- tz - lists.iana.org

British Summer Time end error
by Saras Sing Oct. 1, 2025

Oct. 1, 2025

Hello The United Kingdom British Summer Time ends at every last Sunday of November Thank you

7 6

[PROPOSED] Restore SUPPRESS_TZDIR support if OPENAT_TZDIR
by Paul Eggert Sept. 30, 2025

Sept. 30, 2025

The 09-28 changes inadvertently removed some support for the rarely-used SUPPRESS_TZDIR option. * localtime.c (tzloadbody) [OPENAT_TZDIR && SUPPRESS_TZDIR]: Do not open TZDIR, as when SUPPRESS_TZDIR we want relative pathnames to be relative to the working directory, not to TZDIR. --- localtime.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/localtime.c b/localtime.c index 3937e5dd..186c7bf3 100644 --- a/localtime.c +++ b/localtime.c @@ -667,7 +667,7 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, return ENOTCAPABLE; } - if (OPENAT_TZDIR) { + if (OPENAT_TZDIR && !SUPPRESS_TZDIR) { /* Prefer O_SEARCH or O_PATH if available; O_RDONLY should be OK too, as TZDIR is invariably readable. O_DIRECTORY should be redundant but might help -- 2.48.1

1 0

Use -00 only for invalid time zones
by Dag-Erling Smørgrav Sept. 30, 2025

Sept. 30, 2025

Hi, I've had complaints from FreeBSD users and downstream vendors about the change in 2025a which replaced "UTC" with "-00" when we fail to load a time zone. It turns out quite a few people relied on the old default as a faster way of setting the time zone to UTC, and I can't say I blame them. The attached patch changes tzset_unlocked() slightly so that UNSPEC is only used if TZ was set or zoneinit() returned an error other than ENOENT, so the behavior TZ is unset and TZDEFAULT does not exist is unchanged but we still get an error indication if TZ is set to something we don't understand, or TZDEFAULT exists but can't be read, or TZDEFAULT is readable but not a valid TZif file. DES -- Dag-Erling Smørgrav - des(a)des.no

8 20

Improving tzcode localtime.c security a la FreeBSD
by Paul Eggert Sept. 29, 2025

Sept. 29, 2025

The first difference between FreeBSD and tzcode localtime.c that I'd like to address is that FreeBSD is more cautious about handling outlandish TZ settings like "../foo". It seems to me that tzcode should largely mimic FreeBSD's extra caution. In the long run I'd like tzcode localtime.c to be byte-for-byte identical to FreeBSD localtime.c but it's better to do this one step at a time and this security caution is a good first step. I noticed some minor issues with the FreeBSD implementation of this extra checking, so I propose implementing it in a somewhat different way in tzcode. I'm attaching proposed patches to do that, and have installed them in the tzcode development repository on GitHub. Here are issues I noticed: * With an environment variable setting like TZ="/usr/share/zoneinfo/America/Los_Angeles", in a privileged program FreeBSD first opens the directory /usr/share/zoneinfo to get a directory file descriptor, and then uses fstatat and openat, both with AT_RESOLVE_BENEATH. However, it passes the full name to fstatat which must be a typo; I can't see how that would work reliably for a setuid program, even though the neighboring code suggests it was intended to work. I assume it was intended to pass a relative name to fstatat. * As near as I can make out, in privileged programs FreeBSD rejects adjacent slashes in settings like TZ="/usr/share/zoneinfo//America/Los_Angeles", where there are two slashes after "zoneinfo". It would be nicer to treat those two slashes as if they were one, as the kernel does. * tzcode is intended to be portable to platforms lacking openat and AT_RESOLVE_BENEATH, so for now the attached patches mimic FreeBSD's extra checking without using these two features. There are efficiency motivations for using the two features if available, but I'd like to defer that to a later patchset. * Although FreeBSD takes some steps to treat TZ settings the same regardless of whether they come from the TZ environment variable, it doesn't take this idea to its logical conclusion. It seems to me that it's better to not care where the TZ setting came from, as that's easier to document, understand, and maintain, so the attached patches do that. * FreeBSD uses issetugid, which is not universally available. The attached patches work around this by supplying an issetugid substitute for platforms like GNU/Linux that lack it. The first attached patch is a minor indenting fixup. The second one is a minor efficiency improvement, to use strnlen instead of strlen when that improves asymptotic efficiency. The third patch adopts FreeBSD's idea of treating a TZ setting starting with "/usr/share/zoneinfo/" as if it lacked that prefix. And the fourth and main patch changes the tzcode security checks as described above. These tzcode patches do not address other differences between tzcode and FreeBSD, such as the multithreading differences or the check-for-changes differences. These can wait for later patchsets. Comments and suggestions are welcome as usual.

3 12

[PROPOSED] Also treat space as bogus in abbreviations
by Paul Eggert Sept. 29, 2025

Sept. 29, 2025

* NEWS: Mention this. * localtime.c (scrub_abbrs): Treat ' ' as bogus. --- NEWS | 5 +++++ localtime.c | 1 - 2 files changed, 5 insertions(+), 1 deletion(-) diff --git a/NEWS b/NEWS index 0436c615..2c354038 100644 --- a/NEWS +++ b/NEWS @@ -51,6 +51,11 @@ Unreleased, experimental changes -DHAVE_GETRESUID=[01], and -DHAVE_GETEUID=[01] to enable or disable these system calls' use. + tzset etc. now treat ' ' like '_' in time zone abbreviations, + just as they treat other invalid bytes. This continues the + transition begun in release 96k, which removed spaces in tzdata + because the spaces break time string parsers. + 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 c5bc4e93..aa9ad7bd 100644 --- a/localtime.c +++ b/localtime.c @@ -504,7 +504,6 @@ scrub_abbrs(struct state *sp) for (i = 0; i < sp->charcnt; ++i) switch (sp->chars[i]) { case '\0': - case ' ': case '+': case '-': case '.': case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': -- 2.48.1

1 0

[PATCH 1/3] Prefer mempcpy to doing it by hand
by Paul Eggert Sept. 29, 2025

Sept. 29, 2025

This is less error-prone. Supply a mempcpy if the system lacks it. * Makefile, NEWS: Mention this. * localtime.c (tzloadbody, tzparse): * zic.c (random_dirent, relname, doabbr): * zdump.c (my_snprintf) [!HAVE_SNPRINTF]: (istrftime): Prefer mempcpy to doing it by hand. * private.h (HAVE_MEMPCPY): New macro. (mempcpy) [!HAVE_MEMPCPY]: New function. --- Makefile | 1 + NEWS | 3 +++ localtime.c | 13 +++++++------ private.h | 17 +++++++++++++++++ zdump.c | 11 +++++++---- zic.c | 36 ++++++++++++++++++++---------------- 6 files changed, 55 insertions(+), 26 deletions(-) diff --git a/Makefile b/Makefile index 83122ecc..5b577781 100644 --- a/Makefile +++ b/Makefile @@ -262,6 +262,7 @@ LDLIBS= # -DHAVE_LOCALTIME_RZ=0 if you do not want zdump to use localtime_rz # localtime_rz can make zdump significantly faster, but is nonstandard. # -DHAVE_MALLOC_ERRNO=0 if malloc etc. do not set errno on failure. +# -DHAVE_MEMPCPY=1 if your system has mempcpy, 0 if not (default is guessed) # -DHAVE_POSIX_DECLS=0 if your system's include files do not declare # functions like 'link' or variables like 'tzname' required by POSIX # -DHAVE_SETENV=0 if your system lacks the setenv function diff --git a/NEWS b/NEWS index 5c4786cb..0436c615 100644 --- a/NEWS +++ b/NEWS @@ -51,6 +51,9 @@ Unreleased, experimental changes -DHAVE_GETRESUID=[01], and -DHAVE_GETEUID=[01] to enable or disable these system calls' use. + tzcode now uses mempcpy if available, guessing its availability. + Compile with -DHAVE_MEMPCPY=1 or 0 to override the guess. + tzcode now uses strnlen to improve asymptotic performance a bit. Compile with -DHAVE_STRNLEN=0 if your platform lacks it. diff --git a/localtime.c b/localtime.c index 67500312..316c7504 100644 --- a/localtime.c +++ b/localtime.c @@ -624,6 +624,7 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, } if (!SUPPRESS_TZDIR && name[0] != '/') { + char *cp; if (sizeof lsp->fullname - sizeof tzdirslash <= strnlen(name, sizeof lsp->fullname - sizeof tzdirslash)) return ENAMETOOLONG; @@ -631,8 +632,9 @@ tzloadbody(char const *name, struct state *sp, char tzloadflags, /* Create a string "TZDIR/NAME". Using sprintf here would pull in stdio (and would fail if the resulting string length exceeded INT_MAX!). */ - memcpy(lsp->fullname, tzdirslash, sizeof tzdirslash); - strcpy(lsp->fullname + sizeof tzdirslash, name); + cp = lsp->fullname; + cp = mempcpy(cp, tzdirslash, sizeof tzdirslash); + strcpy(cp, name); name = lsp->fullname; } @@ -1471,12 +1473,11 @@ tzparse(const char *name, struct state *sp, struct state const *basep) } sp->charcnt = charcnt; cp = sp->chars; - memcpy(cp, stdname, stdlen); - cp += stdlen; + cp = mempcpy(cp, stdname, stdlen); *cp++ = '\0'; if (dstlen != 0) { - memcpy(cp, dstname, dstlen); - *(cp + dstlen) = '\0'; + cp = mempcpy(cp, dstname, dstlen); + *cp = '\0'; } return true; } diff --git a/private.h b/private.h index 6ae0c2fd..867340fa 100644 --- a/private.h +++ b/private.h @@ -837,6 +837,23 @@ extern char *asctime_r(struct tm const *restrict, char *restrict); extern char **environ; #endif +#ifndef HAVE_MEMPCPY +# if (defined mempcpy \ + || defined __FreeBSD__ || defined __NetBSD__ || defined __linux__) +# define HAVE_MEMPCPY 1 +# else +# define HAVE_MEMPCPY 0 +# endif +#endif +#if !HAVE_MEMPCPY +static void * +mempcpy(char *restrict s1, char const *restrict s2, size_t n) +{ + char *p = memcpy(s1, s2, n); + return p + n; +} +#endif + #if 2 <= HAVE_TZNAME + (TZ_TIME_T || !HAVE_POSIX_DECLS) extern char *tzname[]; #endif diff --git a/zdump.c b/zdump.c index 8e836e60..93fe05cb 100644 --- a/zdump.c +++ b/zdump.c @@ -927,6 +927,7 @@ my_snprintf(char *s, size_t size, char const *format, ...) int n; va_list args; char const *arg; + char *cp; size_t arglen, slen; char buf[1024]; va_start(args, format); @@ -943,8 +944,9 @@ my_snprintf(char *s, size_t size, char const *format, ...) arglen = n; } slen = arglen < size ? arglen : size - 1; - memcpy(s, arg, slen); - s[slen] = '\0'; + cp = s; + cp = mempcpy(cp, arg, slen); + *cp = '\0'; n = arglen <= INT_MAX ? arglen : -1; va_end(args); return n; @@ -1073,8 +1075,9 @@ istrftime(char *buf, ptrdiff_t size, char const *time_fmt, char fbuf[100]; bool oversized = sizeof fbuf <= f_prefix_copy_size; char *f_prefix_copy = oversized ? xmalloc(f_prefix_copy_size) : fbuf; - memcpy(f_prefix_copy, f, f_prefix_len); - strcpy(f_prefix_copy + f_prefix_len, "X"); + char *cp = f_prefix_copy; + cp = mempcpy(cp, f, f_prefix_len); + strcpy(cp, "X"); formatted_len = strftime(b, s, f_prefix_copy, tm); if (oversized) free(f_prefix_copy); diff --git a/zic.c b/zic.c index 6bdb039d..e78dfd47 100644 --- a/zic.c +++ b/zic.c @@ -1327,10 +1327,10 @@ random_dirent(char const **name, char **namealloc) uint_fast64_t unfair_min = - ((UINTMAX_MAX % base__6 + 1) % base__6); if (!dst) { - dst = xmalloc(size_sum(dirlen, prefixlen + suffixlen + 1)); - memcpy(dst, src, dirlen); - memcpy(dst + dirlen, prefix, prefixlen); - dst[dirlen + prefixlen + suffixlen] = '\0'; + char *cp = dst = xmalloc(size_sum(dirlen, prefixlen + suffixlen + 1)); + cp = mempcpy(cp, src, dirlen); + cp = mempcpy(cp, prefix, prefixlen); + cp[suffixlen] = '\0'; *name = *namealloc = dst; } @@ -1430,15 +1430,17 @@ relname(char const *target, char const *linkname) if (*linkname == '/') { /* Make F absolute too. */ size_t len = strlen(directory); - size_t lenslash = len + (len && directory[len - 1] != '/'); + bool needs_slash = len && directory[len - 1] != '/'; + size_t lenslash = len + needs_slash; size_t targetsize = strlen(target) + 1; + char *cp; if (*directory != '/') return NULL; linksize = size_sum(lenslash, targetsize); - f = result = xmalloc(linksize); - memcpy(result, directory, len); - result[len] = '/'; - memcpy(result + lenslash, target, targetsize); + f = cp = result = xmalloc(linksize); + cp = mempcpy(cp, directory, len); + *cp = '/'; + memcpy(cp + needs_slash, target, targetsize); } for (i = 0; f[i] && f[i] == linkname[i]; i++) if (f[i] == '/') @@ -1448,11 +1450,13 @@ relname(char const *target, char const *linkname) taillen = strlen(f + dir_len); dotdotetcsize = size_sum(size_product(dotdots, 3), taillen + 1); if (dotdotetcsize <= linksize) { + char *cp; if (!result) result = xmalloc(dotdotetcsize); + cp = result; for (i = 0; i < dotdots; i++) - memcpy(result + 3 * i, "../", 3); - memmove(result + 3 * dotdots, f + dir_len, taillen + 1); + cp = mempcpy(cp, "../", 3); + memmove(cp, f + dir_len, taillen + 1); } return result; } @@ -2875,11 +2879,11 @@ doabbr(char *abbr, struct zone const *zp, char const *letters, else if (letters == disable_percent_s) return 0; sprintf(abbr, format, letters); - } else if (isdst) { - strcpy(abbr, slashp + 1); - } else { - memcpy(abbr, format, slashp - format); - abbr[slashp - format] = '\0'; + } else if (isdst) + strcpy(abbr, slashp + 1); + else { + char *abbrend = mempcpy(abbr, format, slashp - format); + *abbrend = '\0'; } len = strlen(abbr); if (!doquotes) -- 2.48.1

1 2

Changes to moderation procedures
by James Mitchell Sept. 29, 2025

Sept. 29, 2025

Hi all, We've made a small update to moderation settings in response to recent spam: * First posts from new members will be held for review. Once approved, the moderation flag will be removed. * All posts from non-members will continue to be held for moderation. A few existing members currently have moderation enabled. You may experience brief delays while these changes take effect—thank you for your patience. If you believe your messages are being held in error, please contact the list owners at tz-owner(a)iana.org<mailto:tz-owner@iana.org>. Kind regards, James Mitchell IANA

1 0

List moderation
by Kim Davies Sept. 24, 2025

Sept. 24, 2025

Hi folks, Our team have removed the list subscriber that flooded the list with emails in the last few minutes, and have put some emergency moderation in place until we can confirm it has died down. Apologies for the noise. kim

1 0

Prevent spam
by stan＠ulbrych.org Sept. 24, 2025

Sept. 24, 2025

Hello, I just received a barrage of emails (over 50) due to spam messages, I suggest the list owners look into first posts requiring approval before they are published. This is done for example in the Python docs list, and allows for such spam to be prevented. Best Regards, Stan Ulbrych

1 0

List moderation
by Kim Davies Sept. 24, 2025

Sept. 24, 2025

Hi folks, Our team have removed the list subscriber that flooded the list with emails in the last few minutes, and have put some emergency moderation in place until we can confirm it has died down. Apologies for the noise. kim

1 0