'u':
359 				/*
360 				** From Arnold Robbins' strftime version 3.0:
361 				** "ISO 8601: Weekday as a decimal number
362 				** [1 (Monday) - 7]"
363 				** (ado, 1993-05-24)
364 				*/
365 				pt = _conv((t->tm_wday == 0) ?
366 					DAYSPERWEEK : t->tm_wday,
367 					"%d", pt, ptlim);
368 				continue;
369 			case 'V':	/* ISO 8601 week number */
370 			case 'G':	/* ISO 8601 year (four digits) */
371 			case 'g':	/* ISO 8601 year (two digits) */
372 /*
373 ** From Arnold Robbins' strftime version 3.0: "the week number of the
374 ** year (the first Monday as the first day of week 1) as a decimal number
375 ** (01-53)."
376 ** (ado, 1993-05-24)
377 **
378 ** From <https://www.cl.cam.ac.uk/~mgk25/iso-time.html> by Markus Kuhn:
379 ** "Week 01 of a year is per definition the first week which has the
380 ** Thursday in this year, which is equivalent to the week which contains
381 ** the fourth day of January. In other words, the first week of a new year
382 ** is the week which has the majority of its days in the new year. Week 01
383 ** might also contain days from the previous year and the week before week
384 ** 01 of a year is the last week (52 or 53) of the previous year even if
385 ** it contains days from the new year. A week starts with Monday (day 1)
386 ** and ends with Sunday (day 7). For example, the first week of the year
387 ** 1997 lasts from 1996-12-30 to 1997-01-05..."
388 ** (ado, 1996-01-02)
389 */
390 				{
391 					int	year;
392 					int	base;
393 					int	yday;
394 					int	wday;
395 					int	w;
396 
397 					year = t->tm_year;
398 					base = TM_YEAR_BASE;
399 					yday = t->tm_yday;
400 					wday = t->tm_wday;
401 					for ( ; ; ) {
402 						int	len;
403 						int	bot;
404 						int	top;
405 
406 						len = isleap_sum(year, base) ?
407 							DAYSPERLYEAR :
408 							DAYSPERNYEAR;
409 						/*
410 						** What yday (-3 ... 3) does
411 						** the ISO year begin on?
412 						*/
413 						bot = ((yday + 11 - wday) %
414 							DAYSPERWEEK) - 3;
415 						/*
416 						** What yday does the NEXT
417 						** ISO year begin on?
418 						*/
419 						top = bot -
420 							(len % DAYSPERWEEK);
421 						if (top < -3)
422 							top += DAYSPERWEEK;
423 						top += len;
424 						if (yday >= top) {
425 							++base;
426 							w = 1;
427 							break;
428 						}
429 						if (yday >= bot) {
430 							w = 1 + ((yday - bot) /
431 								DAYSPERWEEK);
432 							break;
433 						}
434 						--base;
435 						yday += isleap_sum(year, base) ?
436 							DAYSPERLYEAR :
437 							DAYSPERNYEAR;
438 					}
439 #ifdef XPG4_1994_04_09
440 					if ((w == 52 &&
441 						t->tm_mon == TM_JANUARY) ||
442 						(w == 1 &&
443 						t->tm_mon == TM_DECEMBER))
444 							w = 53;
445 #endif /* defined XPG4_1994_04_09 */
446 					if (*format == 'V')
447 						pt = _conv(w, "%02d",
448 							pt, ptlim);
449 					else if (*format == 'g') {
450 						*warnp = IN_ALL;
451 						pt = _yconv(year, base,
452 							false, true,
453 							pt, ptlim);
454 					} else	pt = _yconv(year, base,
455 							true, true,
456 							pt, ptlim);
457 				}
458 				continue;
459 			case 'v':
460 				/*
461 				** From Arnold Robbins' strftime version 3.0:
462 				** "date as dd-bbb-YYYY"
463 				** (ado, 1993-05-24)
464 				*/
465 				pt = _fmt("%e-%b-%Y", t, pt, ptlim, warnp);
466 				continue;
467 			case 'W':
468 				pt = _conv((t->tm_yday + DAYSPERWEEK -
469 					(t->tm_wday ?
470 					(t->tm_wday - 1) :
471 					(DAYSPERWEEK - 1))) / DAYSPERWEEK,
472 					"%02d", pt, ptlim);
473 				continue;
474 			case 'w':
475 				pt = _conv(t->tm_wday, "%d", pt, ptlim);
476 				continue;
477 			case 'X':
478 				pt = _fmt(Locale->X_fmt, t, pt, ptlim, warnp);
479 				continue;
480 			case 'x':
481 				{
482 				enum warn warn2 = IN_SOME;
483 
484 				pt = _fmt(Locale->x_fmt, t, pt, ptlim, &warn2);
485 				if (warn2 == IN_ALL)
486 					warn2 = IN_THIS;
487 				if (warn2 > *warnp)
488 					*warnp = warn2;
489 				}
490 				continue;
491 			case 'y':
492 				*warnp = IN_ALL;
493 				pt = _yconv(t->tm_year, TM_YEAR_BASE,
494 					false, true,
495 					pt, ptlim);
496 				continue;
497 			case 'Y':
498 				pt = _yconv(t->tm_year, TM_YEAR_BASE,
499 					true, true,
500 					pt, ptlim);
501 				continue;
502 			case 'Z':
503 #ifdef TM_ZONE
504 				pt = _add(t->TM_ZONE, pt, ptlim);
505 #elif HAVE_TZNAME
506 				if (t->tm_isdst >= 0)
507 					pt = _add(tzname[t->tm_isdst != 0],
508 						pt, ptlim);
509 #endif
510 				/*
511 				** C99 and later say that %Z must be
512 				** replaced by the empty string if the
513 				** time zone abbreviation is not
514 				** determinable.
515 				*/
516 				continue;
517 			case 'z':
518 #if defined TM_GMTOFF || USG_COMPAT || ALTZONE
519 				{
520 				long		diff;
521 				char const *	sign;
522 				bool negative;
523 
524 # ifdef TM_GMTOFF
525 				diff = t->TM_GMTOFF;
526 # else
527 				/*
528 				** C99 and later say that the UT offset must
529 				** be computed by looking only at
530 				** tm_isdst. This requirement is
531 				** incorrect, since it means the code
532 				** must rely on magic (in this case
533 				** altzone and timezone), and the
534 				** magic might not have the correct
535 				** offset. Doing things correctly is
536 				** tricky and requires disobeying the standard;
537 				** see GNU C strftime for details.
538 				** For now, punt and conform to the
539 				** standard, even though it's incorrect.
540 				**
541 				** C99 and later say that %z must be replaced by
542 				** the empty string if the time zone is not
543 				** determinable, so output nothing if the
544 				** appropriate variables are not available.
545 				*/
546 				if (t->tm_isdst < 0)
547 					continue;
548 				if (t->tm_isdst == 0)
549 #  if USG_COMPAT
550 					diff = -timezone;
551 #  else
552 					continue;
553 #  endif
554 				else
555 #  if ALTZONE
556 					diff = -altzone;
557 #  else
558 					continue;
559 #  endif
560 # endif
561 				negative = diff < 0;
562 				if (diff == 0) {
563 # ifdef TM_ZONE
564 				  negative = t->TM_ZONE[0] == '-';
565 # else
566 				  negative = t->tm_isdst < 0;
567 #  if HAVE_TZNAME
568 				  if (tzname[t->tm_isdst != 0][0] == '-')
569 				    negative = true;
570 #  endif
571 # endif
572 				}
573 				if (negative) {
574 					sign = "-";
575 					diff = -diff;
576 				} else	sign = "+";
577 				pt = _add(sign, pt, ptlim);
578 				diff /= SECSPERMIN;
579 				diff = (diff / MINSPERHOUR) * 100 +
580 					(diff % MINSPERHOUR);
581 				pt = _conv(diff, "%04d", pt, ptlim);
582 				}
583 #endif
584 				continue;
585 			case '+':
586 				pt = _fmt(Locale->date_fmt, t, pt, ptlim,
587 					warnp);
588 				continue;
589 			case '%':
590 			/*
591 			** X311J/88-090 (4.12.3.5): if conversion char is
592 			** undefined, behavior is undefined. Print out the
593 			** character itself as printf(3) also does.
594 			*/
595 			default:
596 				break;
597 			}
598 		}
599 		if (pt == ptlim)
600 			break;
601 		*pt++ = *format;
602 	}
603 	return pt;
604 }
605 
606 static char *
607 _conv(int n, const char *format, char *pt, const char *ptlim)
608 {
609 	char	buf[INT_STRLEN_MAXIMUM(int) + 1];
610 
611 	sprintf(buf, format, n);
612 	return _add(buf, pt, ptlim);
613 }
614 
615 static char *
616 _add(const char *str, char *pt, const char *ptlim)
617 {
618 	while (pt < ptlim && (*pt = *str++) != '\0')
619 		++pt;
620 	return pt;
621 }
622 
623 /*
624 ** POSIX and the C Standard are unclear or inconsistent about
625 ** what %C and %y do if the year is negative or exceeds 9999.
626 ** Use the convention that %C concatenated with %y yields the
627 ** same output as %Y, and that %Y contains at least 4 bytes,
628 ** with more only if necessary.
629 */
630 
631 static char *
632 _yconv(int a, int b, bool convert_top, bool convert_yy,
633        char *pt, const char *ptlim)
634 {
635 	register int	lead;
636 	register int	trail;
637 
638 	int DIVISOR = 100;
639 	trail = a % DIVISOR + b % DIVISOR;
640 	lead = a / DIVISOR + b / DIVISOR + trail / DIVISOR;
641 	trail %= DIVISOR;
642 	if (trail < 0 && lead > 0) {
643 		trail += DIVISOR;
644 		--lead;
645 	} else if (lead < 0 && trail > 0) {
646 		trail -= DIVISOR;
647 		++lead;
648 	}
649 	if (convert_top) {
650 		if (lead == 0 && trail < 0)
651 			pt = _add("-0", pt, ptlim);
652 		else	pt = _conv(lead, "%02d", pt, ptlim);
653 	}
654 	if (convert_yy)
655 		pt = _conv(((trail < 0) ? -trail : trail), "%02d", pt, ptlim);
656 	return pt;
657 }
658