public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Theodore Tso <tytso@mit.edu>
To: SA products <super.aorta@ntlworld.com>
Cc: linux-kernel@vger.kernel.org
Subject: Re: faking time
Date: Mon, 11 Feb 2002 22:47:23 -0500	[thread overview]
Message-ID: <20020211224723.A5514@thunk.org> (raw)
In-Reply-To: <3C67AFD3.722C5471@ntlworld.com>
In-Reply-To: <3C67AFD3.722C5471@ntlworld.com>; from super.aorta@ntlworld.com on Mon, Feb 11, 2002 at 11:49:39AM +0000

[-- Attachment #1: Type: text/plain, Size: 810 bytes --]

On Mon, Feb 11, 2002 at 11:49:39AM +0000, SA products wrote:
> 
> Dear Kernel list,
> 
> I want to fake the time returned by the time() system call so that
> for a limited number of user space programs the time can be set to
> the future or the past without affecting other applications and
> without affecting system time-- Ideally I would like to install a
> loadable module to accomplish this- Any hints ? Any starting points?

Here's an LD_PRELOAD shared library that will do the trick... just
export the environment variable FAKETIME with the time that you'd
like, and then export the LD_PRELOAD environment variable to point
that the faketime.so library, and then execute your program.  All
programs that have these two environment variables set will have their
time faked out accordingly.

							- Ted

[-- Attachment #2: faketime.c --]
[-- Type: text/x-csrc, Size: 27585 bytes --]

/* Convert struct partime into time_t.  */

/* Copyright 1992, 1993, 1994, 1995 Paul Eggert
   Distributed under license by the Free Software Foundation, Inc.

This file is part of RCS.

RCS is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2, or (at your option)
any later version.

RCS is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with RCS; see the file COPYING.
If not, write to the Free Software Foundation,
59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.

Report problems and direct all questions to:

    rcs-bugs@cs.purdue.edu

*/

#include <stdlib.h>
#include <time.h>
#include <sys/time.h>
#include <ctype.h>

#define P(x) x

#define TM_UNDEFINED (-1)
#define TM_DEFINED(x) (0 <= (x))

#define TM_UNDEFINED_ZONE ((long) -24 * 60 * 60)
#define TM_LOCAL_ZONE (TM_UNDEFINED_ZONE - 1)

struct partime {
	/*
	* This structure describes the parsed time.
	* Only the following tm_* values in it are used:
	*	sec, min, hour, mday, mon, year, wday, yday.
	* If TM_UNDEFINED(value), the parser never found the value.
	* The tm_year field is the actual year, not the year - 1900;
	* but see ymodulus below.
	*/
	struct tm tm;

	/*
	* If !TM_UNDEFINED(ymodulus),
	* then tm.tm_year is actually modulo ymodulus.
	*/
	int ymodulus;

	/*
	* Week of year, ISO 8601 style.
	* If TM_UNDEFINED(yweek), the parser never found yweek.
	* Weeks start on Mondays.
	* Week 1 includes Jan 4.
	*/
	int yweek;

	/* Seconds east of UTC; or TM_LOCAL_ZONE or TM_UNDEFINED_ZONE.  */
	long zone;
};

static char *partime (char const *, struct partime *);
static char *parzone (char const *, long *);

static int isleap P((int));
static int month_days P((struct tm const*));
static time_t maketime P((struct partime const*,time_t));


/* Lookup tables for names of months, weekdays, time zones.  */

#define NAME_LENGTH_MAXIMUM 4

struct name_val {
	char name[NAME_LENGTH_MAXIMUM];
	int val;
};


static char const *parse_decimal P((char const*,int,int,int,int,int*,int*));
static char const *parse_fixed P((char const*,int,int*));
static char const *parse_pattern_letter P((char const*,int,struct partime*));
static char const *parse_prefix P((char const*,struct partime*,int*));
static char const *parse_ranged P((char const*,int,int,int,int*));
static int lookup P((char const*,struct name_val const[]));
static int merge_partime P((struct partime*, struct partime const*));
static void undefine P((struct partime*));


static struct name_val const month_names[] = {
	{"jan",0}, {"feb",1}, {"mar",2}, {"apr",3}, {"may",4}, {"jun",5},
	{"jul",6}, {"aug",7}, {"sep",8}, {"oct",9}, {"nov",10}, {"dec",11},
	{"", TM_UNDEFINED}
};

static struct name_val const weekday_names[] = {
	{"sun",0}, {"mon",1}, {"tue",2}, {"wed",3}, {"thu",4}, {"fri",5}, {"sat",6},
	{"", TM_UNDEFINED}
};

#define hr60nonnegative(t)  ((t)/100 * 60  +  (t)%100)
#define hr60(t)  ((t)<0 ? -hr60nonnegative(-(t)) : hr60nonnegative(t))
#define zs(t,s)  {s, hr60(t)}
#define zd(t,s,d)  zs(t, s),  zs((t)+100, d)

static struct name_val const zone_names[] = {
	zs(-1000, "hst"),		/* Hawaii */
	zd(-1000,"hast","hadt"),/* Hawaii-Aleutian */
	zd(- 900,"akst","akdt"),/* Alaska */
	zd(- 800, "pst", "pdt"),/* Pacific */
	zd(- 700, "mst", "mdt"),/* Mountain */
	zd(- 600, "cst", "cdt"),/* Central */
	zd(- 500, "est", "edt"),/* Eastern */
	zd(- 400, "ast", "adt"),/* Atlantic */
	zd(- 330, "nst", "ndt"),/* Newfoundland */
	zs(  000, "utc"),		/* Coordinated Universal */
	zs(  000, "cut"),		/* " */
	zs(  000,  "ut"),		/* Universal */
	zs(  000,   "z"),		/* Zulu (required by ISO 8601) */
	zd(  000, "gmt", "bst"),/* Greenwich Mean, British Summer */
	zs(  000, "wet"),		/* Western Europe */
	zs(  100, "met"),		/* Middle Europe */
	zs(  100, "cet"),		/* Central Europe */
	zs(  200, "eet"),		/* Eastern Europe */
	zs(  530, "ist"),		/* India */
	zd(  900, "jst", "jdt"),/* Japan */
	zd(  900, "kst", "kdt"),/* Korea */
	zd( 1200,"nzst","nzdt"),/* New Zealand */
	{ "lt", 1 },
#if 0
	/* The following names are duplicates or are not well attested.  */
	zs(-1100, "sst"),		/* Samoa */
	zs(-1000, "tht"),		/* Tahiti */
	zs(- 930, "mqt"),		/* Marquesas */
	zs(- 900, "gbt"),		/* Gambier */
	zd(- 900, "yst", "ydt"),/* Yukon - name is no longer used */
	zs(- 830, "pit"),		/* Pitcairn */
	zd(- 500, "cst", "cdt"),/* Cuba */
	zd(- 500, "ast", "adt"),/* Acre */
	zd(- 400, "wst", "wdt"),/* Western Brazil */
	zd(- 400, "ast", "adt"),/* Andes */
	zd(- 400, "cst", "cdt"),/* Chile */
	zs(- 300, "wgt"),		/* Western Greenland */
	zd(- 300, "est", "edt"),/* Eastern South America */
	zs(- 300, "mgt"),		/* Middle Greenland */
	zd(- 200, "fst", "fdt"),/* Fernando de Noronha */
	zs(- 100, "egt"),		/* Eastern Greenland */
	zs(- 100, "aat"),		/* Atlantic Africa */
	zs(- 100, "act"),		/* Azores and Canaries */
	zs(  000, "wat"),		/* West Africa */
	zs(  100, "cat"),		/* Central Africa */
	zd(  100, "mez","mesz"),/* Mittel-Europaeische Zeit */
	zs(  200, "sat"),		/* South Africa */
	zd(  200, "ist", "idt"),/* Israel */
	zs(  300, "eat"),		/* East Africa */
	zd(  300, "ast", "adt"),/* Arabia */
	zd(  300, "msk", "msd"),/* Moscow */
	zd(  330, "ist", "idt"),/* Iran */
	zs(  400, "gst"),		/* Gulf */
	zs(  400, "smt"),		/* Seychelles & Mascarene */
	zd(  400, "esk", "esd"),/* Yekaterinburg */
	zd(  400, "bsk", "bsd"),/* Baku */
	zs(  430, "aft"),		/* Afghanistan */
	zd(  500, "osk", "osd"),/* Omsk */
	zs(  500, "pkt"),		/* Pakistan */
	zd(  500, "tsk", "tsd"),/* Tashkent */
	zs(  545, "npt"),		/* Nepal */
	zs(  600, "bgt"),		/* Bangladesh */
	zd(  600, "nsk", "nsd"),/* Novosibirsk */
	zs(  630, "bmt"),		/* Burma */
	zs(  630, "cct"),		/* Cocos */
	zs(  700, "ict"),		/* Indochina */
	zs(  700, "jvt"),		/* Java */
	zd(  700, "isk", "isd"),/* Irkutsk */
	zs(  800, "hkt"),		/* Hong Kong */
	zs(  800, "pst"),		/* Philippines */
	zs(  800, "sgt"),		/* Singapore */
	zd(  800, "cst", "cdt"),/* China */
	zd(  800, "ust", "udt"),/* Ulan Bator */
	zd(  800, "wst", "wst"),/* Western Australia */
	zd(  800, "ysk", "ysd"),/* Yakutsk */
	zs(  900, "blt"),		/* Belau */
	zs(  900, "mlt"),		/* Moluccas */
	zd(  900, "vsk", "vsd"),/* Vladivostok */
	zd(  930, "cst", "cst"),/* Central Australia */
	zs( 1000, "gst"),		/* Guam */
	zd( 1000, "gsk", "gsd"),/* Magadan */
	zd( 1000, "est", "est"),/* Eastern Australia */
	zd( 1100,"lhst","lhst"),/* Lord Howe */
	zd( 1100, "psk", "psd"),/* Petropavlovsk-Kamchatski */
	zs( 1100,"ncst"),		/* New Caledonia */
	zs( 1130,"nrft"),		/* Norfolk */
	zd( 1200, "ask", "asd"),/* Anadyr */
	zs( 1245,"nz-chat"),	/* Chatham */
	zs( 1300, "tgt"),		/* Tongatapu */
#endif
	{"", -1}
};

	static int
lookup (s, table)
	char const *s;
	struct name_val const table[];
/* Look for a prefix of S in TABLE, returning val for first matching entry.  */
{
	int j;
	char buf[NAME_LENGTH_MAXIMUM];

	for (j = 0;  j < NAME_LENGTH_MAXIMUM;  j++) {
		unsigned char c = *s++;
		buf[j] = isupper (c) ? tolower (c) : c;
		if (!isalpha (c))
			break;
	}
	for (;  table[0].name[0];  table++)
		for (j = 0;  buf[j] == table[0].name[j];  )
			if (++j == NAME_LENGTH_MAXIMUM  ||  !table[0].name[j])
				goto done;
  done:
	return table[0].val;
}


	static void
undefine (t) struct partime *t;
/* Set *T to ``undefined'' values.  */
{
	t->tm.tm_sec = t->tm.tm_min = t->tm.tm_hour = t->tm.tm_mday = t->tm.tm_mon
		= t->tm.tm_year = t->tm.tm_wday = t->tm.tm_yday
		= t->ymodulus = t->yweek
		= TM_UNDEFINED;
	t->zone = TM_UNDEFINED_ZONE;
}

/*
* Array of patterns to look for in a date string.
* Order is important: we look for the first matching pattern
* whose values do not contradict values that we already know about.
* See `parse_pattern_letter' below for the meaning of the pattern codes.
*/
static char const * const patterns[] = {
	/*
	* These traditional patterns must come first,
	* to prevent an ISO 8601 format from misinterpreting their prefixes.
	*/
	"E_n_y", "x", /* RFC 822 */
	"E_n", "n_E", "n", "t:m:s_A", "t:m_A", "t_A", /* traditional */
	"y/N/D$", /* traditional RCS */

	/* ISO 8601:1988 formats, generalized a bit.  */
	"y-N-D$", "4ND$", "Y-N$",
	"RND$", "-R=N$", "-R$", "--N=D$", "N=DT",
	"--N$", "---D$", "DT",
	"Y-d$", "4d$", "R=d$", "-d$", "dT",
	"y-W-X", "yWX", "y=W",
	"-r-W-X", "r-W-XT", "-rWX", "rWXT", "-W=X", "W=XT", "-W",
	"-w-X", "w-XT", "---X$", "XT", "4$",
	"T",
	"h:m:s$", "hms$", "h:m$", "hm$", "h$", "-m:s$", "-ms$", "-m$", "--s$",
	"Y", "Z",

	0
};

	static char const *
parse_prefix (str, t, pi) char const *str; struct partime *t; int *pi;
/*
* Parse an initial prefix of STR, setting *T accordingly.
* Return the first character after the prefix, or 0 if it couldn't be parsed.
* Start with pattern *PI; if success, set *PI to the next pattern to try.
* Set *PI to -1 if we know there are no more patterns to try;
* if *PI is initially negative, give up immediately.
*/
{
	int i = *pi;
	char const *pat;
	unsigned char c;

	if (i < 0)
		return 0;

	/* Remove initial noise.  */
	while (!isalnum (c = *str)  &&  c != '-'  &&  c != '+') {
		if (!c) {
			undefine (t);
			*pi = -1;
			return str;
		}
		str++;
	}

	/* Try a pattern until one succeeds.  */
	while ((pat = patterns[i++]) != 0) {
		char const *s = str;
		undefine (t);
		do {
			if (!(c = *pat++)) {
				*pi = i;
				return s;
			}
		} while ((s = parse_pattern_letter (s, c, t)) != 0);
	}

	return 0;
}

	static char const *
parse_fixed (s, digits, res) char const *s; int digits, *res;
/*
* Parse an initial prefix of S of length DIGITS; it must be a number.
* Store the parsed number into *RES.
* Return the first character after the prefix, or 0 if it couldn't be parsed.
*/
{
	int n = 0;
	char const *lim = s + digits;
	while (s < lim) {
		unsigned d = *s++ - '0';
		if (9 < d)
			return 0;
		n = 10*n + d;
	}
	*res = n;
	return s;
}

	static char const *
parse_ranged (s, digits, lo, hi, res) char const *s; int digits, lo, hi, *res;
/*
* Parse an initial prefix of S of length DIGITS;
* it must be a number in the range LO through HI.
* Store the parsed number into *RES.
* Return the first character after the prefix, or 0 if it couldn't be parsed.
*/
{
	s = parse_fixed (s, digits, res);
	return  s && lo<=*res && *res<=hi  ?  s  :  0;
}

	static char const *
parse_decimal (s, digits, lo, hi, resolution, res, fres)
	char const *s;
	int digits, lo, hi, resolution, *res, *fres;
/*
* Parse an initial prefix of S of length DIGITS;
* it must be a number in the range LO through HI
* and it may be followed by a fraction that is to be computed using RESOLUTION.
* Store the parsed number into *RES; store the fraction times RESOLUTION,
* rounded to the nearest integer, into *FRES.
* Return the first character after the prefix, or 0 if it couldn't be parsed.
*/
{
	s = parse_fixed (s, digits, res);
	if (s && lo<=*res && *res<=hi) {
		int f = 0;
		if ((s[0]==',' || s[0]=='.')  &&  isdigit ((unsigned char) s[1])) {
			char const *s1 = ++s;
			int num10 = 0, denom10 = 10, product;
			while (isdigit ((unsigned char) *++s))
				denom10 *= 10;
			s = parse_fixed (s1, s - s1, &num10);
			product = num10*resolution;
			f = (product + (denom10>>1)) / denom10;
			f -= f & (product%denom10 == denom10>>1); /* round to even */
			if (f < 0  ||  product/resolution != num10)
				return 0; /* overflow */
		}
		*fres = f;
		return s;
	}
	return 0;
}

	static char *
parzone (s, zone) char const *s; long *zone;
/*
* Parse an initial prefix of S; it must denote a time zone.
* Set *ZONE to the number of seconds east of GMT,
* or to TM_LOCAL_ZONE if it is the local time zone.
* Return the first character after the prefix, or 0 if it couldn't be parsed.
*/
{
	char sign;
	int hh, mm, ss;
	int minutesEastOfUTC;
	long offset, z;

	/*
	* The formats are LT, n, n DST, nDST, no, o
	* where n is a time zone name
	* and o is a time zone offset of the form [-+]hh[:mm[:ss]].
	*/
	switch (*s) {
		case '-': case '+':
			z = 0;
			break;

		default:
			minutesEastOfUTC = lookup (s, zone_names);
			if (minutesEastOfUTC == -1)
				return 0;

			/* Don't bother to check rest of spelling.  */
			while (isalpha ((unsigned char) *s))
				s++;

			/* Don't modify LT.  */
			if (minutesEastOfUTC == 1) {
				*zone = TM_LOCAL_ZONE;
				return (char *) s;
			}

			z = minutesEastOfUTC * 60L;

			/* Look for trailing " DST".  */
			if (
				(s[-1]=='T' || s[-1]=='t') &&
				(s[-2]=='S' || s[-2]=='s') &&
				(s[-3]=='D' || s[-3]=='t')
			)
				goto trailing_dst;
			while (isspace ((unsigned char) *s))
				s++;
			if (
				(s[0]=='D' || s[0]=='d') &&
				(s[1]=='S' || s[1]=='s') &&
				(s[2]=='T' || s[2]=='t')
			) {
				s += 3;
			  trailing_dst:
				*zone = z + 60*60;
				return (char *) s;
			}

			switch (*s) {
				case '-': case '+': break;
				default: return (char *) s;
			}
	}
	sign = *s++;

	if (!(s = parse_ranged (s, 2, 0, 23, &hh)))
		return 0;
	mm = ss = 0;
	if (*s == ':')
		s++;
	if (isdigit ((unsigned char) *s)) {
		if (!(s = parse_ranged (s, 2, 0, 59, &mm)))
			return 0;
		if (*s==':' && s[-3]==':' && isdigit ((unsigned char) s[1])) {
			if (!(s = parse_ranged (s + 1, 2, 0, 59, &ss)))
				return 0;
		}
	}
	if (isdigit ((unsigned char) *s))
		return 0;
	offset = (hh*60 + mm)*60L + ss;
	*zone = z + (sign=='-' ? -offset : offset);
	/*
	* ?? Are fractions allowed here?
	* If so, they're not implemented.
	*/
	return (char *) s;
}

	static char const *
parse_pattern_letter (s, c, t) char const *s; int c; struct partime *t;
/*
* Parse an initial prefix of S, matching the pattern whose code is C.
* Set *T accordingly.
* Return the first character after the prefix, or 0 if it couldn't be parsed.
*/
{
	switch (c) {
		case '$': /* The next character must be a non-digit.  */
			if (isdigit ((unsigned char) *s))
				return 0;
			break;

		case '-': case '/': case ':':
			/* These characters stand for themselves.  */
			if (*s++ != c)
				return 0;
			break;

		case '4': /* 4-digit year */
			s = parse_fixed (s, 4, &t->tm.tm_year);
			break;

		case '=': /* optional '-' */
			s  +=  *s == '-';
			break;

		case 'A': /* AM or PM */
			/*
			* This matches the regular expression [AaPp][Mm]?.
			* It must not be followed by a letter or digit;
			* otherwise it would match prefixes of strings like "PST".
			*/
			switch (*s++) {
				case 'A': case 'a':
					if (t->tm.tm_hour == 12)
						t->tm.tm_hour = 0;
					break;

				case 'P': case 'p':
					if (t->tm.tm_hour != 12)
						t->tm.tm_hour += 12;
					break;

				default: return 0;
			}
			switch (*s) {
				case 'M': case 'm': s++; break;
			}
			if (isalnum (*s))
				return 0;
			break;

		case 'D': /* day of month [01-31] */
			s = parse_ranged (s, 2, 1, 31, &t->tm.tm_mday);
			break;

		case 'd': /* day of year [001-366] */
			s = parse_ranged (s, 3, 1, 366, &t->tm.tm_yday);
			t->tm.tm_yday--;
			break;

		case 'E': /* extended day of month [1-9, 01-31] */
			s = parse_ranged (s, (
				isdigit ((unsigned char) s[0]) &&
				isdigit ((unsigned char) s[1])
			) + 1, 1, 31, &t->tm.tm_mday);
			break;

		case 'h': /* hour [00-23 followed by optional fraction] */
			{
				int frac;
				s = parse_decimal (s, 2, 0, 23, 60*60, &t->tm.tm_hour, &frac);
				t->tm.tm_min = frac / 60;
				t->tm.tm_sec = frac % 60;
			}
			break;

		case 'm': /* minute [00-59 followed by optional fraction] */
			s = parse_decimal (s, 2, 0, 59, 60, &t->tm.tm_min, &t->tm.tm_sec);
			break;

		case 'n': /* month name [e.g. "Jan"] */
			if (!TM_DEFINED (t->tm.tm_mon = lookup (s, month_names)))
				return 0;
			/* Don't bother to check rest of spelling.  */
			while (isalpha ((unsigned char) *s))
				s++;
			break;

		case 'N': /* month [01-12] */
			s = parse_ranged (s, 2, 1, 12, &t->tm.tm_mon);
			t->tm.tm_mon--;
			break;

		case 'r': /* year % 10 (remainder in origin-0 decade) [0-9] */
			s = parse_fixed (s, 1, &t->tm.tm_year);
			t->ymodulus = 10;
			break;

		case_R:
		case 'R': /* year % 100 (remainder in origin-0 century) [00-99] */
			s = parse_fixed (s, 2, &t->tm.tm_year);
			t->ymodulus = 100;
			break;

		case 's': /* second [00-60 followed by optional fraction] */
			{
				int frac;
				s = parse_decimal (s, 2, 0, 60, 1, &t->tm.tm_sec, &frac);
				t->tm.tm_sec += frac;
			}
			break;

		case 'T': /* 'T' or 't' */
			switch (*s++) {
				case 'T': case 't': break;
				default: return 0;
			}
			break;

		case 't': /* traditional hour [1-9 or 01-12] */
			s = parse_ranged (s, (
				isdigit ((unsigned char) s[0]) && isdigit ((unsigned char) s[1])
			) + 1, 1, 12, &t->tm.tm_hour);
			break;

		case 'w': /* 'W' or 'w' only (stands for current week) */
			switch (*s++) {
				case 'W': case 'w': break;
				default: return 0;
			}
			break;

		case 'W': /* 'W' or 'w', followed by a week of year [00-53] */
			switch (*s++) {
				case 'W': case 'w': break;
				default: return 0;
			}
			s = parse_ranged (s, 2, 0, 53, &t->yweek);
			break;

		case 'X': /* weekday (1=Mon ... 7=Sun) [1-7] */
			s = parse_ranged (s, 1, 1, 7, &t->tm.tm_wday);
			t->tm.tm_wday--;
			break;

		case 'x': /* weekday name [e.g. "Sun"] */
			if (!TM_DEFINED (t->tm.tm_wday = lookup (s, weekday_names)))
				return 0;
			/* Don't bother to check rest of spelling.  */
			while (isalpha ((unsigned char) *s))
				s++;
			break;

		case 'y': /* either R or Y */
			if (
				isdigit ((unsigned char) s[0]) &&
				isdigit ((unsigned char) s[1]) &&
				!isdigit ((unsigned char) s[2])
			)
				goto case_R;
			/* fall into */
		case 'Y': /* year in full [4 or more digits] */
			{
				int len = 0;
				while (isdigit ((unsigned char) s[len]))
					len++;
				if (len < 4)
					return 0;
				s = parse_fixed (s, len, &t->tm.tm_year);
			}
			break;

		case 'Z': /* time zone */
			s = parzone (s, &t->zone);
			break;

		case '_': /* possibly empty sequence of non-alphanumerics */
			while (!isalnum (*s)  &&  *s)
				s++;
			break;

		default: /* bad pattern */
			return 0;
	}
	return s;
}

	static int
merge_partime (t, u) struct partime *t; struct partime const *u;
/*
* If there is no conflict, merge into *T the additional information in *U
* and return 0.  Otherwise do nothing and return -1.
*/
{
#	define conflict(a,b) ((a) != (b)  &&  TM_DEFINED (a)  &&  TM_DEFINED (b))
	if (
		conflict (t->tm.tm_sec, u->tm.tm_sec) ||
		conflict (t->tm.tm_min, u->tm.tm_min) ||
		conflict (t->tm.tm_hour, u->tm.tm_hour) ||
		conflict (t->tm.tm_mday, u->tm.tm_mday) ||
		conflict (t->tm.tm_mon, u->tm.tm_mon) ||
		conflict (t->tm.tm_year, u->tm.tm_year) ||
		conflict (t->tm.tm_wday, u->tm.tm_yday) ||
		conflict (t->ymodulus, u->ymodulus) ||
		conflict (t->yweek, u->yweek) ||
		(
			t->zone != u->zone &&
			t->zone != TM_UNDEFINED_ZONE &&
			u->zone != TM_UNDEFINED_ZONE
		)
	)
		return -1;
#	undef conflict
#	define merge_(a,b) if (TM_DEFINED (b)) (a) = (b);
	merge_ (t->tm.tm_sec, u->tm.tm_sec)
	merge_ (t->tm.tm_min, u->tm.tm_min)
	merge_ (t->tm.tm_hour, u->tm.tm_hour)
	merge_ (t->tm.tm_mday, u->tm.tm_mday)
	merge_ (t->tm.tm_mon, u->tm.tm_mon)
	merge_ (t->tm.tm_year, u->tm.tm_year)
	merge_ (t->tm.tm_wday, u->tm.tm_yday)
	merge_ (t->ymodulus, u->ymodulus)
	merge_ (t->yweek, u->yweek)
#	undef merge_
	if (u->zone != TM_UNDEFINED_ZONE) t->zone = u->zone;
	return 0;
}

	static char *
partime (s, t) char const *s; struct partime *t;
/*
* Parse a date/time prefix of S, putting the parsed result into *T.
* Return the first character after the prefix.
* The prefix may contain no useful information;
* in that case, *T will contain only undefined values.
*/
{
	struct partime p;

	undefine (t);
	while (*s) {
		int i = 0;
		char const *s1;
		do {
			if (!(s1 = parse_prefix (s, &p, &i)))
				return (char *) s;
		} while (merge_partime (t, &p) != 0);
		s = s1;
	}
	return (char *) s;
}
	

/*
* For maximum portability, use only localtime and gmtime.
* Make no assumptions about the time_t epoch or the range of time_t values.
* Avoid mktime because it's not universal and because there's no easy,
* portable way for mktime to yield the inverse of gmtime.
*/

#define TM_YEAR_ORIGIN 1900

	static int
isleap(y)
	int y;
{
	return (y&3) == 0  &&  (y%100 != 0 || y%400 == 0);
}

static int const month_yday[] = {
	/* days in year before start of months 0-12 */
	0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334, 365
};

/* Yield the number of days in TM's month.  */
	static int
month_days(tm)
	struct tm const *tm;
{
	int m = tm->tm_mon;
	return month_yday[m+1] - month_yday[m]
		+ (m==1 && isleap(tm->tm_year + TM_YEAR_ORIGIN));
}

/*
* Convert UNIXTIME to struct tm form.
* Use gmtime if available and if !LOCALZONE, localtime otherwise.
*/
	static struct tm *
time2tm(time_t unixtime, int localzone)
{
	struct tm *tm;
#	if TZ_must_be_set
		static char const *TZ;
		if (!TZ  &&  !(TZ = getenv("TZ")))
			faterror("The TZ environment variable is not set; please set it to your timezone");
#	endif
	if (localzone  ||  !(tm = gmtime(&unixtime)))
		tm = localtime(&unixtime);
	return tm;
}

/* Yield A - B, measured in seconds.  */
	static time_t
difftm(struct tm const *a, struct tm const *b)
{
	int ay = a->tm_year + (TM_YEAR_ORIGIN - 1);
	int by = b->tm_year + (TM_YEAR_ORIGIN - 1);
	int difference_in_day_of_year = a->tm_yday - b->tm_yday;
	int intervening_leap_days = (
		((ay >> 2) - (by >> 2))
		- (ay/100 - by/100)
		+ ((ay/100 >> 2) - (by/100 >> 2))
	);
	time_t difference_in_years = ay - by;
	time_t difference_in_days = (
		difference_in_years*365
		+ (intervening_leap_days + difference_in_day_of_year)
	);
	return
		(
			(
				24*difference_in_days
				+ (a->tm_hour - b->tm_hour)
			)*60 + (a->tm_min - b->tm_min)
		)*60 + (a->tm_sec - b->tm_sec);
}

/*
* Adjust time T by adding SECONDS.  SECONDS must be at most 24 hours' worth.
* Adjust only T's year, mon, mday, hour, min and sec members;
* plus adjust wday if it is defined.
*/
	static void
adjzone(struct tm *t, long seconds)
{
	/*
	* This code can be off by a second if SECONDS is not a multiple of 60,
	* if T is local time, and if a leap second happens during this minute.
	* But this bug has never occurred, and most likely will not ever occur.
	* Liberia, the last country for which SECONDS % 60 was nonzero,
	* switched to UTC in May 1972; the first leap second was in June 1972.
	*/
	int leap_second = t->tm_sec == 60;
	long sec = seconds + (t->tm_sec - leap_second);
	if (sec < 0) {
	    if ((t->tm_min -= (59-sec)/60) < 0) {
		if ((t->tm_hour -= (59-t->tm_min)/60) < 0) {
		    t->tm_hour += 24;
		    if (TM_DEFINED(t->tm_wday)  &&  --t->tm_wday < 0)
			t->tm_wday = 6;
		    if (--t->tm_mday <= 0) {
			if (--t->tm_mon < 0) {
			    --t->tm_year;
			    t->tm_mon = 11;
			}
			t->tm_mday = month_days(t);
		    }
		}
		t->tm_min += 24 * 60;
	    }
	    sec += 24L * 60 * 60;
	} else
	    if (60 <= (t->tm_min += sec/60))
		if (24 <= (t->tm_hour += t->tm_min/60)) {
		    t->tm_hour -= 24;
		    if (TM_DEFINED(t->tm_wday)  &&  ++t->tm_wday == 7)
			t->tm_wday = 0;
		    if (month_days(t) < ++t->tm_mday) {
			if (11 < ++t->tm_mon) {
			    ++t->tm_year;
			    t->tm_mon = 0;
			}
			t->tm_mday = 1;
		    }
		}
	t->tm_min %= 60;
	t->tm_sec = (int) (sec%60) + leap_second;
}

/*
* Convert TM to time_t, using localtime if LOCALZONE and gmtime otherwise.
* Use only TM's year, mon, mday, hour, min, and sec members.
* Ignore TM's old tm_yday and tm_wday, but fill in their correct values.
* Yield -1 on failure (e.g. a member out of range).
* Posix 1003.1-1990 doesn't allow leap seconds, but some implementations
* have them anyway, so allow them if localtime/gmtime does.
*/
	static time_t
tm2time(struct tm *tm, int localzone)
{
	/* Cache the most recent t,tm pairs; 1 for gmtime, 1 for localtime.  */
	static time_t t_cache[2];
	static struct tm tm_cache[2];

	time_t d, gt;
	struct tm const *gtm;
	/*
	* The maximum number of iterations should be enough to handle any
	* combinations of leap seconds, time zone rule changes, and solar time.
	* 4 is probably enough; we use a bigger number just to be safe.
	*/
	int remaining_tries = 8;

	/* Avoid subscript errors.  */
	if (12 <= (unsigned)tm->tm_mon)
	    return -1;

	tm->tm_yday = month_yday[tm->tm_mon] + tm->tm_mday
		-  (tm->tm_mon<2  ||  ! isleap(tm->tm_year + TM_YEAR_ORIGIN));

	/* Make a first guess.  */
	gt = t_cache[localzone];
	gtm = gt ? &tm_cache[localzone] : time2tm(gt,localzone);

	/* Repeatedly use the error from the guess to improve the guess.  */
	while ((d = difftm(tm, gtm)) != 0) {
		if (--remaining_tries == 0)
			return -1;
		gt += d;
		gtm = time2tm(gt,localzone);
	}
	t_cache[localzone] = gt;
	tm_cache[localzone] = *gtm;

	/*
	* Check that the guess actually matches;
	* overflow can cause difftm to yield 0 even on differing times,
	* or tm may have members out of range (e.g. bad leap seconds).
	*/
	if (   (tm->tm_year ^ gtm->tm_year)
	    |  (tm->tm_mon  ^ gtm->tm_mon)
	    |  (tm->tm_mday ^ gtm->tm_mday)
	    |  (tm->tm_hour ^ gtm->tm_hour)
	    |  (tm->tm_min  ^ gtm->tm_min)
	    |  (tm->tm_sec  ^ gtm->tm_sec))
		return -1;

	tm->tm_wday = gtm->tm_wday;
	return gt;
}

/*
* Check *PT and convert it to time_t.
* If it is incompletely specified, use DEFAULT_TIME to fill it out.
* Use localtime if PT->zone is the special value TM_LOCAL_ZONE.
* Yield -1 on failure.
* ISO 8601 day-of-year and week numbers are not yet supported.
*/
	static time_t
maketime(pt, default_time)
	struct partime const *pt;
	time_t default_time;
{
	int localzone, wday;
	struct tm tm;
	struct tm *tm0 = 0;
	time_t r;

	tm0 = 0; /* Keep gcc -Wall happy.  */
	localzone = pt->zone==TM_LOCAL_ZONE;

	tm = pt->tm;

	if (TM_DEFINED(pt->ymodulus) || !TM_DEFINED(tm.tm_year)) {
	    /* Get tm corresponding to current time.  */
	    tm0 = time2tm(default_time, localzone);
	    if (!localzone)
		adjzone(tm0, pt->zone);
	}

	if (TM_DEFINED(pt->ymodulus))
	    tm.tm_year +=
		(tm0->tm_year + TM_YEAR_ORIGIN)/pt->ymodulus * pt->ymodulus;
	else if (!TM_DEFINED(tm.tm_year)) {
	    /* Set default year, month, day from current time.  */
	    tm.tm_year = tm0->tm_year + TM_YEAR_ORIGIN;
	    if (!TM_DEFINED(tm.tm_mon)) {
		tm.tm_mon = tm0->tm_mon;
		if (!TM_DEFINED(tm.tm_mday))
		    tm.tm_mday = tm0->tm_mday;
	    }
	}

	/* Convert from partime year (Gregorian) to Posix year.  */
	tm.tm_year -= TM_YEAR_ORIGIN;

	/* Set remaining default fields to be their minimum values.  */
	if (!TM_DEFINED(tm.tm_mon)) tm.tm_mon = 0;
	if (!TM_DEFINED(tm.tm_mday)) tm.tm_mday = 1;
	if (!TM_DEFINED(tm.tm_hour)) tm.tm_hour = 0;
	if (!TM_DEFINED(tm.tm_min)) tm.tm_min = 0;
	if (!TM_DEFINED(tm.tm_sec)) tm.tm_sec = 0;

	if (!localzone)
	    adjzone(&tm, -pt->zone);
	wday = tm.tm_wday;

	/* Convert and fill in the rest of the tm.  */
	r = tm2time(&tm, localzone);

	/* Check weekday.  */
	if (r != -1  &&  TM_DEFINED(wday)  &&  wday != tm.tm_wday)
		return -1;

	return r;
}

/* Parse a free-format date in SOURCE, yielding a Unix format time.  */
	static time_t
str2time(char const *source, time_t default_time, long default_zone)
{
	struct partime pt;

	if (*partime(source, &pt))
	    return -1;
	if (pt.zone == TM_UNDEFINED_ZONE)
	    pt.zone = default_zone;
	return maketime(&pt, default_time);
}

time_t time(time_t *ptr)
{
	time_t default_time = 0;
	long default_zone = -5*60*60;
	struct timeval tv;
	time_t t;
	char 	*s;

	gettimeofday(&tv, 0);
	default_time = tv.tv_sec;
	s = getenv("FAKETIME");

	if (s) {
		t = str2time(s, default_time, default_zone);
		if (t == -1)
			t = default_time;
	} else 
		t = default_time;
	if (ptr)
		*ptr = t;
	return t;
}


[-- Attachment #3: Makefile --]
[-- Type: text/plain, Size: 219 bytes --]


all: faketime.so

faketime.o: faketime.c
	cc -fpic -c -o faketime.o faketime.c

faketime.so: faketime.o
	gcc -shared faketime.o -lc -Wl,-soname -Wl,faketime.so -o faketime.so

clean:
	/bin/rm -f faketime.o faketime.so

  parent reply	other threads:[~2002-02-12  5:57 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2002-02-11 11:49 faking time SA products
2002-02-11 13:33 ` Alan Cox
2002-02-11 20:31 ` Pavel Machek
2002-02-11 21:29 ` Nerijus Baliunas
2002-02-11 23:08 ` Kilobug
2002-02-12  3:47 ` Theodore Tso [this message]
2002-02-12  8:33   ` Andreas Ferber
2002-02-12  8:56     ` mulix
2002-02-12 12:39       ` Pavel Machek
2002-02-12 21:43         ` guy keren
2002-02-12  6:00 ` David Schwartz
2002-02-12  6:09 ` Nick 'Sharkey' Moore

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20020211224723.A5514@thunk.org \
    --to=tytso@mit.edu \
    --cc=linux-kernel@vger.kernel.org \
    --cc=super.aorta@ntlworld.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox