freebsd-dev/contrib/tzcode/time2posix.3
Dag-Erling Smørgrav bc42155199 Bring our tzcode up to date.
* Replay 2010[acflm] which had been merged but not recorded.
* Merge 2010n.
* Reorganize (unsplit) the code to match the upstream layout.
* Merge 2022[cdefg].

MFC after:      1 week
Sponsored by:   Klara, Inc.
2023-01-10 16:14:27 +01:00

150 lines
3.4 KiB
Groff

.\" This file is in the public domain, so clarified as of
.\" 1996-06-05 by Arthur David Olson.
.\"
.\" $FreeBSD$
.\"
.Dd December 15, 2022
.Dt TIME2POSIX 3
.Os
.Sh NAME
.Nm time2posix ,
.Nm posix2time
.Nd convert seconds since the Epoch
.Sh LIBRARY
.Lb libc
.Sh SYNOPSIS
.In time.h
.Ft time_t
.Fn time2posix "time_t t"
.Ft time_t
.Fn posix2time "time_t t"
.Sh DESCRIPTION
.St -p1003.1-88
requires the time_t value 536457599 to stand for 1986-12-31 23:59:59 UTC.
This effectively implies that POSIX
.Vt time_t
values cannot include leap
seconds and,
therefore,
that the system time must be adjusted as each leap occurs.
.Pp
If the time package is configured with leap-second support
enabled,
however,
no such adjustment is needed and
.Vt time_t
values continue to increase over leap events
(as a true
.Dq "seconds since..."
value).
This means that these values will differ from those required by POSIX
by the net number of leap seconds inserted since the Epoch.
.Pp
Typically this is not a problem as the type
.Vt time_t
is intended
to be
(mostly)
opaque \(em
.Vt time_t
values should only be obtained-from and
passed-to functions such as
.Xr time 3 ,
.Xr localtime 3 ,
.Xr mktime 3
and
.Xr difftime 3 .
However,
.St -p1003.1-88
gives an arithmetic
expression for directly computing a
.Vt time_t
value from a given date/time,
and the same relationship is assumed by some
(usually older)
applications.
Any programs creating/dissecting
.Vt time_t
values
using such a relationship will typically not handle intervals
over leap seconds correctly.
.Pp
The
.Fn time2posix
and
.Fn posix2time
functions are provided to address this
.Vt time_t
mismatch by converting
between local
.Vt time_t
values and their POSIX equivalents.
This is done by accounting for the number of time-base changes that
would have taken place on a POSIX system as leap seconds were inserted
or deleted.
These converted values can then be used in lieu of correcting the older
applications,
or when communicating with POSIX-compliant systems.
.Pp
The
.Fn time2posix
function
is single-valued.
That is,
every local
.Vt time_t
corresponds to a single POSIX
.Vt time_t .
The
.Fn posix2time
function
is less well-behaved:
for a positive leap second hit the result is not unique,
and for a negative leap second hit the corresponding
POSIX
.Vt time_t
does not exist so an adjacent value is returned.
Both of these are good indicators of the inferiority of the
POSIX representation.
.Pp
The following table summarizes the relationship between a time
T and its conversion to,
and back from,
the POSIX representation over the leap second inserted at the end of June,
1993.
.Bl -column "93/06/30" "23:59:59" "A+0" "X=time2posix(T)"
.It Sy "DATE TIME T X=time2posix(T) posix2time(X)"
.It "93/06/30 23:59:59 A+0 B+0 A+0"
.It "93/06/30 23:59:60 A+1 B+1 A+1 or A+2"
.It "93/07/01 00:00:00 A+2 B+1 A+1 or A+2"
.It "93/07/01 00:00:01 A+3 B+2 A+3"
.El
.Pp
A leap second deletion would look like...
.Bl -column "??/06/30" "23:59:58" "A+0" "X=time2posix(T)"
.It Sy "DATE TIME T X=time2posix(T) posix2time(X)"
.It "??/06/30 23:59:58 A+0 B+0 A+0"
.It "??/07/01 00:00:00 A+1 B+2 A+1"
.It "??/07/01 00:00:01 A+2 B+3 A+2"
.El
.Pp
.D1 No "[Note: posix2time(B+1) => A+0 or A+1]"
.Pp
If leap-second support is not enabled,
local
.Vt time_t
and
POSIX
.Vt time_t
values are equivalent,
and both
.Fn time2posix
and
.Fn posix2time
degenerate to the identity function.
.Sh "SEE ALSO"
.Xr difftime 3 ,
.Xr localtime 3 ,
.Xr mktime 3 ,
.Xr time 3