* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
- * version 2 of the licence, or (at your option) any later version.
+ * version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with this library; if not, write to the
- * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
- * Boston, MA 02111-1307, USA.
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
*
* Author: Ryan Lortie <desrt@desrt.ca>
*/
#include "gbytes.h"
#include "gslice.h"
#include "gdatetime.h"
+#include "gdate.h"
+#include "genviron.h"
+
+#ifdef G_OS_WIN32
+
+#define STRICT
+#include <windows.h>
+#include <wchar.h>
+#endif
/**
* SECTION:timezone
* #GTimeZone is a structure that represents a time zone, at no
* particular point in time. It is refcounted and immutable.
*
+ * Each time zone has an identifier (for example, ‘Europe/London’) which is
+ * platform dependent. See g_time_zone_new() for information on the identifier
+ * formats. The identifier of a time zone can be retrieved using
+ * g_time_zone_get_identifier().
+ *
* A time zone contains a number of intervals. Each interval has
- * an abbreviation to describe it, an offet to UTC and a flag indicating
- * if the daylight savings time is in effect during that interval. A
- * time zone always has at least one interval -- interval 0.
+ * an abbreviation to describe it (for example, ‘PDT’), an offset to UTC and a
+ * flag indicating if the daylight savings time is in effect during that
+ * interval. A time zone always has at least one interval — interval 0. Note
+ * that interval abbreviations are not the same as time zone identifiers
+ * (apart from ‘UTC’), and cannot be passed to g_time_zone_new().
*
* Every UTC time is contained within exactly one interval, but a given
* local time may be contained within zero, one or two intervals (due to
/**
* GTimeZone:
*
- * #GDateTime is an opaque structure whose members cannot be accessed
+ * #GTimeZone is an opaque structure whose members cannot be accessed
* directly.
*
* Since: 2.26
**/
-/* zoneinfo file format {{{1 */
+/* IANA zoneinfo file format {{{1 */
/* unaligned */
typedef struct { gchar bytes[8]; } gint64_be;
guint32 tmp; memcpy (&tmp, &be, sizeof tmp); return GUINT32_FROM_BE (tmp);
}
+/* The layout of an IANA timezone file header */
struct tzhead
{
gchar tzh_magic[4];
guint8 tt_abbrind;
};
-typedef struct
-{
- gint32 gmt_offset;
- gboolean is_dst;
- gboolean is_standard;
- gboolean is_gmt;
- gchar *abbrev;
-} TransitionInfo;
-
-typedef struct
-{
- gint64 time;
- gint info_index;
-} Transition;
-
+/* A Transition Date structure for TZ Rules, an intermediate structure
+ for parsing MSWindows and Environment-variable time zones. It
+ Generalizes MSWindows's SYSTEMTIME struct.
+ */
typedef struct
{
gint year;
gint mday;
gint wday;
gint week;
- gint hour;
- gint min;
- gint sec;
- gboolean isstd;
- gboolean isgmt;
+ gint32 offset; /* hour*3600 + min*60 + sec; can be negative. */
} TimeZoneDate;
+/* POSIX Timezone abbreviations are typically 3 or 4 characters, but
+ Microsoft uses 32-character names. We'll use one larger to ensure
+ we have room for the terminating \0.
+ */
+#define NAME_SIZE 33
+
+/* A MSWindows-style time zone transition rule. Generalizes the
+ MSWindows TIME_ZONE_INFORMATION struct. Also used to compose time
+ zones from tzset-style identifiers.
+ */
typedef struct
{
- gint start_year;
+ guint start_year;
gint32 std_offset;
gint32 dlt_offset;
TimeZoneDate dlt_start;
TimeZoneDate dlt_end;
- const gchar *std_name;
- const gchar *dlt_name;
+ gchar std_name[NAME_SIZE];
+ gchar dlt_name[NAME_SIZE];
} TimeZoneRule;
+/* GTimeZone's internal representation of a Daylight Savings (Summer)
+ time interval.
+ */
+typedef struct
+{
+ gint32 gmt_offset;
+ gboolean is_dst;
+ gchar *abbrev;
+} TransitionInfo;
+
+/* GTimeZone's representation of a transition time to or from Daylight
+ Savings (Summer) time and Standard time for the zone. */
+typedef struct
+{
+ gint64 time;
+ gint info_index;
+} Transition;
-/* GTimeZone structure and lifecycle {{{1 */
+/* GTimeZone structure */
struct _GTimeZone
{
gchar *name;
- GArray *t_info;
- GArray *transitions;
+ GArray *t_info; /* Array of TransitionInfo */
+ GArray *transitions; /* Array of Transition */
gint ref_count;
};
G_LOCK_DEFINE_STATIC (time_zones);
static GHashTable/*<string?, GTimeZone>*/ *time_zones;
+G_LOCK_DEFINE_STATIC (tz_default);
+static GTimeZone *tz_default = NULL;
+G_LOCK_DEFINE_STATIC (tz_local);
+static GTimeZone *tz_local = NULL;
+
+#define MIN_TZYEAR 1916 /* Daylight Savings started in WWI */
+#define MAX_TZYEAR 2999 /* And it's not likely ever to go away, but
+ there's no point in getting carried
+ away. */
-#define MIN_TZYEAR 1900
-#define MAX_TZYEAR 2038
+#ifdef G_OS_UNIX
+static GTimeZone *parse_footertz (const gchar *, size_t);
+#endif
/**
* g_time_zone_unref:
goto again;
}
- g_hash_table_remove (time_zones, tz->name);
+ if (time_zones != NULL)
+ g_hash_table_remove (time_zones, tz->name);
G_UNLOCK(time_zones);
}
- g_array_free (tz->t_info, TRUE);
+ if (tz->t_info != NULL)
+ {
+ guint idx;
+ for (idx = 0; idx < tz->t_info->len; idx++)
+ {
+ TransitionInfo *info = &g_array_index (tz->t_info, TransitionInfo, idx);
+ g_free (info->abbrev);
+ }
+ g_array_free (tz->t_info, TRUE);
+ }
if (tz->transitions != NULL)
g_array_free (tz->transitions, TRUE);
g_free (tz->name);
/* fake zoneinfo creation (for RFC3339/ISO 8601 timezones) {{{1 */
/*
* parses strings of the form h or hh[[:]mm[[[:]ss]]] where:
- * - h[h] is 0 to 23
+ * - h[h] is 0 to 24
* - mm is 00 to 59
* - ss is 00 to 59
+ * If RFC8536, TIME_ is a transition time sans sign,
+ * so colons are required before mm and ss, and hh can be up to 167.
+ * See Internet RFC 8536 section 3.3.1:
+ * https://tools.ietf.org/html/rfc8536#section-3.3.1
+ * and POSIX Base Definitions 8.3 TZ rule time:
+ * https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap08.html#tag_08_03
*/
static gboolean
parse_time (const gchar *time_,
- gint32 *offset)
+ gint32 *offset,
+ gboolean rfc8536)
{
if (*time_ < '0' || '9' < *time_)
return FALSE;
*offset *= 10;
*offset += 60 * 60 * (*time_++ - '0');
- if (*offset > 23 * 60 * 60)
+ if (rfc8536)
+ {
+ /* Internet RFC 8536 section 3.3.1 and POSIX 8.3 TZ together say
+ that a transition time must be of the form [+-]hh[:mm[:ss]] where
+ the hours part can range from -167 to 167. */
+ if ('0' <= *time_ && *time_ <= '9')
+ {
+ *offset *= 10;
+ *offset += 60 * 60 * (*time_++ - '0');
+ }
+ if (*offset > 167 * 60 * 60)
+ return FALSE;
+ }
+ else if (*offset > 24 * 60 * 60)
return FALSE;
if (*time_ == '\0')
if (*time_ == ':')
time_++;
+ else if (rfc8536)
+ return FALSE;
if (*time_ < '0' || '5' < *time_)
return FALSE;
if (*time_ == ':')
time_++;
+ else if (rfc8536)
+ return FALSE;
if (*time_ < '0' || '5' < *time_)
return FALSE;
static gboolean
parse_constant_offset (const gchar *name,
- gint32 *offset)
+ gint32 *offset,
+ gboolean rfc8536)
{
- if (g_strcmp0 (name, "UTC") == 0)
+ /* Internet RFC 8536 section 3.3.1 and POSIX 8.3 TZ together say
+ that a transition time must be numeric. */
+ if (!rfc8536 && g_strcmp0 (name, "UTC") == 0)
{
*offset = 0;
return TRUE;
}
if (*name >= '0' && '9' >= *name)
- return parse_time (name, offset);
+ return parse_time (name, offset, rfc8536);
switch (*name++)
{
case 'Z':
*offset = 0;
- return !*name;
+ /* Internet RFC 8536 section 3.3.1 requires a numeric zone. */
+ return !rfc8536 && !*name;
case '+':
- return parse_time (name, offset);
+ return parse_time (name, offset, rfc8536);
case '-':
- if (parse_time (name, offset))
+ if (parse_time (name, offset, rfc8536))
{
*offset = -*offset;
return TRUE;
}
+ else
+ return FALSE;
default:
return FALSE;
gint32 offset;
TransitionInfo info;
- if (name == NULL || !parse_constant_offset (name, &offset))
+ if (name == NULL || !parse_constant_offset (name, &offset, FALSE))
return;
info.gmt_offset = offset;
info.is_dst = FALSE;
- info.is_standard = TRUE;
- info.is_gmt = TRUE;
info.abbrev = g_strdup (name);
-
+ gtz->name = g_strdup (name);
gtz->t_info = g_array_sized_new (FALSE, TRUE, sizeof (TransitionInfo), 1);
g_array_append_val (gtz->t_info, info);
}
#ifdef G_OS_UNIX
+
+#if defined(__sun) && defined(__SVR4)
+/*
+ * only used by Illumos distros or Solaris < 11: parse the /etc/default/init
+ * text file looking for TZ= followed by the timezone, possibly quoted
+ *
+ */
+static gchar *
+zone_identifier_illumos (void)
+{
+ gchar *resolved_identifier = NULL;
+ gchar *contents = NULL;
+ const gchar *line_start = NULL;
+ gsize tz_len = 0;
+
+ if (!g_file_get_contents ("/etc/default/init", &contents, NULL, NULL) )
+ return NULL;
+
+ /* is TZ= the first/only line in the file? */
+ if (strncmp (contents, "TZ=", 3) == 0)
+ {
+ /* found TZ= on the first line, skip over the TZ= */
+ line_start = contents + 3;
+ }
+ else
+ {
+ /* find a newline followed by TZ= */
+ line_start = strstr (contents, "\nTZ=");
+ if (line_start != NULL)
+ line_start = line_start + 4; /* skip past the \nTZ= */
+ }
+
+ /*
+ * line_start is NULL if we didn't find TZ= at the start of any line,
+ * otherwise it points to what is after the '=' (possibly '\0')
+ */
+ if (line_start == NULL || *line_start == '\0')
+ return NULL;
+
+ /* skip past a possible opening " or ' */
+ if (*line_start == '"' || *line_start == '\'')
+ line_start++;
+
+ /*
+ * loop over the next few characters, building up the length of
+ * the timezone identifier, ending with end of string, newline or
+ * a " or ' character
+ */
+ while (*(line_start + tz_len) != '\0' &&
+ *(line_start + tz_len) != '\n' &&
+ *(line_start + tz_len) != '"' &&
+ *(line_start + tz_len) != '\'')
+ tz_len++;
+
+ if (tz_len > 0)
+ {
+ /* found it */
+ resolved_identifier = g_strndup (line_start, tz_len);
+ g_strchomp (resolved_identifier);
+ g_free (contents);
+ return g_steal_pointer (&resolved_identifier);
+ }
+ else
+ return NULL;
+}
+#endif /* defined(__sun) && defined(__SRVR) */
+
+/*
+ * returns the path to the top of the Olson zoneinfo timezone hierarchy.
+ */
+static const gchar *
+zone_info_base_dir (void)
+{
+ if (g_file_test ("/usr/share/zoneinfo", G_FILE_TEST_IS_DIR))
+ return "/usr/share/zoneinfo"; /* Most distros */
+ else if (g_file_test ("/usr/share/lib/zoneinfo", G_FILE_TEST_IS_DIR))
+ return "/usr/share/lib/zoneinfo"; /* Illumos distros */
+
+ /* need a better fallback case */
+ return "/usr/share/zoneinfo";
+}
+
+static gchar *
+zone_identifier_unix (void)
+{
+ gchar *resolved_identifier = NULL;
+ gsize prefix_len = 0;
+ gchar *canonical_path = NULL;
+ GError *read_link_err = NULL;
+ const gchar *tzdir;
+
+ /* Resolve the actual timezone pointed to by /etc/localtime. */
+ resolved_identifier = g_file_read_link ("/etc/localtime", &read_link_err);
+ if (resolved_identifier == NULL)
+ {
+ gboolean not_a_symlink = g_error_matches (read_link_err,
+ G_FILE_ERROR,
+ G_FILE_ERROR_INVAL);
+ g_clear_error (&read_link_err);
+
+ /* if /etc/localtime is not a symlink, try:
+ * - /var/db/zoneinfo : 'tzsetup' program on FreeBSD and
+ * DragonflyBSD stores the timezone chosen by the user there.
+ * - /etc/timezone : Gentoo, OpenRC, and others store
+ * the user choice there.
+ * - call zone_identifier_illumos iff __sun and __SVR4 are defined,
+ * as a last-ditch effort to parse the TZ= setting from within
+ * /etc/default/init
+ */
+ if (not_a_symlink && (g_file_get_contents ("/var/db/zoneinfo",
+ &resolved_identifier,
+ NULL, NULL) ||
+ g_file_get_contents ("/etc/timezone",
+ &resolved_identifier,
+ NULL, NULL)
+#if defined(__sun) && defined(__SVR4)
+ ||
+ (resolved_identifier = zone_identifier_illumos ())
+#endif
+ ))
+ g_strchomp (resolved_identifier);
+ else
+ {
+ /* Error */
+ g_assert (resolved_identifier == NULL);
+ goto out;
+ }
+ }
+ else
+ {
+ /* Resolve relative path */
+ canonical_path = g_canonicalize_filename (resolved_identifier, "/etc");
+ g_free (resolved_identifier);
+ resolved_identifier = g_steal_pointer (&canonical_path);
+ }
+
+ tzdir = g_getenv ("TZDIR");
+ if (tzdir == NULL)
+ tzdir = zone_info_base_dir ();
+
+ /* Strip the prefix and slashes if possible. */
+ if (g_str_has_prefix (resolved_identifier, tzdir))
+ {
+ prefix_len = strlen (tzdir);
+ while (*(resolved_identifier + prefix_len) == '/')
+ prefix_len++;
+ }
+
+ if (prefix_len > 0)
+ memmove (resolved_identifier, resolved_identifier + prefix_len,
+ strlen (resolved_identifier) - prefix_len + 1 /* nul terminator */);
+
+ g_assert (resolved_identifier != NULL);
+
+out:
+ g_free (canonical_path);
+
+ return resolved_identifier;
+}
+
static GBytes*
-zone_info_unix (const gchar *identifier)
+zone_info_unix (const gchar *identifier,
+ const gchar *resolved_identifier)
{
- gchar *filename;
+ gchar *filename = NULL;
GMappedFile *file = NULL;
GBytes *zoneinfo = NULL;
+ const gchar *tzdir;
+
+ tzdir = g_getenv ("TZDIR");
+ if (tzdir == NULL)
+ tzdir = zone_info_base_dir ();
/* identifier can be a relative or absolute path name;
if relative, it is interpreted starting from /usr/share/zoneinfo
glibc allows both syntaxes, so we should too */
if (identifier != NULL)
{
- const gchar *tzdir;
-
- tzdir = getenv ("TZDIR");
- if (tzdir == NULL)
- tzdir = "/usr/share/zoneinfo";
-
if (*identifier == ':')
identifier ++;
filename = g_build_filename (tzdir, identifier, NULL);
}
else
- filename = g_strdup ("/etc/localtime");
+ {
+ if (resolved_identifier == NULL)
+ goto out;
+
+ filename = g_strdup ("/etc/localtime");
+ }
file = g_mapped_file_new (filename, FALSE, NULL);
if (file != NULL)
g_mapped_file_ref (file));
g_mapped_file_unref (file);
}
+
+ g_assert (resolved_identifier != NULL);
+
+out:
g_free (filename);
+
return zoneinfo;
}
static void
-init_zone_from_iana_info (GTimeZone *gtz, GBytes *zoneinfo)
+init_zone_from_iana_info (GTimeZone *gtz,
+ GBytes *zoneinfo,
+ gchar *identifier /* (transfer full) */)
{
gsize size;
guint index;
- guint32 time_count, type_count, leap_count, isgmt_count;
- guint32 isstd_count, char_count ;
- gpointer tz_transitions, tz_type_index, tz_ttinfo;
- gpointer tz_leaps, tz_isgmt, tz_isstd;
- gchar* tz_abbrs;
- guint timesize = sizeof (gint32), countsize = sizeof (gint32);
- const struct tzhead *header = g_bytes_get_data (zoneinfo, &size);
+ guint32 time_count, type_count;
+ guint8 *tz_transitions, *tz_type_index, *tz_ttinfo;
+ guint8 *tz_abbrs;
+ gsize timesize = sizeof (gint32);
+ gconstpointer header_data = g_bytes_get_data (zoneinfo, &size);
+ const gchar *data = header_data;
+ const struct tzhead *header = header_data;
+ GTimeZone *footertz = NULL;
+ guint extra_time_count = 0, extra_type_count = 0;
+ gint64 last_explicit_transition_time;
g_return_if_fail (size >= sizeof (struct tzhead) &&
memcmp (header, "TZif", 4) == 0);
- if (header->tzh_version == '2')
+ /* FIXME: Handle invalid TZif files better (Issue#1088). */
+
+ if (header->tzh_version >= '2')
{
/* Skip ahead to the newer 64-bit data if it's available. */
header = (const struct tzhead *)
}
time_count = guint32_from_be(header->tzh_timecnt);
type_count = guint32_from_be(header->tzh_typecnt);
- leap_count = guint32_from_be(header->tzh_leapcnt);
- isgmt_count = guint32_from_be(header->tzh_ttisgmtcnt);
- isstd_count = guint32_from_be(header->tzh_ttisstdcnt);
- char_count = guint32_from_be(header->tzh_charcnt);
- g_assert (type_count == isgmt_count);
- g_assert (type_count == isstd_count);
+ if (header->tzh_version >= '2')
+ {
+ const gchar *footer = (((const gchar *) (header + 1))
+ + guint32_from_be(header->tzh_ttisgmtcnt)
+ + guint32_from_be(header->tzh_ttisstdcnt)
+ + 12 * guint32_from_be(header->tzh_leapcnt)
+ + 9 * time_count
+ + 6 * type_count
+ + guint32_from_be(header->tzh_charcnt));
+ const gchar *footerlast;
+ size_t footerlen;
+ g_return_if_fail (footer <= data + size - 2 && footer[0] == '\n');
+ footerlast = memchr (footer + 1, '\n', data + size - (footer + 1));
+ g_return_if_fail (footerlast);
+ footerlen = footerlast + 1 - footer;
+ if (footerlen != 2)
+ {
+ footertz = parse_footertz (footer, footerlen);
+ g_return_if_fail (footertz);
+ extra_type_count = footertz->t_info->len;
+ extra_time_count = footertz->transitions->len;
+ }
+ }
- tz_transitions = (gpointer)(header + 1);
+ tz_transitions = ((guint8 *) (header) + sizeof (*header));
tz_type_index = tz_transitions + timesize * time_count;
tz_ttinfo = tz_type_index + time_count;
tz_abbrs = tz_ttinfo + sizeof (struct ttinfo) * type_count;
- tz_leaps = tz_abbrs + char_count;
- tz_isstd = tz_leaps + (timesize + countsize) * leap_count;
- tz_isgmt = tz_isstd + isstd_count;
+ gtz->name = g_steal_pointer (&identifier);
gtz->t_info = g_array_sized_new (FALSE, TRUE, sizeof (TransitionInfo),
- type_count);
+ type_count + extra_type_count);
gtz->transitions = g_array_sized_new (FALSE, TRUE, sizeof (Transition),
- time_count);
+ time_count + extra_time_count);
for (index = 0; index < type_count; index++)
{
struct ttinfo info = ((struct ttinfo*)tz_ttinfo)[index];
t_info.gmt_offset = gint32_from_be (info.tt_gmtoff);
t_info.is_dst = info.tt_isdst ? TRUE : FALSE;
- t_info.is_standard = ((guint8*)tz_isstd)[index] ? TRUE : FALSE;
- t_info.is_gmt = ((guint8*)tz_isgmt)[index] ? TRUE : FALSE;
- t_info.abbrev = g_strdup (&tz_abbrs[info.tt_abbrind]);
+ t_info.abbrev = g_strdup ((gchar *) &tz_abbrs[info.tt_abbrind]);
g_array_append_val (gtz->t_info, t_info);
}
for (index = 0; index < time_count; index++)
{
Transition trans;
- if (header->tzh_version == '2')
+ if (header->tzh_version >= '2')
trans.time = gint64_from_be (((gint64_be*)tz_transitions)[index]);
else
trans.time = gint32_from_be (((gint32_be*)tz_transitions)[index]);
- trans.info_index = ((guint8*)tz_type_index)[index];
+ last_explicit_transition_time = trans.time;
+ trans.info_index = tz_type_index[index];
g_assert (trans.info_index >= 0);
- g_assert (trans.info_index < gtz->t_info->len);
+ g_assert ((guint) trans.info_index < gtz->t_info->len);
g_array_append_val (gtz->transitions, trans);
}
- g_bytes_unref (zoneinfo);
+
+ if (footertz)
+ {
+ /* Append footer time types. Don't bother to coalesce
+ duplicates with existing time types. */
+ for (index = 0; index < extra_type_count; index++)
+ {
+ TransitionInfo t_info;
+ TransitionInfo *footer_t_info
+ = &g_array_index (footertz->t_info, TransitionInfo, index);
+ t_info.gmt_offset = footer_t_info->gmt_offset;
+ t_info.is_dst = footer_t_info->is_dst;
+ t_info.abbrev = g_steal_pointer (&footer_t_info->abbrev);
+ g_array_append_val (gtz->t_info, t_info);
+ }
+
+ /* Append footer transitions that follow the last explicit
+ transition. */
+ for (index = 0; index < extra_time_count; index++)
+ {
+ Transition *footer_transition
+ = &g_array_index (footertz->transitions, Transition, index);
+ if (time_count <= 0
+ || last_explicit_transition_time < footer_transition->time)
+ {
+ Transition trans;
+ trans.time = footer_transition->time;
+ trans.info_index = type_count + footer_transition->info_index;
+ g_array_append_val (gtz->transitions, trans);
+ }
+ }
+
+ g_time_zone_unref (footertz);
+ }
}
-#endif
+#elif defined (G_OS_WIN32)
static void
-find_relative_date (TimeZoneDate *buffer,
- GTimeZone *tz)
+copy_windows_systemtime (SYSTEMTIME *s_time, TimeZoneDate *tzdate)
+{
+ tzdate->offset
+ = s_time->wHour * 3600 + s_time->wMinute * 60 + s_time->wSecond;
+ tzdate->mon = s_time->wMonth;
+ tzdate->year = s_time->wYear;
+ tzdate->wday = s_time->wDayOfWeek ? s_time->wDayOfWeek : 7;
+
+ if (s_time->wYear)
+ {
+ tzdate->mday = s_time->wDay;
+ tzdate->wday = 0;
+ }
+ else
+ tzdate->week = s_time->wDay;
+}
+
+/* UTC = local time + bias while local time = UTC + offset */
+static gboolean
+rule_from_windows_time_zone_info (TimeZoneRule *rule,
+ TIME_ZONE_INFORMATION *tzi)
{
- GDateTime *dt;
- gint wday;
+ gchar *std_name, *dlt_name;
+
+ std_name = g_utf16_to_utf8 ((gunichar2 *)tzi->StandardName, -1, NULL, NULL, NULL);
+ if (std_name == NULL)
+ return FALSE;
+
+ dlt_name = g_utf16_to_utf8 ((gunichar2 *)tzi->DaylightName, -1, NULL, NULL, NULL);
+ if (dlt_name == NULL)
+ {
+ g_free (std_name);
+ return FALSE;
+ }
+
+ /* Set offset */
+ if (tzi->StandardDate.wMonth)
+ {
+ rule->std_offset = -(tzi->Bias + tzi->StandardBias) * 60;
+ rule->dlt_offset = -(tzi->Bias + tzi->DaylightBias) * 60;
+ copy_windows_systemtime (&(tzi->DaylightDate), &(rule->dlt_start));
+
+ copy_windows_systemtime (&(tzi->StandardDate), &(rule->dlt_end));
+ }
+
+ else
+ {
+ rule->std_offset = -tzi->Bias * 60;
+ rule->dlt_start.mon = 0;
+ }
+ strncpy (rule->std_name, std_name, NAME_SIZE - 1);
+ strncpy (rule->dlt_name, dlt_name, NAME_SIZE - 1);
+
+ g_free (std_name);
+ g_free (dlt_name);
+
+ return TRUE;
+}
+
+static gchar*
+windows_default_tzname (void)
+{
+ const gunichar2 *subkey =
+ L"SYSTEM\\CurrentControlSet\\Control\\TimeZoneInformation";
+ HKEY key;
+ gchar *key_name = NULL;
+ gunichar2 *key_name_w = NULL;
+ if (RegOpenKeyExW (HKEY_LOCAL_MACHINE, subkey, 0,
+ KEY_QUERY_VALUE, &key) == ERROR_SUCCESS)
+ {
+ DWORD size = 0;
+ if (RegQueryValueExW (key, L"TimeZoneKeyName", NULL, NULL,
+ NULL, &size) == ERROR_SUCCESS)
+ {
+ key_name_w = g_malloc ((gint)size);
+
+ if (key_name_w == NULL ||
+ RegQueryValueExW (key, L"TimeZoneKeyName", NULL, NULL,
+ (LPBYTE)key_name_w, &size) != ERROR_SUCCESS)
+ {
+ g_free (key_name_w);
+ key_name = NULL;
+ }
+ else
+ key_name = g_utf16_to_utf8 (key_name_w, -1, NULL, NULL, NULL);
+ }
+ RegCloseKey (key);
+ }
+ return key_name;
+}
+
+typedef struct
+{
+ LONG Bias;
+ LONG StandardBias;
+ LONG DaylightBias;
+ SYSTEMTIME StandardDate;
+ SYSTEMTIME DaylightDate;
+} RegTZI;
+
+static void
+system_time_copy (SYSTEMTIME *orig, SYSTEMTIME *target)
+{
+ g_return_if_fail (orig != NULL);
+ g_return_if_fail (target != NULL);
+
+ target->wYear = orig->wYear;
+ target->wMonth = orig->wMonth;
+ target->wDayOfWeek = orig->wDayOfWeek;
+ target->wDay = orig->wDay;
+ target->wHour = orig->wHour;
+ target->wMinute = orig->wMinute;
+ target->wSecond = orig->wSecond;
+ target->wMilliseconds = orig->wMilliseconds;
+}
+
+static void
+register_tzi_to_tzi (RegTZI *reg, TIME_ZONE_INFORMATION *tzi)
+{
+ g_return_if_fail (reg != NULL);
+ g_return_if_fail (tzi != NULL);
+ tzi->Bias = reg->Bias;
+ system_time_copy (&(reg->StandardDate), &(tzi->StandardDate));
+ tzi->StandardBias = reg->StandardBias;
+ system_time_copy (&(reg->DaylightDate), &(tzi->DaylightDate));
+ tzi->DaylightBias = reg->DaylightBias;
+}
+
+static guint
+rules_from_windows_time_zone (const gchar *identifier,
+ const gchar *resolved_identifier,
+ TimeZoneRule **rules)
+{
+ HKEY key;
+ gchar *subkey = NULL;
+ gchar *subkey_dynamic = NULL;
+ const gchar *key_name;
+ const gchar *reg_key =
+ "SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\Time Zones\\";
+ TIME_ZONE_INFORMATION tzi;
+ DWORD size;
+ guint rules_num = 0;
+ RegTZI regtzi = { 0 }, regtzi_prev;
+ WCHAR winsyspath[MAX_PATH];
+ gunichar2 *subkey_w, *subkey_dynamic_w;
+
+ subkey_dynamic_w = NULL;
+
+ if (GetSystemDirectoryW (winsyspath, MAX_PATH) == 0)
+ return 0;
+
+ g_assert (rules != NULL);
+
+ *rules = NULL;
+ key_name = NULL;
+
+ if (!identifier)
+ key_name = resolved_identifier;
+ else
+ key_name = identifier;
+
+ if (!key_name)
+ return 0;
+
+ subkey = g_strconcat (reg_key, key_name, NULL);
+ subkey_w = g_utf8_to_utf16 (subkey, -1, NULL, NULL, NULL);
+ if (subkey_w == NULL)
+ goto utf16_conv_failed;
+ subkey_dynamic = g_strconcat (subkey, "\\Dynamic DST", NULL);
+ subkey_dynamic_w = g_utf8_to_utf16 (subkey_dynamic, -1, NULL, NULL, NULL);
+ if (subkey_dynamic_w == NULL)
+ goto utf16_conv_failed;
+
+ if (RegOpenKeyExW (HKEY_LOCAL_MACHINE, subkey_w, 0,
+ KEY_QUERY_VALUE, &key) != ERROR_SUCCESS)
+ goto utf16_conv_failed;
+
+ size = sizeof tzi.StandardName;
+
+ /* use RegLoadMUIStringW() to query MUI_Std from the registry if possible, otherwise
+ fallback to querying Std */
+ if (RegLoadMUIStringW (key, L"MUI_Std", tzi.StandardName,
+ size, &size, 0, winsyspath) != ERROR_SUCCESS)
+ {
+ size = sizeof tzi.StandardName;
+ if (RegQueryValueExW (key, L"Std", NULL, NULL,
+ (LPBYTE)&(tzi.StandardName), &size) != ERROR_SUCCESS)
+ goto registry_failed;
+ }
+
+ size = sizeof tzi.DaylightName;
+
+ /* use RegLoadMUIStringW() to query MUI_Dlt from the registry if possible, otherwise
+ fallback to querying Dlt */
+ if (RegLoadMUIStringW (key, L"MUI_Dlt", tzi.DaylightName,
+ size, &size, 0, winsyspath) != ERROR_SUCCESS)
+ {
+ size = sizeof tzi.DaylightName;
+ if (RegQueryValueExW (key, L"Dlt", NULL, NULL,
+ (LPBYTE)&(tzi.DaylightName), &size) != ERROR_SUCCESS)
+ goto registry_failed;
+ }
+
+ RegCloseKey (key);
+ if (RegOpenKeyExW (HKEY_LOCAL_MACHINE, subkey_dynamic_w, 0,
+ KEY_QUERY_VALUE, &key) == ERROR_SUCCESS)
+ {
+ DWORD i, first, last, year;
+ wchar_t s[12];
+
+ size = sizeof first;
+ if (RegQueryValueExW (key, L"FirstEntry", NULL, NULL,
+ (LPBYTE) &first, &size) != ERROR_SUCCESS)
+ goto registry_failed;
+
+ size = sizeof last;
+ if (RegQueryValueExW (key, L"LastEntry", NULL, NULL,
+ (LPBYTE) &last, &size) != ERROR_SUCCESS)
+ goto registry_failed;
+
+ rules_num = last - first + 2;
+ *rules = g_new0 (TimeZoneRule, rules_num);
+
+ for (year = first, i = 0; *rules != NULL && year <= last; year++)
+ {
+ gboolean failed = FALSE;
+ swprintf_s (s, 11, L"%d", year);
+
+ if (!failed)
+ {
+ size = sizeof regtzi;
+ if (RegQueryValueExW (key, s, NULL, NULL,
+ (LPBYTE) ®tzi, &size) != ERROR_SUCCESS)
+ failed = TRUE;
+ }
+
+ if (failed)
+ {
+ g_free (*rules);
+ *rules = NULL;
+ break;
+ }
+
+ if (year > first && memcmp (®tzi_prev, ®tzi, sizeof regtzi) == 0)
+ continue;
+ else
+ memcpy (®tzi_prev, ®tzi, sizeof regtzi);
+
+ register_tzi_to_tzi (®tzi, &tzi);
+
+ if (!rule_from_windows_time_zone_info (&(*rules)[i], &tzi))
+ {
+ g_free (*rules);
+ *rules = NULL;
+ break;
+ }
+
+ (*rules)[i++].start_year = year;
+ }
+
+ rules_num = i + 1;
+
+registry_failed:
+ RegCloseKey (key);
+ }
+ else if (RegOpenKeyExW (HKEY_LOCAL_MACHINE, subkey_w, 0,
+ KEY_QUERY_VALUE, &key) == ERROR_SUCCESS)
+ {
+ size = sizeof regtzi;
+ if (RegQueryValueExW (key, L"TZI", NULL, NULL,
+ (LPBYTE) ®tzi, &size) == ERROR_SUCCESS)
+ {
+ rules_num = 2;
+ *rules = g_new0 (TimeZoneRule, 2);
+ register_tzi_to_tzi (®tzi, &tzi);
+
+ if (!rule_from_windows_time_zone_info (&(*rules)[0], &tzi))
+ {
+ g_free (*rules);
+ *rules = NULL;
+ }
+ }
+
+ RegCloseKey (key);
+ }
+
+utf16_conv_failed:
+ g_free (subkey_dynamic_w);
+ g_free (subkey_dynamic);
+ g_free (subkey_w);
+ g_free (subkey);
+
+ if (*rules)
+ {
+ (*rules)[0].start_year = MIN_TZYEAR;
+ if ((*rules)[rules_num - 2].start_year < MAX_TZYEAR)
+ (*rules)[rules_num - 1].start_year = MAX_TZYEAR;
+ else
+ (*rules)[rules_num - 1].start_year = (*rules)[rules_num - 2].start_year + 1;
+
+ return rules_num;
+ }
+
+ return 0;
+}
+
+#endif
+
+static void
+find_relative_date (TimeZoneDate *buffer)
+{
+ guint wday;
+ GDate date;
+ g_date_clear (&date, 1);
wday = buffer->wday;
/* Get last day if last is needed, first day otherwise */
- dt = g_date_time_new (tz,
- buffer->year,
- buffer->mon + (buffer->week < 5? 0 : 1),
- buffer->week < 5? 1 : 0,
- buffer->hour, buffer->min, buffer->sec);
+ if (buffer->mon == 13 || buffer->mon == 14) /* Julian Date */
+ {
+ g_date_set_dmy (&date, 1, 1, buffer->year);
+ if (wday >= 59 && buffer->mon == 13 && g_date_is_leap_year (buffer->year))
+ g_date_add_days (&date, wday);
+ else
+ g_date_add_days (&date, wday - 1);
+ buffer->mon = (int) g_date_get_month (&date);
+ buffer->mday = (int) g_date_get_day (&date);
+ buffer->wday = 0;
+ }
+ else /* M.W.D */
+ {
+ guint days;
+ guint days_in_month = g_date_get_days_in_month (buffer->mon, buffer->year);
+ GDateWeekday first_wday;
+
+ g_date_set_dmy (&date, 1, buffer->mon, buffer->year);
+ first_wday = g_date_get_weekday (&date);
+
+ if ((guint) first_wday > wday)
+ ++(buffer->week);
+ /* week is 1 <= w <= 5, we need 0-based */
+ days = 7 * (buffer->week - 1) + wday - first_wday;
+
+ /* "days" is a 0-based offset from the 1st of the month.
+ * Adding days == days_in_month would bring us into the next month,
+ * hence the ">=" instead of just ">".
+ */
+ while (days >= days_in_month)
+ days -= 7;
+
+ g_date_add_days (&date, days);
+
+ buffer->mday = g_date_get_day (&date);
+ }
+}
+
+/* Offset is previous offset of local time. Returns 0 if month is 0 */
+static gint64
+boundary_for_year (TimeZoneDate *boundary,
+ gint year,
+ gint32 offset)
+{
+ TimeZoneDate buffer;
+ GDate date;
+ const guint64 unix_epoch_start = 719163L;
+ const guint64 seconds_per_day = 86400L;
+
+ if (!boundary->mon)
+ return 0;
+ buffer = *boundary;
+
+ if (boundary->year == 0)
+ {
+ buffer.year = year;
+
+ if (buffer.wday)
+ find_relative_date (&buffer);
+ }
+
+ g_assert (buffer.year == year);
+ g_date_clear (&date, 1);
+ g_date_set_dmy (&date, buffer.mday, buffer.mon, buffer.year);
+ return ((g_date_get_julian (&date) - unix_epoch_start) * seconds_per_day +
+ buffer.offset - offset);
+}
+
+static void
+fill_transition_info_from_rule (TransitionInfo *info,
+ TimeZoneRule *rule,
+ gboolean is_dst)
+{
+ gint offset = is_dst ? rule->dlt_offset : rule->std_offset;
+ gchar *name = is_dst ? rule->dlt_name : rule->std_name;
+
+ info->gmt_offset = offset;
+ info->is_dst = is_dst;
- buffer->wday = g_date_time_get_day_of_week (dt);
- buffer->mday = g_date_time_get_day_of_month (dt);
+ if (name)
+ info->abbrev = g_strdup (name);
+
+ else
+ info->abbrev = g_strdup_printf ("%+03d%02d",
+ (int) offset / 3600,
+ (int) abs (offset / 60) % 60);
+}
+
+static void
+init_zone_from_rules (GTimeZone *gtz,
+ TimeZoneRule *rules,
+ guint rules_num,
+ gchar *identifier /* (transfer full) */)
+{
+ guint type_count = 0, trans_count = 0, info_index = 0;
+ guint ri; /* rule index */
+ gboolean skip_first_std_trans = TRUE;
+ gint32 last_offset;
+
+ type_count = 0;
+ trans_count = 0;
+
+ /* Last rule only contains max year */
+ for (ri = 0; ri < rules_num - 1; ri++)
+ {
+ if (rules[ri].dlt_start.mon || rules[ri].dlt_end.mon)
+ {
+ guint rulespan = (rules[ri + 1].start_year - rules[ri].start_year);
+ guint transitions = rules[ri].dlt_start.mon > 0 ? 1 : 0;
+ transitions += rules[ri].dlt_end.mon > 0 ? 1 : 0;
+ type_count += rules[ri].dlt_start.mon > 0 ? 2 : 1;
+ trans_count += transitions * rulespan;
+ }
+ else
+ type_count++;
+ }
+
+ gtz->name = g_steal_pointer (&identifier);
+ gtz->t_info = g_array_sized_new (FALSE, TRUE, sizeof (TransitionInfo), type_count);
+ gtz->transitions = g_array_sized_new (FALSE, TRUE, sizeof (Transition), trans_count);
+
+ last_offset = rules[0].std_offset;
+
+ for (ri = 0; ri < rules_num - 1; ri++)
+ {
+ if ((rules[ri].std_offset || rules[ri].dlt_offset) &&
+ rules[ri].dlt_start.mon == 0 && rules[ri].dlt_end.mon == 0)
+ {
+ TransitionInfo std_info;
+ /* Standard */
+ fill_transition_info_from_rule (&std_info, &(rules[ri]), FALSE);
+ g_array_append_val (gtz->t_info, std_info);
+
+ if (ri > 0 &&
+ ((rules[ri - 1].dlt_start.mon > 12 &&
+ rules[ri - 1].dlt_start.wday > rules[ri - 1].dlt_end.wday) ||
+ rules[ri - 1].dlt_start.mon > rules[ri - 1].dlt_end.mon))
+ {
+ /* The previous rule was a southern hemisphere rule that
+ starts the year with DST, so we need to add a
+ transition to return to standard time */
+ guint year = rules[ri].start_year;
+ gint64 std_time = boundary_for_year (&rules[ri].dlt_end,
+ year, last_offset);
+ Transition std_trans = {std_time, info_index};
+ g_array_append_val (gtz->transitions, std_trans);
+
+ }
+ last_offset = rules[ri].std_offset;
+ ++info_index;
+ skip_first_std_trans = TRUE;
+ }
+ else
+ {
+ const guint start_year = rules[ri].start_year;
+ const guint end_year = rules[ri + 1].start_year;
+ gboolean dlt_first;
+ guint year;
+ TransitionInfo std_info, dlt_info;
+ if (rules[ri].dlt_start.mon > 12)
+ dlt_first = rules[ri].dlt_start.wday > rules[ri].dlt_end.wday;
+ else
+ dlt_first = rules[ri].dlt_start.mon > rules[ri].dlt_end.mon;
+ /* Standard rules are always even, because before the first
+ transition is always standard time, and 0 is even. */
+ fill_transition_info_from_rule (&std_info, &(rules[ri]), FALSE);
+ fill_transition_info_from_rule (&dlt_info, &(rules[ri]), TRUE);
+
+ g_array_append_val (gtz->t_info, std_info);
+ g_array_append_val (gtz->t_info, dlt_info);
+
+ /* Transition dates. We hope that a year which ends daylight
+ time in a southern-hemisphere country (i.e., one that
+ begins the year in daylight time) will include a rule
+ which has only a dlt_end. */
+ for (year = start_year; year < end_year; year++)
+ {
+ gint32 dlt_offset = (dlt_first ? last_offset :
+ rules[ri].dlt_offset);
+ gint32 std_offset = (dlt_first ? rules[ri].std_offset :
+ last_offset);
+ /* NB: boundary_for_year returns 0 if mon == 0 */
+ gint64 std_time = boundary_for_year (&rules[ri].dlt_end,
+ year, dlt_offset);
+ gint64 dlt_time = boundary_for_year (&rules[ri].dlt_start,
+ year, std_offset);
+ Transition std_trans = {std_time, info_index};
+ Transition dlt_trans = {dlt_time, info_index + 1};
+ last_offset = (dlt_first ? rules[ri].dlt_offset :
+ rules[ri].std_offset);
+ if (dlt_first)
+ {
+ if (skip_first_std_trans)
+ skip_first_std_trans = FALSE;
+ else if (std_time)
+ g_array_append_val (gtz->transitions, std_trans);
+ if (dlt_time)
+ g_array_append_val (gtz->transitions, dlt_trans);
+ }
+ else
+ {
+ if (dlt_time)
+ g_array_append_val (gtz->transitions, dlt_trans);
+ if (std_time)
+ g_array_append_val (gtz->transitions, std_trans);
+ }
+ }
+
+ info_index += 2;
+ }
+ }
+ if (ri > 0 &&
+ ((rules[ri - 1].dlt_start.mon > 12 &&
+ rules[ri - 1].dlt_start.wday > rules[ri - 1].dlt_end.wday) ||
+ rules[ri - 1].dlt_start.mon > rules[ri - 1].dlt_end.mon))
+ {
+ /* The previous rule was a southern hemisphere rule that
+ starts the year with DST, so we need to add a
+ transition to return to standard time */
+ TransitionInfo info;
+ guint year = rules[ri].start_year;
+ Transition trans;
+ fill_transition_info_from_rule (&info, &(rules[ri - 1]), FALSE);
+ g_array_append_val (gtz->t_info, info);
+ trans.time = boundary_for_year (&rules[ri - 1].dlt_end,
+ year, last_offset);
+ trans.info_index = info_index;
+ g_array_append_val (gtz->transitions, trans);
+ }
+}
+
+/*
+ * parses date[/time] for parsing TZ environment variable
+ *
+ * date is either Mm.w.d, Jn or N
+ * - m is 1 to 12
+ * - w is 1 to 5
+ * - d is 0 to 6
+ * - n is 1 to 365
+ * - N is 0 to 365
+ *
+ * time is either h or hh[[:]mm[[[:]ss]]]
+ * - h[h] is 0 to 24
+ * - mm is 00 to 59
+ * - ss is 00 to 59
+ */
+static gboolean
+parse_mwd_boundary (gchar **pos, TimeZoneDate *boundary)
+{
+ gint month, week, day;
+
+ if (**pos == '\0' || **pos < '0' || '9' < **pos)
+ return FALSE;
+
+ month = *(*pos)++ - '0';
+
+ if ((month == 1 && **pos >= '0' && '2' >= **pos) ||
+ (month == 0 && **pos >= '0' && '9' >= **pos))
+ {
+ month *= 10;
+ month += *(*pos)++ - '0';
+ }
+
+ if (*(*pos)++ != '.' || month == 0)
+ return FALSE;
+
+ if (**pos == '\0' || **pos < '1' || '5' < **pos)
+ return FALSE;
+
+ week = *(*pos)++ - '0';
+
+ if (*(*pos)++ != '.')
+ return FALSE;
+
+ if (**pos == '\0' || **pos < '0' || '6' < **pos)
+ return FALSE;
+
+ day = *(*pos)++ - '0';
+
+ if (!day)
+ day += 7;
+
+ boundary->year = 0;
+ boundary->mon = month;
+ boundary->week = week;
+ boundary->wday = day;
+ return TRUE;
+}
+
+/*
+ * This parses two slightly different ways of specifying
+ * the Julian day:
+ *
+ * - ignore_leap == TRUE
+ *
+ * Jn This specifies the Julian day with n between 1 and 365. Leap days
+ * are not counted. In this format, February 29 can't be represented;
+ * February 28 is day 59, and March 1 is always day 60.
+ *
+ * - ignore_leap == FALSE
+ *
+ * n This specifies the zero-based Julian day with n between 0 and 365.
+ * February 29 is counted in leap years.
+ */
+static gboolean
+parse_julian_boundary (gchar** pos, TimeZoneDate *boundary,
+ gboolean ignore_leap)
+{
+ gint day = 0;
+ GDate date;
- if (buffer->week < 5)
+ while (**pos >= '0' && '9' >= **pos)
{
- if (wday < buffer->wday)
- buffer->wday -= 7;
-
- buffer->mday += (buffer->week - 1) * 7;
+ day *= 10;
+ day += *(*pos)++ - '0';
}
- else if (wday > buffer->wday)
- buffer->wday += 7;
+ if (ignore_leap)
+ {
+ if (day < 1 || 365 < day)
+ return FALSE;
+ if (day >= 59)
+ day++;
+ }
+ else
+ {
+ if (day < 0 || 365 < day)
+ return FALSE;
+ /* GDate wants day in range 1->366 */
+ day++;
+ }
- buffer->mday += wday - buffer->wday;
- buffer->wday = wday;
+ g_date_clear (&date, 1);
+ g_date_set_julian (&date, day);
+ boundary->year = 0;
+ boundary->mon = (int) g_date_get_month (&date);
+ boundary->mday = (int) g_date_get_day (&date);
+ boundary->wday = 0;
- g_date_time_unref (dt);
+ return TRUE;
}
-/* Offset is previous offset of local time */
-static gint64
-boundary_for_year (TimeZoneDate *boundary,
- gint year,
- gint32 prev_offset,
- gint32 std_offset)
+static gboolean
+parse_tz_boundary (const gchar *identifier,
+ TimeZoneDate *boundary)
{
- TimeZoneDate buffer;
- GDateTime *dt;
- GTimeZone *tz;
- gint64 t;
- gint32 offset;
- gchar *identifier;
-
- buffer = *boundary;
+ gchar *pos;
- if (boundary->isgmt)
- offset = 0;
- else if (boundary->isstd)
- offset = std_offset;
+ pos = (gchar*)identifier;
+ /* Month-week-weekday */
+ if (*pos == 'M')
+ {
+ ++pos;
+ if (!parse_mwd_boundary (&pos, boundary))
+ return FALSE;
+ }
+ /* Julian date which ignores Feb 29 in leap years */
+ else if (*pos == 'J')
+ {
+ ++pos;
+ if (!parse_julian_boundary (&pos, boundary, TRUE))
+ return FALSE ;
+ }
+ /* Julian date which counts Feb 29 in leap years */
+ else if (*pos >= '0' && '9' >= *pos)
+ {
+ if (!parse_julian_boundary (&pos, boundary, FALSE))
+ return FALSE;
+ }
else
- offset = prev_offset;
-
- G_UNLOCK (time_zones);
+ return FALSE;
- identifier = g_strdup_printf ("%+03d:%02d:%02d",
- (int) offset / 3600,
- (int) abs (offset / 60) % 60,
- (int) abs (offset) % 3600);
- tz = g_time_zone_new (identifier);
- g_free (identifier);
+ /* Time */
- if (boundary->year == 0)
+ if (*pos == '/')
+ return parse_constant_offset (pos + 1, &boundary->offset, TRUE);
+ else
{
- buffer.year = year;
-
- if (buffer.wday)
- find_relative_date (&buffer, tz);
+ boundary->offset = 2 * 60 * 60;
+ return *pos == '\0';
}
+}
- g_assert (buffer.year == year);
+static guint
+create_ruleset_from_rule (TimeZoneRule **rules, TimeZoneRule *rule)
+{
+ *rules = g_new0 (TimeZoneRule, 2);
+
+ (*rules)[0].start_year = MIN_TZYEAR;
+ (*rules)[1].start_year = MAX_TZYEAR;
+
+ (*rules)[0].std_offset = -rule->std_offset;
+ (*rules)[0].dlt_offset = -rule->dlt_offset;
+ (*rules)[0].dlt_start = rule->dlt_start;
+ (*rules)[0].dlt_end = rule->dlt_end;
+ strcpy ((*rules)[0].std_name, rule->std_name);
+ strcpy ((*rules)[0].dlt_name, rule->dlt_name);
+ return 2;
+}
- dt = g_date_time_new (tz,
- buffer.year, buffer.mon, buffer.mday,
- buffer.hour, buffer.min, buffer.sec);
- t = g_date_time_to_unix (dt);
- g_date_time_unref (dt);
+static gboolean
+parse_offset (gchar **pos, gint32 *target)
+{
+ gchar *buffer;
+ gchar *target_pos = *pos;
+ gboolean ret;
- g_time_zone_unref (tz);
+ while (**pos == '+' || **pos == '-' || **pos == ':' ||
+ (**pos >= '0' && '9' >= **pos))
+ ++(*pos);
- G_LOCK (time_zones);
+ buffer = g_strndup (target_pos, *pos - target_pos);
+ ret = parse_constant_offset (buffer, target, FALSE);
+ g_free (buffer);
- return t;
+ return ret;
}
-static void
-init_zone_from_rules (GTimeZone *gtz,
- TimeZoneRule *rules,
- gint rules_num)
+static gboolean
+parse_identifier_boundary (gchar **pos, TimeZoneDate *target)
{
- TransitionInfo info[2];
- Transition trans;
- gint type_count, trans_count;
- gint year, i, x, y;
- gint32 last_offset;
+ gchar *buffer;
+ gchar *target_pos = *pos;
+ gboolean ret;
- type_count = 0;
- trans_count = 0;
+ while (**pos != ',' && **pos != '\0')
+ ++(*pos);
+ buffer = g_strndup (target_pos, *pos - target_pos);
+ ret = parse_tz_boundary (buffer, target);
+ g_free (buffer);
- /* Last rule only contains max year */
- for (i = 0; i < rules_num - 1; i++)
- {
- if (rules[i].dlt_start.mon)
- {
- type_count += 2;
- trans_count += 2 * (rules[i+1].start_year - rules[i].start_year);
- }
- else
- type_count++;
- }
+ return ret;
+}
- x = 0;
- y = 0;
+static gboolean
+set_tz_name (gchar **pos, gchar *buffer, guint size)
+{
+ gboolean quoted = **pos == '<';
+ gchar *name_pos = *pos;
+ guint len;
- /* If standard time happens before daylight time in first rule
- * with daylight, skip first transition so the minimum is in
- * standard time and the first transition is in daylight time */
- for (i = 0; i < rules_num - 1 && rules[0].dlt_start.mon == 0; i++);
+ g_assert (size != 0);
- if (i < rules_num -1 && rules[i].dlt_start.mon > 0 &&
- rules[i].dlt_start.mon > rules[i].dlt_end.mon)
+ if (quoted)
{
- trans_count--;
- x = -1;
+ name_pos++;
+ do
+ ++(*pos);
+ while (g_ascii_isalnum (**pos) || **pos == '-' || **pos == '+');
+ if (**pos != '>')
+ return FALSE;
}
+ else
+ while (g_ascii_isalpha (**pos))
+ ++(*pos);
- gtz->t_info = g_array_sized_new (FALSE, TRUE, sizeof (TransitionInfo), type_count);
- gtz->transitions = g_array_sized_new (FALSE, TRUE, sizeof (Transition), trans_count);
+ /* Name should be three or more characters */
+ /* FIXME: Should return FALSE if the name is too long.
+ This should simplify code later in this function. */
+ if (*pos - name_pos < 3)
+ return FALSE;
- last_offset = rules[0].std_offset;
+ memset (buffer, 0, size);
+ /* name_pos isn't 0-terminated, so we have to limit the length expressly */
+ len = (guint) (*pos - name_pos) > size - 1 ? size - 1 : (guint) (*pos - name_pos);
+ strncpy (buffer, name_pos, len);
+ *pos += quoted;
+ return TRUE;
+}
- for (i = 0; i < rules_num - 1; i++)
- {
- if (rules[i].dlt_start.mon)
- {
- /* Standard */
- info[0].gmt_offset = rules[i].std_offset;
- info[0].is_dst = FALSE;
- info[0].is_standard = rules[i].dlt_end.isstd;
- info[0].is_gmt = rules[i].dlt_end.isgmt;
+static gboolean
+parse_identifier_boundaries (gchar **pos, TimeZoneRule *tzr)
+{
+ if (*(*pos)++ != ',')
+ return FALSE;
- if (rules[i].std_name)
- info[0].abbrev = g_strdup (rules[i].std_name);
+ /* Start date */
+ if (!parse_identifier_boundary (pos, &(tzr->dlt_start)) || *(*pos)++ != ',')
+ return FALSE;
- else
- info[0].abbrev = g_strdup_printf ("%+03d%02d",
- (int) rules[i].std_offset / 3600,
- (int) abs (rules[i].std_offset / 60) % 60);
+ /* End date */
+ if (!parse_identifier_boundary (pos, &(tzr->dlt_end)))
+ return FALSE;
+ return TRUE;
+}
+/*
+ * Creates an array of TimeZoneRule from a TZ environment variable
+ * type of identifier. Should free rules afterwards
+ */
+static guint
+rules_from_identifier (const gchar *identifier,
+ TimeZoneRule **rules)
+{
+ gchar *pos;
+ TimeZoneRule tzr;
- /* Daylight */
- info[1].gmt_offset = rules[i].dlt_offset;
- info[1].is_dst = TRUE;
- info[1].is_standard = rules[i].dlt_start.isstd;
- info[1].is_gmt = rules[i].dlt_start.isgmt;
+ g_assert (rules != NULL);
- if (rules[i].dlt_name)
- info[1].abbrev = g_strdup (rules[i].dlt_name);
+ *rules = NULL;
- else
- info[1].abbrev = g_strdup_printf ("%+03d%02d",
- (int) rules[i].dlt_offset / 3600,
- (int) abs (rules[i].dlt_offset / 60) % 60);
+ if (!identifier)
+ return 0;
- if (rules[i].dlt_start.mon < rules[i].dlt_end.mon)
- {
- g_array_append_val (gtz->t_info, info[1]);
- g_array_append_val (gtz->t_info, info[0]);
- }
- else
- {
- g_array_append_val (gtz->t_info, info[0]);
- g_array_append_val (gtz->t_info, info[1]);
- }
+ pos = (gchar*)identifier;
+ memset (&tzr, 0, sizeof (tzr));
+ /* Standard offset */
+ if (!(set_tz_name (&pos, tzr.std_name, NAME_SIZE)) ||
+ !parse_offset (&pos, &(tzr.std_offset)))
+ return 0;
+
+ if (*pos == 0)
+ {
+ return create_ruleset_from_rule (rules, &tzr);
+ }
+
+ /* Format 2 */
+ if (!(set_tz_name (&pos, tzr.dlt_name, NAME_SIZE)))
+ return 0;
+ parse_offset (&pos, &(tzr.dlt_offset));
+ if (tzr.dlt_offset == 0) /* No daylight offset given, assume it's 1
+ hour earlier that standard */
+ tzr.dlt_offset = tzr.std_offset - 3600;
+ if (*pos == '\0')
+#ifdef G_OS_WIN32
+ /* Windows allows us to use the US DST boundaries if they're not given */
+ {
+ guint i, rules_num = 0;
- /* Transition dates */
- for (year = rules[i].start_year; year < rules[i+1].start_year; year++)
+ /* Use US rules, Windows' default is Pacific Standard Time */
+ if ((rules_num = rules_from_windows_time_zone ("Pacific Standard Time",
+ NULL,
+ rules)))
+ {
+ for (i = 0; i < rules_num - 1; i++)
{
- if (rules[i].dlt_start.mon < rules[i].dlt_end.mon)
- {
- /* Daylight Data */
- trans.info_index = y;
- trans.time = boundary_for_year (&rules[i].dlt_start, year,
- last_offset, rules[i].std_offset);
- g_array_insert_val (gtz->transitions, x++, trans);
- last_offset = rules[i].dlt_offset;
-
- /* Standard Data */
- trans.info_index = y+1;
- trans.time = boundary_for_year (&rules[i].dlt_end, year,
- last_offset, rules[i].std_offset);
- g_array_insert_val (gtz->transitions, x++, trans);
- last_offset = rules[i].std_offset;
- }
- else
- {
- /* Standard Data */
- trans.info_index = y;
- trans.time = boundary_for_year (&rules[i].dlt_end, year,
- last_offset, rules[i].std_offset);
- if (x >= 0)
- g_array_insert_val (gtz->transitions, x++, trans);
- else
- x++;
- last_offset = rules[i].std_offset;
-
- /* Daylight Data */
- trans.info_index = y+1;
- trans.time = boundary_for_year (&rules[i].dlt_start, year,
- last_offset, rules[i].std_offset);
- g_array_insert_val (gtz->transitions, x++, trans);
- last_offset = rules[i].dlt_offset;
- }
+ (*rules)[i].std_offset = - tzr.std_offset;
+ (*rules)[i].dlt_offset = - tzr.dlt_offset;
+ strcpy ((*rules)[i].std_name, tzr.std_name);
+ strcpy ((*rules)[i].dlt_name, tzr.dlt_name);
}
- y += 2;
+ return rules_num;
}
else
- {
- /* Standard */
- info[0].gmt_offset = rules[i].std_offset;
- info[0].is_dst = FALSE;
- info[0].is_standard = FALSE;
- info[0].is_gmt = FALSE;
-
- if (rules[i].std_name)
- info[0].abbrev = g_strdup (rules[i].std_name);
-
- else
- info[0].abbrev = g_strdup_printf ("%+03d%02d",
- (int) rules[i].std_offset / 3600,
- (int) abs (rules[i].std_offset / 60) % 60);
-
- g_array_append_val (gtz->t_info, info[0]);
+ return 0;
+ }
+#else
+ return 0;
+#endif
+ /* Start and end required (format 2) */
+ if (!parse_identifier_boundaries (&pos, &tzr))
+ return 0;
- last_offset = rules[i].std_offset;
+ return create_ruleset_from_rule (rules, &tzr);
+}
- y++;
- }
+#ifdef G_OS_UNIX
+static GTimeZone *
+parse_footertz (const gchar *footer, size_t footerlen)
+{
+ gchar *tzstring = g_strndup (footer + 1, footerlen - 2);
+ GTimeZone *footertz = NULL;
+
+ /* FIXME: The allocation for tzstring could be avoided by
+ passing a gsize identifier_len argument to rules_from_identifier
+ and changing the code in that function to stop assuming that
+ identifier is nul-terminated. */
+ TimeZoneRule *rules;
+ guint rules_num = rules_from_identifier (tzstring, &rules);
+
+ g_free (tzstring);
+ if (rules_num > 1)
+ {
+ footertz = g_slice_new0 (GTimeZone);
+ init_zone_from_rules (footertz, rules, rules_num, NULL);
+ footertz->ref_count++;
}
+ g_free (rules);
+ return footertz;
}
+#endif
/* Construction {{{1 */
/**
* g_time_zone_new:
- * @identifier: (allow-none): a timezone identifier
+ * @identifier: (nullable): a timezone identifier
+ *
+ * A version of g_time_zone_new_identifier() which returns the UTC time zone
+ * if @identifier could not be parsed or loaded.
*
- * Creates a #GTimeZone corresponding to @identifier.
+ * If you need to check whether @identifier was loaded successfully, use
+ * g_time_zone_new_identifier().
+ *
+ * Returns: (transfer full) (not nullable): the requested timezone
+ * Deprecated: 2.68: Use g_time_zone_new_identifier() instead, as it provides
+ * error reporting. Change your code to handle a potentially %NULL return
+ * value.
+ *
+ * Since: 2.26
+ **/
+GTimeZone *
+g_time_zone_new (const gchar *identifier)
+{
+ GTimeZone *tz = g_time_zone_new_identifier (identifier);
+
+ /* Always fall back to UTC. */
+ if (tz == NULL)
+ tz = g_time_zone_new_utc ();
+
+ g_assert (tz != NULL);
+
+ return g_steal_pointer (&tz);
+}
+
+/**
+ * g_time_zone_new_identifier:
+ * @identifier: (nullable): a timezone identifier
+ *
+ * Creates a #GTimeZone corresponding to @identifier. If @identifier cannot be
+ * parsed or loaded, %NULL is returned.
*
* @identifier can either be an RFC3339/ISO 8601 time offset or
- * something that would pass as a valid value for the
- * <varname>TZ</varname> environment variable (including %NULL).
+ * something that would pass as a valid value for the `TZ` environment
+ * variable (including %NULL).
+ *
+ * In Windows, @identifier can also be the unlocalized name of a time
+ * zone for standard time, for example "Pacific Standard Time".
+ *
+ * Valid RFC3339 time offsets are `"Z"` (for UTC) or
+ * `"±hh:mm"`. ISO 8601 additionally specifies
+ * `"±hhmm"` and `"±hh"`. Offsets are
+ * time values to be added to Coordinated Universal Time (UTC) to get
+ * the local time.
*
- * Valid RFC3339 time offsets are <literal>"Z"</literal> (for UTC) or
- * <literal>"±hh:mm"</literal>. ISO 8601 additionally specifies
- * <literal>"±hhmm"</literal> and <literal>"±hh"</literal>.
+ * In UNIX, the `TZ` environment variable typically corresponds
+ * to the name of a file in the zoneinfo database, an absolute path to a file
+ * somewhere else, or a string in
+ * "std offset [dst [offset],start[/time],end[/time]]" (POSIX) format.
+ * There are no spaces in the specification. The name of standard
+ * and daylight savings time zone must be three or more alphabetic
+ * characters. Offsets are time values to be added to local time to
+ * get Coordinated Universal Time (UTC) and should be
+ * `"[±]hh[[:]mm[:ss]]"`. Dates are either
+ * `"Jn"` (Julian day with n between 1 and 365, leap
+ * years not counted), `"n"` (zero-based Julian day
+ * with n between 0 and 365) or `"Mm.w.d"` (day d
+ * (0 <= d <= 6) of week w (1 <= w <= 5) of month m (1 <= m <= 12), day
+ * 0 is a Sunday). Times are in local wall clock time, the default is
+ * 02:00:00.
*
- * The <varname>TZ</varname> environment variable typically corresponds
- * to the name of a file in the zoneinfo database, but there are many
- * other possibilities. Note that those other possibilities are not
- * currently implemented, but are planned.
+ * In Windows, the "tzn[+|–]hh[:mm[:ss]][dzn]" format is used, but also
+ * accepts POSIX format. The Windows format uses US rules for all time
+ * zones; daylight savings time is 60 minutes behind the standard time
+ * with date and time of change taken from Pacific Standard Time.
+ * Offsets are time values to be added to the local time to get
+ * Coordinated Universal Time (UTC).
*
* g_time_zone_new_local() calls this function with the value of the
- * <varname>TZ</varname> environment variable. This function itself is
- * independent of the value of <varname>TZ</varname>, but if @identifier
- * is %NULL then <filename>/etc/localtime</filename> will be consulted
- * to discover the correct timezone.
- *
- * See <ulink
- * url='http://tools.ietf.org/html/rfc3339#section-5.6'>RFC3339
- * §5.6</ulink> for a precise definition of valid RFC3339 time offsets
- * (the <varname>time-offset</varname> expansion) and ISO 8601 for the
- * full list of valid time offsets. See <ulink
- * url='http://www.gnu.org/s/libc/manual/html_node/TZ-Variable.html'>The
- * GNU C Library manual</ulink> for an explanation of the possible
- * values of the <varname>TZ</varname> environment variable.
+ * `TZ` environment variable. This function itself is independent of
+ * the value of `TZ`, but if @identifier is %NULL then `/etc/localtime`
+ * will be consulted to discover the correct time zone on UNIX and the
+ * registry will be consulted or GetTimeZoneInformation() will be used
+ * to get the local time zone on Windows.
+ *
+ * If intervals are not available, only time zone rules from `TZ`
+ * environment variable or other means, then they will be computed
+ * from year 1900 to 2037. If the maximum year for the rules is
+ * available and it is greater than 2037, then it will followed
+ * instead.
+ *
+ * See
+ * [RFC3339 §5.6](http://tools.ietf.org/html/rfc3339#section-5.6)
+ * for a precise definition of valid RFC3339 time offsets
+ * (the `time-offset` expansion) and ISO 8601 for the
+ * full list of valid time offsets. See
+ * [The GNU C Library manual](http://www.gnu.org/s/libc/manual/html_node/TZ-Variable.html)
+ * for an explanation of the possible
+ * values of the `TZ` environment variable. See
+ * [Microsoft Time Zone Index Values](http://msdn.microsoft.com/en-us/library/ms912391%28v=winembedded.11%29.aspx)
+ * for the list of time zones on Windows.
*
* You should release the return value by calling g_time_zone_unref()
* when you are done with it.
*
- * Returns: the requested timezone
- *
- * Since: 2.26
- **/
+ * Returns: (transfer full) (nullable): the requested timezone, or %NULL on
+ * failure
+ * Since: 2.68
+ */
GTimeZone *
-g_time_zone_new (const gchar *identifier)
+g_time_zone_new_identifier (const gchar *identifier)
{
GTimeZone *tz = NULL;
-
- G_LOCK (time_zones);
- if (time_zones == NULL)
- time_zones = g_hash_table_new (g_str_hash, g_str_equal);
+ TimeZoneRule *rules;
+ gint rules_num;
+ gchar *resolved_identifier = NULL;
if (identifier)
{
+ G_LOCK (time_zones);
+ if (time_zones == NULL)
+ time_zones = g_hash_table_new (g_str_hash, g_str_equal);
+
tz = g_hash_table_lookup (time_zones, identifier);
if (tz)
{
G_UNLOCK (time_zones);
return tz;
}
+ else
+ resolved_identifier = g_strdup (identifier);
+ }
+ else
+ {
+ G_LOCK (tz_default);
+#ifdef G_OS_UNIX
+ resolved_identifier = zone_identifier_unix ();
+#elif defined (G_OS_WIN32)
+ resolved_identifier = windows_default_tzname ();
+#endif
+ if (tz_default)
+ {
+ /* Flush default if changed. If the identifier couldn’t be resolved,
+ * we’re going to fall back to UTC eventually, so don’t clear out the
+ * cache if it’s already UTC. */
+ if (!(resolved_identifier == NULL && g_str_equal (tz_default->name, "UTC")) &&
+ g_strcmp0 (tz_default->name, resolved_identifier) != 0)
+ {
+ g_clear_pointer (&tz_default, g_time_zone_unref);
+ }
+ else
+ {
+ tz = g_time_zone_ref (tz_default);
+ G_UNLOCK (tz_default);
+
+ g_free (resolved_identifier);
+ return tz;
+ }
+ }
}
tz = g_slice_new0 (GTimeZone);
- tz->name = g_strdup (identifier);
tz->ref_count = 0;
zone_for_constant_offset (tz, identifier);
+ if (tz->t_info == NULL &&
+ (rules_num = rules_from_identifier (identifier, &rules)))
+ {
+ init_zone_from_rules (tz, rules, rules_num, g_steal_pointer (&resolved_identifier));
+ g_free (rules);
+ }
+
if (tz->t_info == NULL)
{
#ifdef G_OS_UNIX
- GBytes *zoneinfo = zone_info_unix (identifier);
- if (!zoneinfo)
- zone_for_constant_offset (tz, "UTC");
- else
+ GBytes *zoneinfo = zone_info_unix (identifier, resolved_identifier);
+ if (zoneinfo != NULL)
{
- init_zone_from_iana_info (tz, zoneinfo);
+ init_zone_from_iana_info (tz, zoneinfo, g_steal_pointer (&resolved_identifier));
g_bytes_unref (zoneinfo);
}
-#elif defined G_OS_WIN32
+#elif defined (G_OS_WIN32)
+ if ((rules_num = rules_from_windows_time_zone (identifier,
+ resolved_identifier,
+ &rules)))
+ {
+ init_zone_from_rules (tz, rules, rules_num, g_steal_pointer (&resolved_identifier));
+ g_free (rules);
+ }
#endif
}
- if (tz->t_info != NULL)
+#if defined (G_OS_WIN32)
+ if (tz->t_info == NULL)
+ {
+ if (identifier == NULL)
+ {
+ TIME_ZONE_INFORMATION tzi;
+
+ if (GetTimeZoneInformation (&tzi) != TIME_ZONE_ID_INVALID)
+ {
+ rules = g_new0 (TimeZoneRule, 2);
+
+ if (rule_from_windows_time_zone_info (&rules[0], &tzi))
+ {
+ memset (rules[0].std_name, 0, NAME_SIZE);
+ memset (rules[0].dlt_name, 0, NAME_SIZE);
+
+ rules[0].start_year = MIN_TZYEAR;
+ rules[1].start_year = MAX_TZYEAR;
+
+ init_zone_from_rules (tz, rules, 2, g_steal_pointer (&resolved_identifier));
+ }
+
+ g_free (rules);
+ }
+ }
+ }
+#endif
+
+ g_free (resolved_identifier);
+
+ /* Failed to load the timezone. */
+ if (tz->t_info == NULL)
{
+ g_slice_free (GTimeZone, tz);
+
if (identifier)
- g_hash_table_insert (time_zones, tz->name, tz);
+ G_UNLOCK (time_zones);
+ else
+ G_UNLOCK (tz_default);
+
+ return NULL;
+ }
+
+ g_assert (tz->name != NULL);
+ g_assert (tz->t_info != NULL);
+
+ if (identifier)
+ g_hash_table_insert (time_zones, tz->name, tz);
+ else if (tz->name)
+ {
+ /* Caching reference */
+ g_atomic_int_inc (&tz->ref_count);
+ tz_default = tz;
}
+
g_atomic_int_inc (&tz->ref_count);
- G_UNLOCK (time_zones);
+
+ if (identifier)
+ G_UNLOCK (time_zones);
+ else
+ G_UNLOCK (tz_default);
return tz;
}
GTimeZone *
g_time_zone_new_utc (void)
{
- return g_time_zone_new ("UTC");
+ static GTimeZone *utc = NULL;
+ static gsize initialised;
+
+ if (g_once_init_enter (&initialised))
+ {
+ utc = g_time_zone_new_identifier ("UTC");
+ g_assert (utc != NULL);
+ g_once_init_leave (&initialised, TRUE);
+ }
+
+ return g_time_zone_ref (utc);
}
/**
* zone may change between invocations to this function; for example,
* if the system administrator changes it.
*
- * This is equivalent to calling g_time_zone_new() with the value of the
- * <varname>TZ</varname> environment variable (including the possibility
- * of %NULL).
+ * This is equivalent to calling g_time_zone_new() with the value of
+ * the `TZ` environment variable (including the possibility of %NULL).
*
* You should release the return value by calling g_time_zone_unref()
* when you are done with it.
GTimeZone *
g_time_zone_new_local (void)
{
- return g_time_zone_new (getenv ("TZ"));
+ const gchar *tzenv = g_getenv ("TZ");
+ GTimeZone *tz;
+
+ G_LOCK (tz_local);
+
+ /* Is time zone changed and must be flushed? */
+ if (tz_local && g_strcmp0 (g_time_zone_get_identifier (tz_local), tzenv))
+ g_clear_pointer (&tz_local, g_time_zone_unref);
+
+ if (tz_local == NULL)
+ tz_local = g_time_zone_new_identifier (tzenv);
+ if (tz_local == NULL)
+ tz_local = g_time_zone_new_utc ();
+
+ tz = g_time_zone_ref (tz_local);
+
+ G_UNLOCK (tz_local);
+
+ return tz;
+}
+
+/**
+ * g_time_zone_new_offset:
+ * @seconds: offset to UTC, in seconds
+ *
+ * Creates a #GTimeZone corresponding to the given constant offset from UTC,
+ * in seconds.
+ *
+ * This is equivalent to calling g_time_zone_new() with a string in the form
+ * `[+|-]hh[:mm[:ss]]`.
+ *
+ * It is possible for this function to fail if @seconds is too big (greater than
+ * 24 hours), in which case this function will return the UTC timezone for
+ * backwards compatibility. To detect failures like this, use
+ * g_time_zone_new_identifier() directly.
+ *
+ * Returns: (transfer full): a timezone at the given offset from UTC, or UTC on
+ * failure
+ * Since: 2.58
+ */
+GTimeZone *
+g_time_zone_new_offset (gint32 seconds)
+{
+ GTimeZone *tz = NULL;
+ gchar *identifier = NULL;
+
+ /* Seemingly, we should be using @seconds directly to set the
+ * #TransitionInfo.gmt_offset to avoid all this string building and parsing.
+ * However, we always need to set the #GTimeZone.name to a constructed
+ * string anyway, so we might as well reuse its code.
+ * g_time_zone_new_identifier() should never fail in this situation. */
+ identifier = g_strdup_printf ("%c%02u:%02u:%02u",
+ (seconds >= 0) ? '+' : '-',
+ (ABS (seconds) / 60) / 60,
+ (ABS (seconds) / 60) % 60,
+ ABS (seconds) % 60);
+ tz = g_time_zone_new_identifier (identifier);
+
+ if (tz == NULL)
+ tz = g_time_zone_new_utc ();
+ else
+ g_assert (g_time_zone_get_offset (tz, 0) == seconds);
+
+ g_assert (tz != NULL);
+ g_free (identifier);
+
+ return tz;
}
#define TRANSITION(n) g_array_index (tz->transitions, Transition, n)
#define TRANSITION_INFO(n) g_array_index (tz->t_info, TransitionInfo, n)
/* Internal helpers {{{1 */
-/* Note that interval 0 is *before* the first transition time, so
- * interval 1 gets transitions[0].
+/* NB: Interval 0 is before the first transition, so there's no
+ * transition structure to point to which TransitionInfo to
+ * use. Rule-based zones are set up so that TI 0 is always standard
+ * time (which is what's in effect before Daylight time got started
+ * in the early 20th century), but IANA tzfiles don't follow that
+ * convention. The tzfile documentation says to use the first
+ * standard-time (i.e., non-DST) tinfo, so that's what we do.
*/
inline static const TransitionInfo*
interval_info (GTimeZone *tz,
if (interval && tz->transitions && interval <= tz->transitions->len)
index = (TRANSITION(interval - 1)).info_index;
else
- index = 0;
+ {
+ for (index = 0; index < tz->t_info->len; index++)
+ {
+ TransitionInfo *tzinfo = &(TRANSITION_INFO(index));
+ if (!tzinfo->is_dst)
+ return tzinfo;
+ }
+ index = 0;
+ }
+
return &(TRANSITION_INFO(index));
}
guint interval)
{
if (tz->transitions && interval < tz->transitions->len)
- return (TRANSITION(interval)).time - 1;
+ {
+ gint64 lim = (TRANSITION(interval)).time;
+ return lim - (lim != G_MININT64);
+ }
return G_MAXINT64;
}
}
-inline static gboolean
-interval_isgmt (GTimeZone *tz,
- guint interval)
-{
- g_return_val_if_fail (tz->t_info != NULL, 0);
- return interval_info (tz, interval)->is_gmt;
-}
-
-inline static gboolean
-interval_isstandard (GTimeZone *tz,
- guint interval)
-{
- return interval_info (tz, interval)->is_standard;
-}
-
inline static gchar*
interval_abbrev (GTimeZone *tz,
guint interval)
GTimeType type,
gint64 *time_)
{
- gint i;
- guint intervals;
+ guint i, intervals;
+ gboolean interval_is_dst;
if (tz->transitions == NULL)
return 0;
*time_ = interval_local_start (tz, i);
}
- else if (interval_isdst (tz, i) != type)
- /* it's in this interval, but dst flag doesn't match.
- * check neighbours for a better fit. */
+ else
{
- if (i && *time_ <= interval_local_end (tz, i - 1))
- i--;
-
- else if (i < intervals &&
- *time_ >= interval_local_start (tz, i + 1))
- i++;
+ interval_is_dst = interval_isdst (tz, i);
+ if ((interval_is_dst && type != G_TIME_TYPE_DAYLIGHT) ||
+ (!interval_is_dst && type == G_TIME_TYPE_DAYLIGHT))
+ {
+ /* it's in this interval, but dst flag doesn't match.
+ * check neighbours for a better fit. */
+ if (i && *time_ <= interval_local_end (tz, i - 1))
+ i--;
+
+ else if (i < intervals &&
+ *time_ >= interval_local_start (tz, i + 1))
+ i++;
+ }
}
}
* @type: the #GTimeType of @time_
* @time_: a number of seconds since January 1, 1970
*
- * Finds an the interval within @tz that corresponds to the given @time_.
+ * Finds an interval within @tz that corresponds to the given @time_.
* The meaning of @time_ depends on @type.
*
* If @type is %G_TIME_TYPE_UNIVERSAL then this function will always
* succeed (since universal time is monotonic and continuous).
*
- * Otherwise @time_ is treated is local time. The distinction between
+ * Otherwise @time_ is treated as local time. The distinction between
* %G_TIME_TYPE_STANDARD and %G_TIME_TYPE_DAYLIGHT is ignored except in
* the case that the given @time_ is ambiguous. In Toronto, for example,
* 01:30 on November 7th 2010 occurred twice (once inside of daylight
GTimeType type,
gint64 time_)
{
- gint i;
- guint intervals;
+ guint i, intervals;
+ gboolean interval_is_dst;
if (tz->transitions == NULL)
return 0;
return -1;
}
- else if (interval_isdst (tz, i) != type)
+ else
{
- if (i && time_ <= interval_local_end (tz, i - 1))
- i--;
+ interval_is_dst = interval_isdst (tz, i);
+ if ((interval_is_dst && type != G_TIME_TYPE_DAYLIGHT) ||
+ (!interval_is_dst && type == G_TIME_TYPE_DAYLIGHT))
+ {
+ if (i && time_ <= interval_local_end (tz, i - 1))
+ i--;
- else if (i < intervals && time_ >= interval_local_start (tz, i + 1))
- i++;
+ else if (i < intervals && time_ >= interval_local_start (tz, i + 1))
+ i++;
+ }
}
return i;
return interval_isdst (tz, (guint)interval);
}
+/**
+ * g_time_zone_get_identifier:
+ * @tz: a #GTimeZone
+ *
+ * Get the identifier of this #GTimeZone, as passed to g_time_zone_new().
+ * If the identifier passed at construction time was not recognised, `UTC` will
+ * be returned. If it was %NULL, the identifier of the local timezone at
+ * construction time will be returned.
+ *
+ * The identifier will be returned in the same format as provided at
+ * construction time: if provided as a time offset, that will be returned by
+ * this function.
+ *
+ * Returns: identifier for this timezone
+ * Since: 2.58
+ */
+const gchar *
+g_time_zone_get_identifier (GTimeZone *tz)
+{
+ g_return_val_if_fail (tz != NULL, NULL);
+
+ return tz->name;
+}
+
/* Epilogue {{{1 */
/* vim:set foldmethod=marker: */