- tz - lists.iana.org

[PROPOSED] Say more clearly that the tz mailing list is public
by Paul Eggert Dec. 10, 2025

Dec. 10, 2025

* CONTRIBUTING, africa, asia, australasia, backzone, europe, * northamerica, southamerica, theory.html, tz-art.html * tz-link.html: Say that the tz mailing list and its archive are public, and that people should not contribute confidential information. --- CONTRIBUTING | 6 +++++- africa | 6 +++--- asia | 6 +++--- australasia | 6 +++--- backzone | 8 ++++---- europe | 8 ++++---- northamerica | 6 +++--- southamerica | 6 +++--- theory.html | 11 ++++++++--- tz-art.html | 7 +++++-- tz-link.html | 12 +++++++++--- 11 files changed, 50 insertions(+), 32 deletions(-) diff --git a/CONTRIBUTING b/CONTRIBUTING index 1f064fa0..c5fa803f 100644 --- a/CONTRIBUTING +++ b/CONTRIBUTING @@ -10,7 +10,8 @@ warning, the data entries do not cover all of civil time before 1970, and undoubtedly errors remain in the code and data. Feel free to fill gaps or fix mistakes, and please email improvements to <tz(a)iana.org> for use in the future. In your email, please give -reliable sources that reviewers can check. +reliable sources that reviewers can check. The mailing list and its +archives are public, so please do not send confidential information. ## Contributing technical changes @@ -84,6 +85,9 @@ If you use Git the following workflow may be helpful: git format-patch main + * Check that the patch files and your email setup contain only + information that you want to make public. + * After reviewing the patch files, send the patches to <tz(a)iana.org> for others to review. diff --git a/africa b/africa index 11738a1e..ec067bee 100644 --- a/africa +++ b/africa @@ -4,9 +4,9 @@ # 2009-05-17 by Arthur David Olson. # This file is by no means authoritative; if you think you know better, -# go ahead and edit the file (and please send any changes to -# tz(a)iana.org for general use in the future). For more, please see -# the file CONTRIBUTING in the tz distribution. +# go ahead and edit the file, and please send any changes to +# the public mailing list tz(a)iana.org for general use in the future. +# For more, please see the file CONTRIBUTING in the tz distribution. # From Paul Eggert (2018-05-27): # diff --git a/asia b/asia index d3d1059a..dd76bdfc 100644 --- a/asia +++ b/asia @@ -4,9 +4,9 @@ # 2009-05-17 by Arthur David Olson. # This file is by no means authoritative; if you think you know better, -# go ahead and edit the file (and please send any changes to -# tz(a)iana.org for general use in the future). For more, please see -# the file CONTRIBUTING in the tz distribution. +# go ahead and edit the file, and please send any changes to +# the public mailing list tz(a)iana.org for general use in the future. +# For more, please see the file CONTRIBUTING in the tz distribution. # From Paul Eggert (2019-07-11): # diff --git a/australasia b/australasia index 5cf8f117..e52aab10 100644 --- a/australasia +++ b/australasia @@ -937,9 +937,9 @@ Zone Pacific/Efate 11:13:16 - LMT 1912 Jan 13 # Vila # NOTES # This file is by no means authoritative; if you think you know better, -# go ahead and edit the file (and please send any changes to -# tz(a)iana.org for general use in the future). For more, please see -# the file CONTRIBUTING in the tz distribution. +# go ahead and edit the file, and please send any changes to +# the public mailing list tz(a)iana.org for general use in the future. +# For more, please see the file CONTRIBUTING in the tz distribution. # From Paul Eggert (2018-11-18): # diff --git a/backzone b/backzone index dd7a8675..3c7e720a 100644 --- a/backzone +++ b/backzone @@ -2,10 +2,10 @@ # This file is in the public domain. -# This file is by no means authoritative; if you think you know -# better, go ahead and edit it (and please send any changes to -# tz(a)iana.org for general use in the future). For more, please see -# the file CONTRIBUTING in the tz distribution. +# This file is by no means authoritative; if you think you know better, +# go ahead and edit the file, and please send any changes to +# the public mailing list tz(a)iana.org for general use in the future. +# For more, please see the file CONTRIBUTING in the tz distribution. # When proposing changes to this file, please use 'git format-patch' # format, either by attaching the resulting .patch file to your email, diff --git a/europe b/europe index 4d7e1f3b..af521710 100644 --- a/europe +++ b/europe @@ -4,9 +4,9 @@ # 2009-05-17 by Arthur David Olson. # This file is by no means authoritative; if you think you know better, -# go ahead and edit the file (and please send any changes to -# tz(a)iana.org for general use in the future). For more, please see -# the file CONTRIBUTING in the tz distribution. +# go ahead and edit the file, and please send any changes to +# the public mailing list tz(a)iana.org for general use in the future. +# For more, please see the file CONTRIBUTING in the tz distribution. # From Paul Eggert (2017-02-10): # @@ -42,7 +42,7 @@ # <https://www.jstor.org/stable/1774359>. He writes: # "It is requested that corrections and additions to these tables # may be sent to Mr. John Milne, Royal Geographical Society, -# Savile Row, London." Nowadays please email them to tz(a)iana.org. +# Savile Row, London." Nowadays please see the file CONTRIBUTING. # # Byalokoz EL. New Counting of Time in Russia since July 1, 1919. # This Russian-language source was consulted by Vladimir Karpinsky; see diff --git a/northamerica b/northamerica index bb3d1b43..16b79af2 100644 --- a/northamerica +++ b/northamerica @@ -6,9 +6,9 @@ # also includes Central America and the Caribbean # This file is by no means authoritative; if you think you know better, -# go ahead and edit the file (and please send any changes to -# tz(a)iana.org for general use in the future). For more, please see -# the file CONTRIBUTING in the tz distribution. +# go ahead and edit the file, and please send any changes to +# the public mailing list tz(a)iana.org for general use in the future. +# For more, please see the file CONTRIBUTING in the tz distribution. # From Paul Eggert (1999-03-22): # A reliable and entertaining source about time zones is diff --git a/southamerica b/southamerica index 06fee598..bcd4abdf 100644 --- a/southamerica +++ b/southamerica @@ -4,9 +4,9 @@ # 2009-05-17 by Arthur David Olson. # This file is by no means authoritative; if you think you know better, -# go ahead and edit the file (and please send any changes to -# tz(a)iana.org for general use in the future). For more, please see -# the file CONTRIBUTING in the tz distribution. +# go ahead and edit the file, and please send any changes to +# the public mailing list tz(a)iana.org for general use in the future. +# For more, please see the file CONTRIBUTING in the tz distribution. # From Paul Eggert (2016-12-05): # diff --git a/theory.html b/theory.html index cdc6093a..6e52a929 100644 --- a/theory.html +++ b/theory.html @@ -1584,9 +1584,14 @@ Sources for time on other planets: </section> <footer> - <hr> - This file is in the public domain, so clarified as of 2009-05-17 by - Arthur David Olson. +<hr> +This web page is in the public domain, so clarified as of +2009-05-17 by Arthur David Olson. +<br> +Please send corrections to this web page to the +<a href="mailto:tz@iana.org">time zone mailing list</a>. +The mailing list and its archives are public, +so please do not send confidential information. </footer> </body> </html> diff --git a/tz-art.html b/tz-art.html index 01a00297..3039a86f 100644 --- a/tz-art.html +++ b/tz-art.html @@ -611,13 +611,16 @@ but maybe in your time zone I haven’t finished yet. So stay tuned!” <li><a href="tz-link.html">Time Zone and Daylight Saving Time Data</a></li> </ul> + +<footer> <hr> -<address> This web page is in the public domain, so clarified as of 2009-05-17 by Arthur David Olson. <br> Please send corrections to this web page to the <a href="mailto:tz@iana.org">time zone mailing list</a>. -</address> +The mailing list and its archives are public, +so please do not send confidential information. +</footer> </body> </html> diff --git a/tz-link.html b/tz-link.html index 669a3cb8..9267fb6f 100644 --- a/tz-link.html +++ b/tz-link.html @@ -211,7 +211,10 @@ location.</p> The <code><abbr>tz</abbr></code> code and data are by no means authoritative. If you find errors, please email changes to <a href="mailto:tz@iana.org"><code>tz(a)iana.org</code></a>, -the time zone mailing list. See +the time zone mailing list. +The mailing list and its archives are public, +so please do not send confidential information. +See <a href="https://lists.iana.org/postorius/lists/tz.iana.org/">the mailing list’s main page</a> to subscribe or to browse its archive of old messages. <a href="https://tzdata-meta.timtimeonline.com">Metadata for mailing list @@ -283,8 +286,9 @@ collaboration among governments, the time zone database volunteer community, and data distributors downstream. <p> If your government plans to change its time zone boundaries or -daylight saving rules, please send email to <a -href="mailto:tz@iana.org"><code>tz(a)iana.org</code></a> well in advance, +daylight saving rules, please send email as described in +"<a href="#changes">Changes to the <code><abbr>tz</abbr></code> database</a>". +Do this well in advance, as this will lessen confusion and will coordinate updates to many cell phones, computers, and other devices around the world. In your email, please cite the legislation or regulation that specifies @@ -1250,6 +1254,8 @@ This web page is in the public domain, so clarified as of <br> Please send corrections to this web page to the <a href="mailto:tz@iana.org">time zone mailing list</a>. +The mailing list and its archives are public, +so please do not send confidential information. </footer> </body> </html> -- 2.51.0

1 0

An odd juxtaposition...
by Arthur Olson Dec. 10, 2025

Dec. 10, 2025

...of sex appeal and time zones at the Hyattsville, Maryland Mall at Prince George's: a watch kiosk called "Time Zone 2" by a Victoria's Secret store. An image is part of the social media posting linked to below. @dashdashado https://bsky.app/profile/dashdashado.bsky.social/post/3m7ld53qtj22s

3 2

Greenland parliament to vote February on time zone change
by Paul Eggert Dec. 8, 2025

Dec. 8, 2025

Hanna Hviid reported in Sermitsiaq last month[1] (in Danish) that Inatsisartut, Greenland's parliament, plans to debate in February whether to revert the 2023 change to Greenland's clocks. It seems like there's a good chance something will pass. Although the details are unclear, presumably this would affect clocks starting in fall 2026. I found out about this by running across last month's "Iceland’s Old Timezone Controversy Echoes in Greenland"[2], by the Iceland Review's Michael Chapman. [1]: https://www.sermitsiaq.ag/samfund/pele-beskylder-udvalg-for-at-forhale-over… [2]: https://www.icelandreview.com/news/icelands-old-timezone-controversy-echoes…

2 3

Request for Clarification or Alternative Label for “Indian/Mauritius” Time Zone
by James Lai Dec. 4, 2025

Dec. 4, 2025

Hello, I hope you are well. I am writing regarding the naming of the IANA time zone “Indian/Mauritius.” Many users in Mauritius have reported confusion because the word “Indian” is often interpreted as referring to India rather than the Indian Ocean region. This creates misunderstandings when scheduling international meetings, especially on platforms like Zoom, Google Calendar, and others that rely directly on IANA time zone identifiers. I understand that renaming time zones is avoided due to compatibility concerns. However, I would like to kindly ask whether: 1. An alternative label such as “IndianOcean/Mauritius” could be considered, or 2. A documented alias or clarification could be added to help reduce confusion for end users. Even a non-breaking solution (e.g., an alias) would greatly help clarify the intended meaning without disrupting existing systems. Thank you for your time and for maintaining this important global standard. Kind regards, Kind regards, [cid:asset3_0cde460c-bb22-4017-a4dd-b19716304dc9.png] James Lai IT | Head of Operations Intercontinental Trust Limited Level 3, Alexander House, 35 Cybercity, Ebene 72201, Mauritius E: jlai(a)intercontinentaltrust.com<mailto:jlai@intercontinentaltrust.com> W: www.intercontinentaltrust.com<http://www.intercontinentaltrust.com/> T: (230) 403 0800 DID: (230) 403 0869 M: (230) 5258 5889 F: (230) 403 0801 BRN: C07023546 [cid:itl_emailbanner_03.10.25v2_d5878fa8-0af2-4ccb-b63e-55a10f1a1c87.jpg] Intercontinental Trust Ltd is regulated by the Financial Services Commission in Mauritius. If you are not the intended recipient, please advise the sender immediately and delete this message. [cid:facebooklogo_9db4cba5-e3da-4106-9849-2d5d2e6d37db.gif]<https://www.facebook.com/IntercontinentalTrust?fref=ts> [cid:LinkedInlogo_86f1ad59-c943-4f8d-a70a-699b2322ca7e.gif] <http://www.linkedin.com/company/2502998?trk=pro_other_cmpy> [cid:Twitterlogo_3b4b7685-ec4d-4906-8971-931729a1de68.gif] <https://twitter.com/ITLMauritius> Disclaimer<http://intercontinentaltrust.com/disclaimer> Please consider the environment before you print

8 10

backward file for zonenow.tab?
by kirill.gagarski＠gmail.com Nov. 17, 2025

Nov. 17, 2025

Hi. Our software tries to expose a list of supported timezones to an end user for setting up some periodic scheduled task in the UI. Till now we've tried to populate the list of Java supported timezones, filtering out aliases (from backward file). I noticed that zonedata now ships with zonenow.tab file which is exactly what we need for our UI and meets our needs more precisely. Unfortunately it seems quite impossible to migrate old data to this new notation automatically. For example, hte fact that 'Europe/Berlin' is covered by 'Europe/Paris' in zonenow.tab is not mentioned anywhere. Is there a way to determine this fact from tzdata without deep analysis of the rules done by myself? Other simple question: is there a way to know what is "now" in zonenow? Is it defined by the release date of a new version?

4 6

Calendrical Issues Russia section
by Kirill Gagarski Nov. 14, 2025

Nov. 14, 2025

I fully understand that I am trying to engage into a 30 y.o discussion now, but the Russia section in calendars file repeats quite widespread myths about Eternal Calendar with 30th February. I used to believe in that myself, given quite a little information on the internet about the issue. Even Wikipedia spread this myth back in the day. Even the Russian version of Wikipedia spreaded the myth: https://ru.wikipedia.org/w/index.php?title=%D0%9D%D0%B5%D0%BF%D1%80%D0%B5%D… Nowadays, Wikipedia is more clean about the issue, providing some source links: https://ru.wikipedia.org/wiki/%D0%9D%D0%B5%D0%BF%D1%80%D0%B5%D1%80%D1%8B%D0… https://en.wikipedia.org/wiki/Soviet_calendar The truth is: - Gregorian calendar was continuously used during the mentioned time - 30th February in USSR is an absolute myth (to be more clear, this is a declined proposal of a new calendar) - Mentioned 5-day weeks and 6-day weeks were in fact an attempt to relabel weekdays (Monday-Sunday) - Relabeling was done in a way, not conflicting with traditional weekdays (they used words like "third day of six-week day" or colors). Traditional names were used in parallel I could not say, the calendar file does not say the truth (after all, it raises doubts in a second quoted message), and all this information is indeed very interesting from a historical point of view, but does not seem relevant to Calendrical Issues. In fact 100 years later no one (unless someone is specifically interested in these weird weekdays) cares if 12th of December of 1937 was "sixth day of six-day week", because it was Sunday at the same time. I'd suggest removing the whole Russia section from this file (you can keep the small note about switching from Julia to Gregorian calendar in 1918), since for the purpose of it it raises nothing but confusion.

2 2

[PROPOSED] Fix unlikely backslash bug in scripts
by Paul Eggert Nov. 12, 2025

Nov. 12, 2025

* tzselect.ksh, workman.sh: Use printf, not echo, if the argument might in theory contain a backslash. --- tzselect.ksh | 9 +++++---- workman.sh | 2 +- 2 files changed, 6 insertions(+), 5 deletions(-) diff --git a/tzselect.ksh b/tzselect.ksh index 7cd8679d..85a8c670 100644 --- a/tzselect.ksh +++ b/tzselect.ksh @@ -145,9 +145,11 @@ do t*) # Undocumented option, used for developer testing. zonetabtype=$OPTARG;; -help) - exec echo "$usage";; + say "$usage" + exit;; -version) - exec echo "tzselect $PKGVERSION$TZVERSION";; + say "tzselect $PKGVERSION$TZVERSION" + exit;; -*) say >&2 "$0: -$opt$OPTARG: unknown option; try '$0 --help'"; exit 1;; *) @@ -516,8 +518,7 @@ while ' ="$distance_table" ) echo >&2 'Please select one of the following timezones,' - echo >&2 'listed roughly in increasing order' \ - "of distance from $coord". + say >&2 "listed roughly in increasing order of distance from $coord." doselect $regions region=$select_result tz=$( diff --git a/workman.sh b/workman.sh index 91621b10..5da55654 100644 --- a/workman.sh +++ b/workman.sh @@ -51,6 +51,6 @@ elif (type nroff && type perl) >/dev/null 2>&1; then } ' else - echo >&2 "$0: please install groff, or mandoc and col" + printf >&2 '%s\n' "$0: please install groff, or mandoc and col" exit 1 fi -- 2.51.0

1 0

Melbourne's Shrine of Remembrance
by Tim Parenti Nov. 11, 2025

Nov. 11, 2025

Earlier this year, Julian O'Shea released a short video detailing a peculiarity about the Shrine of Remembrance in Melbourne, "How Daylight Savings Broke this $24 Million Building": https://www.youtube.com/watch?v=mfzsBMUiGGQ The shrine, originally dedicated in 1934 to honor those who served in World War I, was designed with a small aperture in the roof. Each Remembrance Day (11 November) at 11:00am local time, the position of the sun would be perfectly aligned to shine through the aperture, allowing the light of its rays to fall upon an inscribed stone inside, thus commemorating the hour of the Armistice. The problem? Several decades later, in 1971, Victoria adopted DST, but of course the aperture had been tuned to the position of the sun at 11:00am AEST. So what do you do to preserve this meaningful element of an annual ceremony with important dignitaries, when it's effectively been moved an hour earlier to 11:00am AEDT instead? The answer involves a clever system of mirrors. (And the general public still gets to witness the light following its true path an hour later!) Since this is quite a grand physical representation of how we interact with our timekeeping, I've added this to tz-art.html, as well as a few fixups to unclosed HTML tags nearby. -- Tim Parenti

4 3

Cape Town
by Justin Singh Nov. 11, 2025

Nov. 11, 2025

Hello The Timezone for Cape Town is UTC+01:00 The western Northern Cape and Western Cape differ, however. Everywhere on land west of 22°30′ E effectively experiences year-round daylight saving time because of its location in true UTC+01:00 but still being in South African Standard Time. Sunrise and sunset are thus relatively late in Cape Town, compared to the rest of the country.

6 7

Nunavik to stop changing its clocks after March 2026?
by Paul Eggert Nov. 8, 2025

Nov. 8, 2025

Yesterday’s CBC News report “Nunavimmiut have wound back their clocks for the last time”[1] is so light on details that for now I am merely attaching a draft patch to this email for review, rather than installing the patch into the development repository. I found no announcement of the change at www.makivvik.ca. Although I asked for more details by using the form at <https://www.makivvik.ca/contact/>, these sorts of contact forms often do not work, so if anyone has access to more-authoritative info than the CBC report, please let us know. Among other things, details in the news report make me wonder how likely Makivvik would reconsider its decision before the change would take effect about a year from now. [1]: https://www.cbc.ca/news/canada/north/nunavik-daylight-savings-9.6969650

1 0

Phoenix AZ timezone issue
by akbar＠marketing-qa.com Nov. 5, 2025

Nov. 5, 2025

🧭 Problem Summary Arizona remains one of the few U.S. states that does not observe Daylight Saving Time (DST) since 1968. However, it identifies its time zone as Mountain Standard Time (MST) — which does change to MDT when DST applies elsewhere. Because of this, every spring and fall, Arizona’s effective time alignment shifts: In summer, Arizona matches Pacific Daylight Time (PDT). In winter, it matches Mountain Standard Time (MST). This causes systemic issues in digital calendars, automated scheduling, flight bookings, and cross-state business coordination, requiring frequent manual adjustments. 💡 Recommended Solution Create and declare a dedicated time zone: “Phoenix Standard Time.” This would be a fixed, non-DST time zone officially recognized by IANA (the Internet Assigned Numbers Authority) and major platforms (Microsoft, Apple, Google). All digital devices could use it as a unique time zone identifier that never shifts with DST — eliminating confusion and rescheduling errors. It would function similarly to “UTC+7” year-round but would display as “Phoenix Standard Time (PST)” in software systems. Thank you Akbar Jaffer Akbar Jaffer Office: (415) 335-6950 Mobile: (650) 430-0232 Web: Marketing-QA.com <http://www.marketing-qa.com/> Meeting: Book time W/Akbar <https://calendly.com/ajaffer/> LinkedIn <https://www.linkedin.com/in/akbarjaffer/> | POI Blog <https://pointofintersection.org/>

8 9

How to select zones for a TimeZone picker
by Guillermo Rodriguez Garcia Nov. 3, 2025

Nov. 3, 2025

Hi all, I'm building a TimeZone picker with the following constraints: 1. At least one distinct timezone per country (so that e.g. there is an entry for Atlantic/Reykjavik instead of mapping IS to Africa/Abidjan) 2. Within a country, do not show timezones that currently have identical rules to another timezone in the same country (i.e. historical differences don't matter). Because of (1) I started from zone.tab rather than zone1970.tab. However, I'm not sure how to handle (2). For example, Spain has three zones in zone.tab: Europe/Madrid, Africa/Ceuta, and Atlantic/Canary. But as far as most users are concerned, as of today there are only two timezones in Spain: Madrid and Canary Islands. Ceuta follows Madrid rules, so there should be no need to show this as an option to the user. Another example: the U.S. lists 29 zones, but only 8 are distinct today (ignoring historical changes). So, the question: Any suggestions on how to filter timezones within each country to drop those that as of today are not relevant except for historical differences? Thank you, Guillermo Rodriguez Garcia guille.rodriguez(a)gmail.com

8 17

[PROPOSED] tzfree now preserves errno
by Paul Eggert Nov. 3, 2025

Nov. 3, 2025

* Makefile, NEWS, newtzset.3: Mention this. * localtime.c (FREE_PRESERVES_ERRNO): Default to 1 if 'free' is known to preserve errno, 0 otherwise. (tzfree): Preserve errno if 'free' does not. --- Makefile | 2 ++ NEWS | 2 ++ localtime.c | 15 +++++++++++++++ newtzset.3 | 4 ++++ 4 files changed, 23 insertions(+) diff --git a/Makefile b/Makefile index 0fe8a9f8..0d33fa17 100644 --- a/Makefile +++ b/Makefile @@ -239,6 +239,8 @@ LDLIBS= # -DEPOCH_OFFSET=N if the 'time' function returns a value N greater # than what POSIX specifies, assuming local time is UT. # For example, N is 252460800 on AmigaOS. +# -DFREE_PRESERVES_ERRNO=[01] if the 'free' function munges or preserves errno +# (default is guessed) # -DHAVE_DECL_ASCTIME_R=0 if <time.h> does not declare asctime_r # on POSIX platforms predating POSIX.1-2024 # -DHAVE_DECL_ENVIRON if <unistd.h> declares 'environ' diff --git a/NEWS b/NEWS index 1ed9b2eb..a9b2d321 100644 --- a/NEWS +++ b/NEWS @@ -93,6 +93,8 @@ Unreleased, experimental changes option is incompatible with POSIX.1-2024 and earlier. It also costs CPU time and memory. + tzfree now preserves errno, consistently with POSIX.1-2024 'free'. + tzcode now uses mempcpy if available, guessing its availability. Compile with -DHAVE_MEMPCPY=1 or 0 to override the guess. diff --git a/localtime.c b/localtime.c index b85941f2..46acf3bd 100644 --- a/localtime.c +++ b/localtime.c @@ -2061,10 +2061,25 @@ tzalloc(char const *name) return sp; } +#ifndef FREE_PRESERVES_ERRNO +# if ((defined _POSIX_VERSION && 202405 <= _POSIX_VERSION) \ + || (defined __GLIBC__ && 2 < __GLIBC__ + (33 <= __GLIBC_MINOR__)) \ + || defined __OpenBSD__ || defined __sun) +# define FREE_PRESERVES_ERRNO 1 +# else +# define FREE_PRESERVES_ERRNO 0 +# endif +#endif + void tzfree(timezone_t sp) { + int err; + if (!FREE_PRESERVES_ERRNO) + err = errno; free(sp); + if (!FREE_PRESERVES_ERRNO) + errno = err; } /* diff --git a/newtzset.3 b/newtzset.3 index 3801ab1b..028cfd25 100644 --- a/newtzset.3 +++ b/newtzset.3 @@ -379,6 +379,10 @@ If successful, the function returns a nonnull pointer to the newly allocated object. Otherwise, it returns a null pointer and sets .IR errno . +The +.B tzfree +function does not modify +.IR errno . .SH ERRORS .TP .B EOVERFLOW -- 2.51.0

1 0

[PROPOSED] Improve TZ="/..." security and speed (NetBSD only)
by Paul Eggert Oct. 28, 2025

Oct. 28, 2025

* NEWS: Mention security. * localtime.c (O_REGULAR): Default to 0. (tzloadbody): Open TZif files with O_REGULAR. Suppress unnecessary call to 'stat' if O_REGULAR is present. --- NEWS | 3 +++ localtime.c | 7 +++++-- 2 files changed, 8 insertions(+), 2 deletions(-) diff --git a/NEWS b/NEWS index 64d32bf3..1ed9b2eb 100644 --- a/NEWS +++ b/NEWS @@ -40,6 +40,9 @@ Unreleased, experimental changes and reject relative names containing ".." directory components; formerly, only privileged programs did those two things. These changes were inspired by similar behavior in FreeBSD. + On NetBSD, unprivileged programs now use O_REGULAR to check + whether a TZ setting starting with '/' names a regular file, + avoiding a minor security race still present elsewhere. TZ strings taken from tzalloc arguments are now treated with no less caution than TZ strings taken from the environment, as the old undocumented behavior would have been hard to explain. diff --git a/localtime.c b/localtime.c index 0d2d2a93..b85941f2 100644 --- a/localtime.c +++ b/localtime.c @@ -355,6 +355,9 @@ static int openat(int dd, char const *path, int oflag) { unreachable (); } #ifndef O_PATH # define O_PATH 0 #endif +#ifndef O_REGULAR +# define O_REGULAR 0 +#endif #ifndef O_RESOLVE_BENEATH # define O_RESOLVE_BENEATH 0 #endif @@ -900,7 +903,7 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, register int tzheadsize = sizeof(struct tzhead); int dd = AT_FDCWD; int oflags = (O_RDONLY | O_BINARY | O_CLOEXEC | O_CLOFORK - | O_IGNORE_CTTY | O_NOCTTY); + | O_IGNORE_CTTY | O_NOCTTY | O_REGULAR); int err; struct stat st; st.st_ctime = 0; @@ -927,7 +930,7 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, continue; else if (issetugid()) return ENOTCAPABLE; - else { + else if (!O_REGULAR) { /* Check for devices, as their mere opening could have unwanted side effects. Though racy, there is no portable way to fix the races. This check is needed -- 2.48.1

1 0

Iana build failures on Windows
by Manuela Friedrich Oct. 28, 2025

Oct. 28, 2025

Hello Paul, I have hit a compile error for zic.c on Windows. Git commit 9e0bf0b42d68b3d9ad0bc7ff70fc92d87f8b40d2 zic.c zic.c(152): error C2081: 'mode_t': name in formal parameter list illegal zic.c(2437): warning C4116: unnamed type definition in parentheses NMAKE : U1077: '"C:\Program Files (x86)\Microsoft Visual Studio\2017\Professional\VC\Tools\MSVC\14.12.25827\bin\HostX64\x64\cl.EXE"' : return code '0x2' Stop. Can you please advise? Please do not release a new version with this issue present. Thank you! Regards Manuela Friedrich

4 17

[PROPOSED] Be more like OpenBSD wrt ".."
by Paul Eggert Oct. 25, 2025

Oct. 25, 2025

* localtime.c (tzloadbody): Do not worry about settings like TZ=".." (i.e., trailing ".." component) as the later read will fail anyway. This simplifies the code slightly, and agrees with the security-conscious OpenBSD libc. --- localtime.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/localtime.c b/localtime.c index f1a82370..0d2d2a93 100644 --- a/localtime.c +++ b/localtime.c @@ -941,13 +941,13 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, if (relname[0] != '/') { if (!OPENAT_TZDIR || !O_RESOLVE_BENEATH) { - /* Fail if a relative name contains a ".." component, - as such a name could read a file outside TZDIR + /* Fail if a relative name contains a non-terminal ".." component, + as such a name could read a non-directory outside TZDIR when AT_FDCWD and O_RESOLVE_BENEATH are not available. */ char const *component; for (component = relname; component[0]; component++) if (component[0] == '.' && component[1] == '.' - && ((component[2] == '/') | !component[2]) + && component[2] == '/' && (component == relname || component[-1] == '/')) return ENOTCAPABLE; } -- 2.48.1

1 0

[PATCH 1/5] Clarify <sys/auxv.h> vs getauxval
by Paul Eggert Oct. 25, 2025

Oct. 25, 2025

FreeBSD has <sys/auxv.h> but not getauxval, so clarify ifdeffery to reflect this. This doesn’t fix any bugs (since FreeBSD also has issetugid) but makes things clearer. * Makefile, NEWS: Update to match new behavior and fix a few old typos. * localtime.c: Move the <sys/auv.h> code to be closer together. (HAVE_SYS_AUXV_H) [!HAVE_ISSETUGID]: New macro. * private.h (HAVE_GETAUXVAL): Remove. All uses replaced by HAVE_SYS_AUXV_H. --- Makefile | 12 ++++++------ NEWS | 9 +++++---- localtime.c | 25 +++++++++++++++++++------ private.h | 13 ------------- 4 files changed, 30 insertions(+), 29 deletions(-) diff --git a/Makefile b/Makefile index 3b259c80..0fe8a9f8 100644 --- a/Makefile +++ b/Makefile @@ -246,7 +246,6 @@ LDLIBS= # -DHAVE_DIRECT_H if mkdir needs <direct.h> (MS-Windows) # -DHAVE_FCHMOD=0 if your system lacks the fchmod function # -DHAVE__GENERIC=0 if _Generic does not work* -# -DHAVE_GETAUXVAL=1 if getauxval works, 0 otherwise (default is guessed) # -DHAVE_GETEUID=0 if gete?[ug]id do not work # -DHAVE_GETRANDOM if getrandom works (e.g., GNU/Linux), # -DHAVE_GETRANDOM=0 to avoid using getrandom @@ -258,7 +257,9 @@ LDLIBS= # ctime_r and asctime_r incompatibly with POSIX.1-2017 and earlier # (Solaris when _POSIX_PTHREAD_SEMANTICS is not defined). # -DHAVE_INTTYPES_H=0 if <inttypes.h> does not work*+ -# -DHAVE_ISSETUID=1 if issetugid works, 0 otherwise (default is guessed) +# -DHAVE_ISSETUGID=1 if issetugid works, 0 otherwise (default is guessed) +# If 0, you may also use -DHAVE_SYS_AUXV_H=1 if <sys/auxv.h> works, +# 0 otherwise (default is guessed). # -DHAVE_LINK=0 if your system lacks a link function # -DHAVE_LOCALTIME_R=0 if your system lacks a localtime_r function # -DHAVE_LOCALTIME_RZ=0 if you do not want zdump to use localtime_rz @@ -268,8 +269,8 @@ LDLIBS= # -DHAVE_POSIX_DECLS=0 if your system's include files do not declare # variables like 'tzname' required by POSIX # -DHAVE_PWD_H=0 if your system lacks pwd.h, grp.h and corresponding functions -# The following additional options may be needed: -# -Dgid_t=T, -Duid_t=T to define gid_t, uid_t to be type T +# If 0, you may also need -Dgid_t=G -Duid_t=U +# to define gid_t and uid_t to be types G and U. # -DHAVE_SETENV=0 if your system lacks the setenv function # -DHAVE_SETMODE=[01] if your system lacks or has the setmode and getmode # functions (default is guessed) @@ -285,8 +286,7 @@ LDLIBS= # -DHAVE_STRUCT_TIMESPEC=0 if your system lacks struct timespec+ # -DHAVE_SYMLINK=0 if your system lacks the symlink function # -DHAVE_SYS_STAT_H=0 if <sys/stat.h> does not work* -# The following additional option may be needed: -# -Dmode_t=T to define mode_t to be type T +# If 0, you may also need -Dmode_t=M to define mode_t to be type M. # -DHAVE_TZSET=0 if your system lacks a tzset function # -DHAVE_UNISTD_H=0 if <unistd.h> does not work* # -DHAVE_UTMPX_H=0 if <utmpx.h> does not work* diff --git a/NEWS b/NEWS index 37ac9fba..64d32bf3 100644 --- a/NEWS +++ b/NEWS @@ -44,12 +44,13 @@ Unreleased, experimental changes no less caution than TZ strings taken from the environment, as the old undocumented behavior would have been hard to explain. tzset etc. no longer use the 'access' system call to check access; - instead they now use the system calls issetuid, getauxval, + instead they now use the system calls issetugid, getauxval, getresuid/getresgid, and geteuid/getegid/getuid/getgid (whichever first works) to test whether a program is privileged. - Compile with -DHAVE_ISSETUID=[01], -DHAVE_GETAUXVAL=[01], - -DHAVE_GETRESUID=[01], and -DHAVE_GETEUID=[01] to enable or - disable these system calls' use. + Compile with -DHAVE_SYS_AUXV_H=[01] to enable or disable + <sys/auxv.h> which (if it defines AT_SECURE) enables getauxval, + and compile with -DHAVE_ISSETUGID=[01], -DHAVE_GETRESUID=[01], and + -DHAVE_GETEUID=[01] to enable or disable the other calls' use. The new CFLAGS option -DTZ_CHANGE_INTERVAL=N makes tzset etc. check for TZif file changes if the in-memory data are N seconds diff --git a/localtime.c b/localtime.c index a8c442dd..f2d35813 100644 --- a/localtime.c +++ b/localtime.c @@ -19,10 +19,6 @@ #include "tzfile.h" #include <fcntl.h> -#if HAVE_GETAUXVAL && !HAVE_ISSETUGID -# include <sys/auxv.h> -#endif - #if HAVE_SYS_STAT_H # include <sys/stat.h> # ifndef S_ISREG @@ -366,12 +362,29 @@ static int openat(int dd, char const *path, int oflag) { unreachable (); } # define O_SEARCH 0 #endif -/* Return 1 if the process is privileged, 0 otherwise. */ #if !HAVE_ISSETUGID + +# if !defined HAVE_SYS_AUXV_H && defined __has_include +# if __has_include(<sys/auxv.h>) +# define HAVE_SYS_AUXV_H 1 +# endif +# endif +# ifndef HAVE_SYS_AUXV_H +# if defined __GLIBC__ && 2 < __GLIBC__ + (19 <= __GLIBC_MINOR__) +# define HAVE_SYS_AUXV_H 1 +# else +# define HAVE_SYS_AUXV_H 0 +# endif +# endif +# if HAVE_SYS_AUXV_H +# include <sys/auxv.h> +# endif + +/* Return 1 if the process is privileged, 0 otherwise. */ static int issetugid(void) { -# if HAVE_GETAUXVAL && defined AT_SECURE +# if HAVE_SYS_AUXV_H && defined AT_SECURE unsigned long val; errno = 0; val = getauxval(AT_SECURE); diff --git a/private.h b/private.h index 8d3d6404..4daa4864 100644 --- a/private.h +++ b/private.h @@ -294,19 +294,6 @@ extern int optind; # endif #endif -#if !defined HAVE_GETAUXVAL && defined __has_include -# if __has_include(<sys/auxv.h>) -# define HAVE_GETAUXVAL 1 -# endif -#endif -#ifndef HAVE_GETAUXVAL -# if defined __GLIBC__ && 2 < __GLIBC__ + (19 <= __GLIBC_MINOR__) -# define HAVE_GETAUXVAL 1 -# else -# define HAVE_GETAUXVAL 0 -# endif -#endif - #ifndef HAVE_ISSETUGID # if (defined __FreeBSD__ || defined __NetBSD__ || defined __OpenBSD__ \ || (defined __linux__ && !defined __GLIBC__) /* Android, musl, etc. */ \ -- 2.51.0

1 4

[PROPOSED] Shorten GCC_INSTRUMENT à la GCC 14
by Paul Eggert Oct. 25, 2025

Oct. 25, 2025

* Makefile (GCC_INSTRUMENT): Switch from -fsanitize-undefined-trap-on-error to its shorter alias -fsanitize-trap=all, introduced in GCC 14. --- Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Makefile b/Makefile index f57042df..3b259c80 100644 --- a/Makefile +++ b/Makefile @@ -380,7 +380,7 @@ LDLIBS= # Select instrumentation via "make GCC_INSTRUMENT='whatever'". GCC_INSTRUMENT = \ -fsanitize=undefined -fsanitize-address-use-after-scope \ - -fsanitize-undefined-trap-on-error -fstack-protector + -fsanitize-trap=all -fstack-protector # Omit -fanalyzer from GCC_DEBUG_FLAGS, as it makes GCC too slow. GCC_DEBUG_FLAGS = -DGCC_LINT -g3 -O3 \ $(GCC_INSTRUMENT) \ -- 2.51.0

1 0

[PROPOSED 0/3] zic options -D, -m, and -u from FreeBSD
by Paul Eggert Oct. 24, 2025

Oct. 24, 2025

This thread contains proposed patches to add support for the zic options -D, -m, and -u, which have been in FreeBSD zic for many years. The implementation differs from that of FreeBSD, to fix some unlikely security races. I installed these patches in the development repository on GitHub. Paul Eggert (3): New zic option -D, inspired by FreeBSD New zic option -m, inspired by FreeBSD New zic option -u, inspired by FreeBSD Makefile | 8 ++ NEWS | 2 + zic.8 | 32 +++++++ zic.c | 259 +++++++++++++++++++++++++++++++++++++++++++++++++++++-- 4 files changed, 293 insertions(+), 8 deletions(-) -- 2.48.1

2 5

Possible timezone change in Argentina (starting in April?)
by Mariano Absatz - el Baby Oct. 24, 2025

Oct. 24, 2025

Hi, back here after 12+ years for a heads up... There is a Law proposal approved by the Cámara de Diputados (equivalent of the House) but yet to be approved in the Senate regarding Argentina's timezone. Argentina is currently in UTC-3 with no DST. The proposal establishes the Official Time for the whole country as UTC-4 (which is geographically more accurate), with a DST of UTC-3. The DST would start at 00:00 on the first April Sunday and would end at 00:00 on the first September Sunday. The Executive is authorized to modify these dates and to coordinate with other countries of MERCOSUR to apply changes on the same moment. This is just a heads up, since the Senate did not approve or modify the proposal. If this is approved and enacted before April 2026, we can assume the current status is DST and at 00:00 Sunday April 5th 2026 we would be moving the clock back to 23:00 Saturday April 4th. I'll keep you posted as this evolves, but don't hold your breath since on October 26th there are Legislative Elections (midterm), and half of the House and one third of the Senate will change. The new legislators will take office on December 10th. By the end of December the ordinary session of Congress ends; it may be extended by the Executive Branch, but there is no certainty about which bills can be discussed. References (all in Spanish): * Law proposal (official Cámara de Diputados site): https://www.hcdn.gob.ar/diputados/jcobos/proyecto.html?exp=1110-D-2024 (click on the "Ver documento original" button) * Status in the Senate: https://www.senado.gob.ar/parlamentario/comisiones/verExp/14.25/CD/PL (click on the "Texto Original" button for the signed proposal sent from the Cámara de Diputados; click on the "Trámite Legislativo" button for the process status in the Senate) * Brief article about the proposal on a reputed fact checking site: https://chequeado.com/el-explicador/cambio-de-huso-horario-en-la-argentina-… -- Mariano Absatz - el Baby www.clueless.ar

3 3

[PROPOSED 1/2] Refactor duplicate duplicate-option code
by Paul Eggert Oct. 24, 2025

Oct. 24, 2025

* zic.c (duplicate_options): New function. (main): Use it. --- zic.c | 73 +++++++++++++++++++---------------------------------------- 1 file changed, 23 insertions(+), 50 deletions(-) diff --git a/zic.c b/zic.c index 67bcaa43..34e7d91c 100644 --- a/zic.c +++ b/zic.c @@ -708,6 +708,13 @@ close_file(FILE *stream, char const *dir, char const *name, } } +ATTRIBUTE_NORETURN static void +duplicate_options(char const *opt) +{ + fprintf(stderr, _("%s: More than one %s option specified\n"), progname, opt); + exit(EXIT_FAILURE); +} + ATTRIBUTE_NORETURN static void usage(FILE *stream, int status) { @@ -1063,73 +1070,39 @@ main(int argc, char **argv) error(_("invalid option: -b '%s'"), optarg); break; case 'd': - if (directory == NULL) - directory = optarg; - else { - fprintf(stderr, - _("%s: More than one -d option" - " specified\n"), - progname); - return EXIT_FAILURE; - } + if (directory) + duplicate_options("-d"); + directory = optarg; break; case 'l': - if (lcltime == NULL) - lcltime = optarg; - else { - fprintf(stderr, - _("%s: More than one -l option" - " specified\n"), - progname); - return EXIT_FAILURE; - } + if (lcltime) + duplicate_options("-l"); + lcltime = optarg; break; case 'p': - if (psxrules == NULL) - psxrules = optarg; - else { - fprintf(stderr, - _("%s: More than one -p option" - " specified\n"), - progname); - return EXIT_FAILURE; - } + if (psxrules) + duplicate_options("-p"); + psxrules = optarg; break; case 't': - if (tzdefault != NULL) { - fprintf(stderr, - _("%s: More than one -t option" - " specified\n"), - progname); - return EXIT_FAILURE; - } + if (tzdefault) + duplicate_options("-t"); tzdefault = optarg; break; case 'y': warning(_("-y ignored")); break; case 'L': - if (leapsec == NULL) - leapsec = optarg; - else { - fprintf(stderr, - _("%s: More than one -L option" - " specified\n"), - progname); - return EXIT_FAILURE; - } + if (leapsec) + duplicate_options("-L"); + leapsec = optarg; break; case 'v': noise = true; break; case 'r': - if (timerange_given) { - fprintf(stderr, - _("%s: More than one -r option" - " specified\n"), - progname); - return EXIT_FAILURE; - } + if (timerange_given) + duplicate_options("-r"); if (! timerange_option(optarg)) { fprintf(stderr, _("%s: invalid time range: %s\n"), -- 2.48.1

1 1

South Africa Daylight Saving Time
by Saras Sing Oct. 21, 2025

Oct. 21, 2025

https://www.timeanddate.com/time/change/south-africa?year=1942 https://www.timeanddate.com/time/change/south-africa?year=1944 https://www.timeanddate.com/time/change/south-africa Hello South Africa has Daylight Saving Time every year from September to March South African Daylight Saving Time starts at every first Sunday of September from GMT/UTC+02:00 00:00 to 01:00 to GMT/UTC+03:00 South African Daylight Saving Time ends at every last Sunday of March from GMT/UTC+03:00 01:00 to 00:00 back to South Africa Standard Time GMT/UTC+02:00 Thank you

5 4

[PROPOSED 1/3] zic now uses is_digit
by Paul Eggert Oct. 18, 2025

Oct. 18, 2025

Refactor zic.c to use is_digit like others use. * zic.c (is_digit): New static function, copied from localtime.c and from zdump.c. It’s too bad we can’t avoid duplication by putting this into private.h, because gcc -Wunused-function would then complain about it. Perhaps ‘inline’ would pacify gcc but then we’d have to port ‘inline’ to ancient compilers and anyway ‘inline’ is supposed to be just a performance hint which is not needed here. --- zic.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) diff --git a/zic.c b/zic.c index 39e1e2bc..30426621 100644 --- a/zic.c +++ b/zic.c @@ -229,6 +229,13 @@ static bool rulesub(struct rule * rp, const char * dayp, const char * timep); static zic_t tadd(zic_t t1, zic_t t2); +/* Is C an ASCII digit? */ +static bool +is_digit(char c) +{ + return '0' <= c && c <= '9'; +} + /* Bound on length of what %z can expand to. */ enum { PERCENT_Z_LEN_BOUND = sizeof "+995959" - 1 }; @@ -1852,7 +1859,7 @@ gethms(char const *string, char const *errstring) &hh, &hhx, &mm, &mmx, &ss, &ssx, &tenths, &xr, &xs)) { default: ok = false; break; case 8: - ok = '0' <= xr && xr <= '9'; + ok = is_digit(xr); ATTRIBUTE_FALLTHROUGH; case 7: ok &= ssx == '.'; @@ -3952,7 +3959,7 @@ newabbr(const char *string) cp = string; mp = NULL; - while (is_alpha(*cp) || ('0' <= *cp && *cp <= '9') + while (is_alpha(*cp) || is_digit(*cp) || *cp == '-' || *cp == '+') ++cp; if (noise && cp - string < 3) -- 2.51.0

1 2

[PROPOSED] New compile-time option THREAD_TM_MULTI
by Paul Eggert Oct. 14, 2025

Oct. 14, 2025

* Makefile, NEWS: Mention it. * localtime.c (THREAD_TM_MULTI): Default to 0. The remaining changes affect only the THREAD_TM_MULTI case. (enum tm_multi): New enum. (N_TM_MULTI): New constant. (tm_multi_key, tm_multi_key_err): New static vars. (tm_multi): New static function. (localtime, gmtime, offtime): Use it. --- Makefile | 8 ++++++++ NEWS | 9 +++++++++ localtime.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++++++--- 3 files changed, 70 insertions(+), 3 deletions(-) diff --git a/Makefile b/Makefile index 8939869a..b172542a 100644 --- a/Makefile +++ b/Makefile @@ -316,6 +316,14 @@ LDLIBS= # This can improve paralellism and thus save real time # if many threads call tzcode functions simultaneously. # It also costs CPU time and thus energy. +# -DTHREAD_TM_MULTI to have gmtime, localtime, and offtime +# return different struct tm * addresses in different threads. +# This supports unportable programs that call +# gmtime/localtime/offtime when they should call +# gmtime_r/localtime_r/offtime_r to avoid races. +# Because the corresponding storage is freed on thread exit, +# this option is incompatible with POSIX.1-2024 and earlier. +# It also costs CPU time and memory. # -Dtime_tz=\"T\" to use T as the time_t type, rather than the system time_t # This is intended for internal use only; it mangles external names. # -DTZ_CHANGE_INTERVAL=N if functions depending on TZ should check diff --git a/NEWS b/NEWS index 5bfb7bb4..46bf48ed 100644 --- a/NEWS +++ b/NEWS @@ -80,6 +80,15 @@ Unreleased, experimental changes rarely changing and many threads call tzcode simultaneously. It costs more CPU time and energy. + The new CFLAGS option -TTHREAD_TM_MULTI causes localtime to return + a pointer to thread-specific memory, as FreeBSD does, instead of + to the same memory in all threads. This supports unportable + programs that incorrectly use localtime instead of localtime_r. + This option affects gmtime and offtime similarly to localtime. + Because the corresponding storage is freed on thread exit, this + option is incompatible with POSIX.1-2024 and earlier. It also + costs CPU time and memory. + tzcode now uses mempcpy if available, guessing its availability. Compile with -DHAVE_MEMPCPY=1 or 0 to override the guess. diff --git a/localtime.c b/localtime.c index 1191f810..709b954c 100644 --- a/localtime.c +++ b/localtime.c @@ -52,6 +52,10 @@ struct stat { char st_ctime, st_dev, st_ino; } # define THREAD_RWLOCK 0 #endif +#ifndef THREAD_TM_MULTI +# define THREAD_TM_MULTI 0 +#endif + #if THREAD_SAFE # include <pthread.h> @@ -171,6 +175,52 @@ once(once_t *once_control, void (*init_routine)(void)) #endif } +enum tm_multi { LOCALTIME_TM_MULTI, GMTIME_TM_MULTI, OFFTIME_TM_MULTI }; + +#if THREAD_SAFE && THREAD_TM_MULTI + +enum { N_TM_MULTI = OFFTIME_TM_MULTI + 1 }; +static pthread_key_t tm_multi_key; +static int tm_multi_key_err; + +static void +tm_multi_key_init(void) +{ + tm_multi_key_err = pthread_key_create(&tm_multi_key, free); +} + +#endif + +/* Return TMP, or a thread-specific struct tm * selected by WHICH. */ +static struct tm * +tm_multi(struct tm *tmp, ATTRIBUTE_MAYBE_UNUSED enum tm_multi which) +{ +#if THREAD_SAFE && THREAD_TM_MULTI + /* It is OK to check is_threaded() separately here; even if it + returns a different value in other places in the caller, + this function's behavior is still valid. */ + if (is_threaded()) { + /* Try to get a thread-specific struct tm *. + Fall back on TMP if this fails. */ + static pthread_once_t tm_multi_once = PTHREAD_ONCE_INIT; + pthread_once(&tm_multi_once, tm_multi_key_init); + if (!tm_multi_key_err) { + struct tm *p = pthread_getspecific(tm_multi_key); + if (!p) { + p = malloc(N_TM_MULTI * sizeof *p); + if (p && pthread_setspecific(tm_multi_key, p) != 0) { + free(p); + p = NULL; + } + } + if (p) + return &p[which]; + } + } +#endif + return tmp; +} + /* Unless intptr_t is missing, pacify gcc -Wcast-qual on char const * exprs. Use this carefully, as the casts disable type checking. This is a macro so that it can be used in static initializers. */ @@ -2156,7 +2206,7 @@ localtime(const time_t *timep) # if !SUPPORT_C89 static struct tm tm; # endif - return localtime_tzset(timep, &tm, true); + return localtime_tzset(timep, tm_multi(&tm, LOCALTIME_TM_MULTI), true); } struct tm * @@ -2208,7 +2258,7 @@ gmtime(const time_t *timep) # if !SUPPORT_C89 static struct tm tm; # endif - return gmtime_r(timep, &tm); + return gmtime_r(timep, tm_multi(&tm, GMTIME_TM_MULTI)); } # if STD_INSPIRED @@ -2229,7 +2279,7 @@ offtime(time_t const *timep, long offset) # if !SUPPORT_C89 static struct tm tm; # endif - return offtime_r(timep, offset, &tm); + return offtime_r(timep, offset, tm_multi(&tm, OFFTIME_TM_MULTI)); } # endif -- 2.48.1

1 0

[PROPOSED 1/3] Define fullname only if needed
by Paul Eggert Oct. 14, 2025

Oct. 14, 2025

* localtime.c (union local_storage): Define fullname member only if needed. This saves space on unlikely platforms where sizeof (struct file_analysis) < PATH_MAX, and is closer to FreeBSD. (tzloadbody): Mention fullname only if declared. --- localtime.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/localtime.c b/localtime.c index a78962a6..c9836b2d 100644 --- a/localtime.c +++ b/localtime.c @@ -771,7 +771,7 @@ union local_storage { struct state st; } u; -#ifdef PATH_MAX +#if defined PATH_MAX && !OPENAT_TZDIR && !SUPPRESS_TZDIR /* The name of the file to be opened. */ char fullname[PATH_MAX]; #endif @@ -890,7 +890,7 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, cp = mempcpy(lsp, tzdirslash, tzdirslashlen); cp = mempcpy(cp, name, namelen); *cp = '\0'; -#ifdef PATH_MAX +#if defined PATH_MAX && !OPENAT_TZDIR && !SUPPRESS_TZDIR name = lsp->fullname; #else name = (char *) lsp; -- 2.48.1

1 2