August 2026 - tz - lists.iana.org

Initialism for Alberta Time (ABT) — UTC-6
by Daylight Saving Time Sept. 1, 2026

Sept. 1, 2026

To the Time Zone Database, We are writing to share information about Alberta’s transition to permanent Daylight Saving Time<https://www.alberta.ca/albertas-new-time-system-abt>, also known as Alberta Time (ABT). Beginning November 1, 2026, Alberta will no longer switch between Daylight Saving Time and Mountain Standard Time. Instead, the province will remain on Alberta Time year-round (UTC-6:00), eliminating the need to change clocks twice a year. We are recommending the use of the abbreviation 'ABT' as a convenient way to refer to Alberta Time. If you require additional information or wish to discuss this further, please feel free to reach out. Sincerely, Alberta Time Team Ministry of Service Alberta and Red Tape Reduction Government of Alberta [AB Logo blue CMYK - no tagline] Classification: Protected A

6 5

Europe/Sofia: end of DST in 1996 appears to have no legal basis; also 1983-1990 transition times
by Cult Consult Sept. 1, 2026

Sept. 1, 2026

Hello, I would like to report a likely error in Europe/Sofia's 1996 data, found while verifying tzdb's Bulgarian history against primary legal sources, and to flag a related open question for 1983-1990. This is a preliminary report: we have initiated formal archive requests to establish what clocks actually did, and will follow up with documentary evidence either way. ## Autumn 1996 tzdb ends Bulgarian DST on 1996-10-27 (Europe/Sofia is on the shared E-Eur rules for this period, with no Bulgaria-specific citation in the europe file). We can find no Bulgarian legal basis for an October end in 1996: * The DST regime in force through 1996 was established by Council of Ministers Decree No. 6 of 26 January 1979 (State Gazette / Държавен вестник No. 12/1979), which provides for summer time "until 30 September" each year. * A last-Sunday-of-October end appears in Bulgarian law for the first time in Decree No. 94 of 13 March 1997 (State Gazette No. 24/1997). Its Article 2 also fixes the autumn change at 04:00 local time. The full text is in the government's legal-information system PRIS (pris.government.bg). The decree's repeal clause enumerates the entire prior regime: Decree 6/1979 and item 12 of Directive (Разпореждане) No. 15 of 11 March 1980, the latter marked as never published. * We find no 1996 act changing the end date. PRIS records Council of Ministers acts since 1990 and contains none on clock time before the 1997 decree. If the law was followed, summer time in Bulgaria ended in late September 1996, and Europe/Sofia is one hour ahead of reality (+03 instead of +02) for roughly 1996-09-29 through 1996-10-27. The report tzdb cites for Bulgaria (Plamen Simenov via Steffen Thorsen, 1999-09-09) describes the 1997 decree's rules and does not address autumn 1996. Since law and practice can differ, before proposing a patch we have initiated: - a request to the Bulgarian National Library for scans of five national dailies for 27-30 September and 25-28 October 1996 (the clock-change notices published to the public); - freedom-of-information requests to the Council of Ministers and the State Archives Agency for the unpublished Directive 15/1980 and all clock-time acts 1979-1997; - inquiries with the Bulgarian Institute of Metrology and the BTA news-agency archive. We will report what the archives show, in either direction. ## Why this matters to us We build astrological software, which makes a historical UTC offset an input to a computation rather than a display detail. The quantity most sensitive to it is the ascendant, the ecliptic degree rising at the given moment, which completes a full circuit every day. Measured at Sofia's latitude (42.698 N) for 1 October 1996, sampled every minute of the day: a one-hour error in the offset moves the ascendant by between 11.6 and 27.1 degrees, and it lands in a different one of the twelve 30-degree sectors in exactly half of the minutes of that day. The 50% is not a coincidence; a one-hour arc averages 15 degrees against 30-degree sectors. So for a Bulgarian birth between 29 September and 27 October 1996, a wrong offset does not produce a slightly wrong result. Half the time it produces a categorically different one, and there is nothing in the output to indicate which. That is also why we are asking about 1983-1990, where the published record looks incomplete rather than wrong, and why we would rather see a documented gap than a confident value. The same sensitivity applies well outside our field: anything that reconstructs past local civil time from records inherits whatever the database says about a date nobody can re-observe. We are not asking for a change that happens to suit us. If the archives show that clocks in practice followed the October date despite the decree, we will report that, with the same evidence and in the same detail. ## 1983-1990 transition times For 1983-1990 tzdb has Sofia on C-Eur rules (02:00s/03:00d transitions), apparently ex-Shanks and uncited; Bulgarian retrospective sources consistently describe midnight transitions for the whole 1979-1996 period, and tzdb itself moves Sofia to midnight rules (E-Eur) in 1991 with no documented reason. Our archive requests cover all sixteen transition nights of 1983-1990 (the last Sundays of March and September) in the national dailies of the period, asking specifically for the announced hour. If any list members have contemporaneous Bulgarian sources (newspapers, broadcast listings) for these transitions, we would be glad to hear of them. For completeness: the same review confirmed tzdb is correct on the 04:00 autumn hour for 1997-1998, the decree text matches tzdb exactly, against several Bulgarian secondary sources that claim 03:00. Best regards, Ivan Petrov Ivanov Sofia, Bulgaria

2 1

[PROPOSED] Document Canada/{Pacific,Mountain} glitches
by Paul Eggert Aug. 31, 2026

Aug. 31, 2026

* NEWS: Document 2026b and 2026c changes to the backward compatibility links Canada/Pacific and Canada/Mountain. They will now use the abbreviations MST and CST, respectively, for affected timestamps. That’s confusing, and is an argument for using numeric abbreviations instead, but for now just document the situation. These old names became obsolescent in 1993 and are little used. --- NEWS | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/NEWS b/NEWS index cca58326..b6b57a78 100644 --- a/NEWS +++ b/NEWS @@ -75,6 +75,11 @@ Release 2026c - 2026-07-08 10:23:58 -0700 02:00 instead, for the same reason we introduced a similarly temporary hack for British Columbia in 2026b. + As the change affects both America/Edmonton and its backward + compatibility link, the obsolescent setting TZ="Canada/Mountain" + will now use the abbreviation CST for affected timestamps, + akin to TZ="Canada/Pacific" behavior introduced in 2026b. + Although another TZDB release will likely be needed soon because Northwest Territories will likely follow Alberta, the legal formalities have not yet taken place. @@ -137,6 +142,17 @@ Release 2026b - 2026-04-22 23:06:43 -0700 limitation in CLDR 48.1 (2026-01-08). This temporary hack is planned to be removed after CLDR is fixed. + To match longstanding practice for -07 standard time in North + America, such as for America/Dawson, America/Dawson_Creek, + America/Fort_Nelson and America/Whitehose, affected BC + timestamps use the abbreviation MST. As the change affects both + America/Vancouver and its backward compatibility link, the + obsolescent setting TZ="Canada/Pacific" will now use the + abbreviation MST. Future TZDB releases may change the + abbreviation if something else becomes popular across areas + observing -07, and may change it to "-07" if no alphabetic + consensus emerges. + Changes to code zic no longer mishandles a last transition to a new time type. -- 2.55.0

3 2

[PROPOSED] Port to Haiku
by Paul Eggert Aug. 30, 2026

Aug. 30, 2026

* Makefile (UTF8_LOCALE_MISSING): Also test [[:alpha:]], which does not work on Haiku in UTF-8 locales. (now.ck): Work even if the platform, like Haiku, supports symlinks but not hard links. * NEWS: Mention the port. * localtime.c (issetugid) [!HAVE_ISSETUGID]: Define to localtime_issetugid, to avoid a clash on Haiku which declares this function but by default does not define it. * private.h (HAVE_DECL_TIMEGM, TM_GMTOFF, TM_ZONE) [__HAIKU__]: Default to BSD behavior. --- Makefile | 6 +++--- NEWS | 2 ++ localtime.c | 4 ++++ private.h | 4 ++-- 4 files changed, 11 insertions(+), 5 deletions(-) diff --git a/Makefile b/Makefile index 5d463b74..96a7462e 100644 --- a/Makefile +++ b/Makefile @@ -922,7 +922,7 @@ check_mild: check_web check_zishrink \ UTF8_LOCALE_MISSING = \ { test ! '$(UTF8_LOCALE)' \ || ! printf 'A\304\200B\n' \ - | LC_ALL='$(UTF8_LOCALE)' grep -q '^A.B$$' >/dev/null 2>&1 \ + | LC_ALL='$(UTF8_LOCALE)' grep -q '^A[[:alpha:]]B$$' >/dev/null 2>&1 \ || { export LC_ALL='$(UTF8_LOCALE)'; false; }; } character-set.ck: $(ENCHILADA) @@ -1006,11 +1006,11 @@ now.ck: checknow.awk date tzdata.zi zdump zic zone1970.tab zonenow.tab now=$(CHECK_NOW_TIMESTAMP) && \ future=$$(($(CHECK_NOW_FUTURE_SECS) + $$now)) && \ ./zdump -i -t $$now,$$future \ - $$(find "$$PWD/$@d"/????*/ -type f) \ + $$(find "$$PWD/$@d"/????*/ -type f -o -type l) \ >$(a)d/zdump-now.tab && \ ./zdump -i -t 0,$$future \ $$(find "$$PWD/$@d" -name Etc -prune \ - -o -type f ! -name '*.tab' -print) \ + -o $ -type f -o -type l $ ! -name '*.tab' -print) \ >$(a)d/zdump-1970.tab && \ $(AWK) \ -v now=$$now \ diff --git a/NEWS b/NEWS index a60f119f..71595a8f 100644 --- a/NEWS +++ b/NEWS @@ -65,6 +65,8 @@ Unreleased, experimental changes functions have not used since 2026c’s removal of the old posixrules feature. + tzcode has been ported to Haiku. + localtime.c now works again by default on AIX and DragonFly BSD. Changes to documentation diff --git a/localtime.c b/localtime.c index 7bc9387f..608ddaa6 100644 --- a/localtime.c +++ b/localtime.c @@ -348,6 +348,10 @@ static int fstatat(int dd, char const *path, struct stat *st, int flags) # include <sys/auxv.h> # endif +/* Avoid clash if headers declare but libraries do not define issetugid. */ +# undef issetugid +# define issetugid localtime_issetugid + /* Return 1 if the process is privileged, 0 otherwise. */ static int issetugid(void) diff --git a/private.h b/private.h index f130bc49..f2861c94 100644 --- a/private.h +++ b/private.h @@ -820,7 +820,7 @@ void tzset(void); # if (202311 <= __STDC_VERSION__ \ || defined __GLIBC__ || defined __tm_zone /* musl */ \ || defined __FreeBSD__ || defined __NetBSD__ || defined __OpenBSD__ \ - || defined __DragonFly__ \ + || defined __DragonFly__ || defined __HAIKU__ \ || (defined __APPLE__ && defined __MACH__)) # define HAVE_DECL_TIMEGM 1 # else @@ -915,7 +915,7 @@ time_t posix2time(time_t); || defined __GLIBC__ \ || defined __tm_zone /* musl */ \ || defined __FreeBSD__ || defined __NetBSD__ || defined __OpenBSD__ \ - || defined __DragonFly__ \ + || defined __DragonFly__ || defined __HAIKU__ \ || (defined __APPLE__ && defined __MACH__)) # if !defined TM_GMTOFF && !defined NO_TM_GMTOFF # define TM_GMTOFF tm_gmtoff -- 2.55.0

2 1

[PROPOSED] Port back to DragonFly BSD
by Paul Eggert Aug. 30, 2026

Aug. 30, 2026

Build out-of-the-box on DragonFly BSD, instead of requiring the builder to specify CFLAGS by hand. The following changes take effect only if __DragonFly__ is defined. * NEWS: Mention the fixes. * localtime.c (HAVE___ISTHREADED): * private.h (HAVE_ISSETUGID, HAVE_DECL_TIMEGM, HAVE_MEMPCPY): Default to 1, not to 0. (TM_GMTOFF) [!NO_TM_GMTOFF]: Default to tm_gmtoff. (TM_ZONE) [!NO_TM_ZONE]: Default to tm_zone. --- NEWS | 3 +-- localtime.c | 2 +- private.h | 6 +++++- 3 files changed, 7 insertions(+), 4 deletions(-) diff --git a/NEWS b/NEWS index ef9689f3..a60f119f 100644 --- a/NEWS +++ b/NEWS @@ -65,8 +65,7 @@ Unreleased, experimental changes functions have not used since 2026c’s removal of the old posixrules feature. - localtime.c now works on AIX 7.3 by default again, working - around AIX’s POSIX-conformance glitch with st_ctim. + localtime.c now works again by default on AIX and DragonFly BSD. Changes to documentation diff --git a/localtime.c b/localtime.c index 276f8890..7bc9387f 100644 --- a/localtime.c +++ b/localtime.c @@ -64,7 +64,7 @@ struct stat { char st_ctime, st_dev, st_ino; }; # endif # if THREAD_PREFER_SINGLE # ifndef HAVE___ISTHREADED -# if defined __FreeBSD__ || defined __OpenBSD__ +# if defined __FreeBSD__ || defined __OpenBSD__ || defined __DragonFly__ # define HAVE___ISTHREADED 1 # else # define HAVE___ISTHREADED 0 diff --git a/private.h b/private.h index 2e782435..f130bc49 100644 --- a/private.h +++ b/private.h @@ -311,6 +311,7 @@ extern int optind; #ifndef HAVE_ISSETUGID # if (defined __FreeBSD__ || defined __NetBSD__ || defined __OpenBSD__ \ + || defined __DragonFly__ \ || (defined __linux__ && !defined __GLIBC__) /* Android, musl, etc. */ \ || (defined __APPLE__ && defined __MACH__) || defined __sun) # define HAVE_ISSETUGID 1 @@ -819,6 +820,7 @@ void tzset(void); # if (202311 <= __STDC_VERSION__ \ || defined __GLIBC__ || defined __tm_zone /* musl */ \ || defined __FreeBSD__ || defined __NetBSD__ || defined __OpenBSD__ \ + || defined __DragonFly__ \ || (defined __APPLE__ && defined __MACH__)) # define HAVE_DECL_TIMEGM 1 # else @@ -847,7 +849,8 @@ extern char **environ; #ifndef HAVE_MEMPCPY # if (defined mempcpy \ - || defined __FreeBSD__ || defined __NetBSD__ || defined __linux__) + || defined __FreeBSD__ || defined __NetBSD__ || defined __DragonFly__ \ + || defined __linux__) # define HAVE_MEMPCPY 1 # else # define HAVE_MEMPCPY 0 @@ -912,6 +915,7 @@ time_t posix2time(time_t); || defined __GLIBC__ \ || defined __tm_zone /* musl */ \ || defined __FreeBSD__ || defined __NetBSD__ || defined __OpenBSD__ \ + || defined __DragonFly__ \ || (defined __APPLE__ && defined __MACH__)) # if !defined TM_GMTOFF && !defined NO_TM_GMTOFF # define TM_GMTOFF tm_gmtoff -- 2.53.0

1 1

[PROPOSED 0/8] overflow fixes and speedups for localtime.c
by Paul Eggert Aug. 29, 2026

Aug. 29, 2026

This series of patches to localtime.c fix some unlikely integer overflows, along with some major performance speedups for unlikely TZif files. Thanks to David Sarkisyan for privately reporting performance issues. Paul Eggert (8): year_days refactoring Check bulging TZif TZ string abbrs Refactor transtime for future use locasub need not decrement seconds Fix unlikely overflows in time_t reduction Fix performance with extreme TZif transitions Be more generous re standard/wall and UT/local Port tzfile_changed back to AIX NEWS | 16 +++ localtime.c | 346 ++++++++++++++++++++-------------------------------- private.h | 3 +- strftime.c | 9 +- zdump.c | 4 +- zic.c | 6 +- 6 files changed, 156 insertions(+), 228 deletions(-) -- 2.55.0

1 8

Regional District of East Kootenay announces survey for time change
by James Bellaire Aug. 28, 2026

Aug. 28, 2026

The press release reads (in part): A new public survey has been launched to gather input on time observance in the southeast corner of BC. The survey is open through August 3 and is being done in partnership between the Regional District of East Kootenay, Town of Golden, and Columbia Shuswap Regional District (CSRD) Electoral Area A. “A dedicated project page has been set up that has the online survey along with other information. While it is being hosted by the RDEK, the survey covers the entire southeast corner of BC that is still on Mountain Time and is being done in partnership with all the local governments in the East Kootenay, Golden and CSRD Area A,” says RDEK Communications Manager, Loree Duczek. The survey can be found on https://engage.rdek.bc.ca/timechange and will remain open until 11:59pm on August 3. https://www.rdek.bc.ca/news/entry/regional_survey_launches_to_gather_feedba… The three options presented are "No Change" (continue to change clocks), Permanent Standard Time (UTC -7 or BC Time) or Permanent Daylight Time (UTC -6 or Alberta Time).

5 13

[PROPOSED] NWT regulations came into force today
by Paul Eggert Aug. 28, 2026

Aug. 28, 2026

* northamerica (America/Inuvik): Record the legal change date, with the usual temporary hack for CLDR. With the hack, this change merely records the law; it does not affect TZif data. --- northamerica | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-) diff --git a/northamerica b/northamerica index 9d05e891..6d26fdaa 100644 --- a/northamerica +++ b/northamerica @@ -2445,15 +2445,10 @@ Zone America/Fort_Nelson -8:10:47 - LMT 1884 # to fix a year-round standard time: # https://www.ntlegislativeassembly.ca/sites/default/files/legacy/bill_13_-_r… # -# Recent private correspondence with staffers in the Northwest Territories -# Department of Justice responsible for implementing the change indicates that, -# while the legal process is moving forward and the change is intended to come -# into force no later than 2026-11-01, at time of writing the expected -# regulation's official registration is not anticipated until September 2026. -# -# For most of NWT, this is already covered by the change to America/Edmonton, -# which alias America/Yellowknife links to. For now, assume below for -# America/Inuvik that the NWT change goes forward from 2026-11-01 02:00. +# From Paul Eggert (2026-08-21): +# The regulations were registered 2026-08-18 and came into force today: +# https://www.justice.gov.nt.ca/en/files/legislation/interpretation/interpret… +# Use a temporary hack like Alberta’s. # Rule NAME FROM TO - IN ON AT SAVE LETTER/S Rule NT_YK 1918 only - Apr 14 2:00 1:00 D @@ -2496,7 +2491,10 @@ Zone America/Cambridge_Bay 0 - -00 1920 # trading post est.? Zone America/Inuvik 0 - -00 1953 # Inuvik founded -8:00 NT_YK P%sT 1979 Apr lastSun 2:00 -7:00 NT_YK M%sT 1980 - -7:00 Canada M%sT 2026 Nov 1 2:00 # presumed + -7:00 Canada M%sT 2026 Aug 21 + # Temporary hack; see above. + -7:00 Canada MDT 2026 Nov 1 2:00 + # End of temporary hack. -6:00 - CST Zone America/Whitehorse -9:00:12 - LMT 1900 Aug 20 -9:00 NT_YK Y%sT 1965 -- 2.53.0

3 6

[PROPOSED] Check invalid/bulging TZif files more carefully
by Paul Eggert Aug. 27, 2026

Aug. 27, 2026

* localtime.c (tzloadbody): If the TZif file exceeds our capacity, return EOVERFLOW rather than silently pretending the TZif file was smaller than it is. Also, if the TZ string was invalid, return EFTYPE instead of ignoring the invalidity. (tzparse): Don’t fail merely because two types slots are not available. Fail only if we need more slots than are available. --- localtime.c | 25 ++++++++++++++----------- 1 file changed, 14 insertions(+), 11 deletions(-) diff --git a/localtime.c b/localtime.c index be3fe1d5..ca3ba9d4 100644 --- a/localtime.c +++ b/localtime.c @@ -1232,13 +1232,13 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, break; } if ((tzloadflags & TZLOAD_TZSTRING) && nread > 2 && - up->buf[0] == '\n' && up->buf[nread - 1] == '\n' && - sp->typecnt + 2 <= TZ_MAX_TYPES) { + up->buf[0] == '\n' && up->buf[nread - 1] == '\n') { struct state *ts = &lsp->u.st; up->buf[nread - 1] = '\0'; - if (tzparse(&up->buf[1], ts, sp)) { - + if (!tzparse(&up->buf[1], ts, sp)) + return EFTYPE; + else { /* Attempt to reuse existing abbreviations. Without this, America/Anchorage would consume 50 bytes for abbreviations, as @@ -1270,6 +1270,8 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, } } if (gotabbr == ts->typecnt) { + if (TZ_MAX_TYPES - sp->typecnt < ts->typecnt) + return EOVERFLOW; sp->charcnt = charcnt; /* Ignore any trailing, no-op transitions generated @@ -1288,10 +1290,8 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, || (0 < sp->timecnt && t <= sp->ats[sp->timecnt - 1])) continue; - if (TZ_MAX_TIMES <= sp->timecnt) { - sp->goahead = false; - break; - } + if (TZ_MAX_TIMES <= sp->timecnt) + return EOVERFLOW; sp->ats[sp->timecnt] = t; sp->types[sp->timecnt] = (sp->typecnt + ts->types[i]); @@ -1764,17 +1764,20 @@ tzparse(const char *name, struct state *sp, struct state const *basep) if (reversed || (starttime < endtime && endtime - starttime < yearsecs)) { - if (TZ_MAX_TIMES - 2 < timecnt) - break; sp->ats[timecnt] = janfirst; if (! increment_overflow_time(&sp->ats[timecnt], janoffset + starttime) - && atlo <= sp->ats[timecnt]) + && atlo <= sp->ats[timecnt]) { + if (TZ_MAX_TIMES <= timecnt) + return false; sp->types[timecnt++] = !reversed; + } sp->ats[timecnt] = janfirst; if (! increment_overflow_time(&sp->ats[timecnt], janoffset + endtime) && atlo <= sp->ats[timecnt]) { + if (TZ_MAX_TIMES <= timecnt) + return false; sp->types[timecnt++] = reversed; } } -- 2.53.0

1 1

Two administrative patches
by Tim Parenti Aug. 27, 2026

Aug. 27, 2026

Two mostly administrative patches tonight… 0001 reimplements the distinction in tz-art.html's references to xkcd between "strips" and "panels", as this had gotten a bit muddied recently. 0002 uses UT timestamps in UNTIL fields to be clearer that all of Alaska transitioned across the International Date Line at once. This is more easily maintainable and follows from a similar change in September 2022 for most continental US zones. The Juneau example given in tz-how-to.html is also updated to match, and its commentary is updated to better reflect that the transfer happened in the afternoon and not at midnight. -- Tim Parenti

1 0