- tz - lists.iana.org

Introducing myself
by Stephen Colebourne Aug. 29, 2013

Aug. 29, 2013

On 29 August 2013 20:39, Kevin Lyda <kevin(a)ie.suberic.net> wrote: > No solution will avoid politics. And someone starting a number of threads on > this topic clearly has their own political agenda. I should introduce myself, as not everyone will know who I am, or what I represent. I am Stephen Colebourne - a Java developer of 15+ years - recognised as a Java Champion, a leader in the Java community - author of Joda-Time, the most widely adopted 3rd party API for date and time in Java - co-spec lead of JSR-310, the effort to write a better API for the Java JDK - have written APIs to parse the tzdb - have designed APIs to expose that data to developers - work very closely with Oracle, and have had conversations with CLDR in the past My "political agenda" is to avoid the effort I've put in over the last 12 years on date and time in Java being damaged. Recent changes give the appearance of a lack of understanding of the importance of pre-1970 data and of cross-border localization. I'm sorry that I've had to write so much and object so strongly. I want to find a solution we can all agree on. It seems harder than it should be. Stephen

3 2

Problematic/dubious commits in 2013
by Stephen Colebourne Aug. 29, 2013

Aug. 29, 2013

I have reviewed every commit since the end of 2012. Here are the ones I consider to be problematic. I suggest that others have another review of these, as it may be that not everyone has taken the time to do this. - One zone for each inhabited ISO-3166 region should not have been removed: https://github.com/eggert/tz/commit/d3b025adb25554ee10b986850371e573df92733… - Ensuing changes were wrong: https://github.com/eggert/tz/commit/df99923bde035a263493f3435db63db15df14d53 - Reversion did not revert everything, specifically the reverted links are still in the backward file, not the main file, indicating that they are deprecated which should not be the case (eg for Europe/Zagreb): https://github.com/eggert/tz/commit/ee40570d7d937de749fe89676dec2ff1526ee68b - Pre-1970 damage: https://github.com/eggert/tz/commit/5903db75e5974fbdede1fa05bcef53ff645ac13d https://github.com/eggert/tz/commit/9432f1d23d3c5bbcc14c26f781185c01679edeef - Montreal removal (looks like it changes pre-1970 history) https://github.com/eggert/tz/commit/45dcf69b45087cff50282d4da64b86a7d705ddf3 - Navajo - America/ShipRock removal https://github.com/eggert/tz/commit/d58b59dd8a1e4c10abdbab8857ae521f6c2d2536 https://github.com/eggert/tz/commit/81be53c16b97ce0bc561ce3891ce2503227e3369 - Other possibly dubious commits tzselect changes https://github.com/eggert/tz/commit/cc3f93886bcfe36dfedc8f340827fded6e43539c https://github.com/eggert/tz/commit/ebdf0c5d0ce06b20dcd7e984b7370548f74dc210 Not all of these cause damage, but some clearly do. I'm unhappy with most of them, as they break the principles I value. Stephen

1 0

Principles
by Stephen Colebourne Aug. 29, 2013

Aug. 29, 2013

It seems like there are conflicts in the principles behind the tzdb. Here are my thoughts: What is the tzdb? Generally accepted: - widely used - ubiquitous - reliable - accurate post-1970 - handles changes occurring now More controversial: - IDs used in GUIs (they are whether we like it or not) - IDs have meaning (they do whether we like it or not) - pre-1970 is in-scope (data is there whether we like it or not) - pre-1970 data may not be accurate (users like data, care less about its accuracy) - pre-1970 LMT is accurate for the location in the tzdb ID What do end users expect? (developers and non-developers) - users often pick zones based on raw tzdb ID - users don't know the history of zone rules in their location - users do know that they live in an area that might have special zone rules, eg. Busingen/Navajo - users want minimal cognitive effort, pick it and it works - users don't understand the nuances of a cross-border zone, they just associate with their region What do I expect? - Stability - No deletions of data, except where _proven_ wrong - Additions focussed on today and the future - Historical changes only following research - No new time-zones in the main files unless they differ post-1970 - At least one time-zone ID for each ISO-3166 region - Using all files except backward should meet goals above - Backward file is for deprecations only I don't think that any of this is unreasonable, or different from how the tzdb has previously worked. Stephen

1 0

Namibia Time Change
by Hank W. Aug. 26, 2013

Aug. 26, 2013

Reminder: The annual switch from West Africa Time (UTC+1) to West Africa Summer Time (UTC+2) in Namibia only is scheduled for this coming Sunday, 1 September, at 2 AM WAT (1 AM UTC).

2 2

Unable to get the tz database
by Peddi, Radhika (Radhika) Aug. 24, 2013

Aug. 24, 2013

Hi, This is Radhika from Avaya. When I am trying to get tz database from the link http://www.twinsun.com/tz/tz-link.htm, by using FTP distribution, I am getting Host Not Found error. Please find the attached screen shot for the same. When I am using below shell command on my linux setup wget 'ftp://elsie.nci.nih.gov/pub/tz*.tar.gz', I am getting below error --2013-08-20 03:14:03-- ftp://elsie.nci.nih.gov/pub/tz*.tar.gz => `.listing' Resolving elsie.nci.nih.gov... failed: Name or service not known. wget: unable to resolve host address `elsie.nci.nih.gov' Could you please help me how to get rid of this problem. Thanks & Regards, Radhika Peddi|Avaya India Pvt. Ltd.| Wing 'A', Level -5, Tower-11, Cybercity, Magarpatta City, Hadapsar | Pune, 411028|India Voice +91 20 4101 8311| Email: rpeddi<mailto:rpeddi@avaya.com>@avaya.com [ Please consider the environment before printing this email ]

2 3

infinite loop in time2sub
by enh Aug. 22, 2013

Aug. 22, 2013

i recently upgraded Android's copy of tzcode to be closer to tzcode2013d. with Android's signed 32-bit time_t i'm seeing a change in behavior for out-of-range dates. previously given something like this, mktime would return an error, but now it hangs in the binary search in time2sub. struct tm t; memset(&t, 0, sizeof(tm)); t.tm_year = 200; t.tm_mon = 2; t.tm_mday = 10; changing the if (t == lo) { ++t; if (t <= lo) return WRONG; ++lo; } else if (t == hi) { --t; if (t >= hi) return WRONG; --hi; } back to if (t == lo) { if (t == TIME_T_MAX) return WRONG; ++t; ++lo; } else if (t == hi) { if (t == TIME_T_MIN) return WRONG; --t; --hi; } fixes things. looking at the git history in https://github.com/eggert/tz.gitit looks like the old code, found in Android up to and including jb-mr2 (but no longer in AOSP ToT), was never in upstream tzcode. here's where the upstream tzcode's overflow checks were introduced: commit 46fa7e98177cae664f673c6034c9ac74d0ccd0ec Author: Arthur David Olson <olsona@elsie> Date: Tue Aug 9 09:55:06 2005 -0400 overflow work SCCS-file: localtime.c SCCS-SID: 7.96 diff --git a/localtime.c b/localtime.c index 1c3e9b0..d3edb5a 100644 --- a/localtime.c +++ b/localtime.c @@ -1616,9 +1616,13 @@ const int do_norm_secs; if (dir != 0) { if (t == lo) { ++t; + if (t <= lo) + return WRONG; ++lo; } else if (t == hi) { --t; + if (t >= hi) + return WRONG; --hi; } if (lo > hi) looking at the Android AOSP git history, it looks like we found and fixed this bug years ago but never talked to upstream about it: commit 2093d350be21ff086f9e145404877941b9a42c5c Author: David 'Digit' Turner <digit(a)google.com> Date: Wed Sep 9 17:41:59 2009 -0700 Fix an infinite loop in time2sub. The problem is that time_t is signed, and the original code relied on the fact that (X + c < X) in case of overflow for c >= 0. Unfortunately, this condition is only guaranteed by the standard for unsigned arithmetic, and the gcc 4.4.0 optimizer did completely remove the corresponding test from the code. This resulted in a missing boundary check, and an infinite loop. The problem is solved by testing explicitely for TIME_T_MIN and TIME_T_MAX in the loop that uses this. Also fix increment_overflow and long_increment_overflow which were buggy for exactly the same reasons. Note: a similar fix is needed for system/core/libcutils diff --git a/libc/tzcode/localtime.c b/libc/tzcode/localtime.c index 9c5f218..83c1011 100644 --- a/libc/tzcode/localtime.c +++ b/libc/tzcode/localtime.c @@ -75,6 +75,31 @@ static __inline__ void _tzUnlock(void) pthread_mutex_unlock(&_tzMutex); } +/* Complex computations to determine the min/max of time_t depending + * on TYPE_BIT / TYPE_SIGNED / TYPE_INTEGRAL. + * These macros cannot be used in pre-processor directives, so we + * let the C compiler do the work, which makes things a bit funky. + */ +static const time_t TIME_T_MAX = + TYPE_INTEGRAL(time_t) ? + ( TYPE_SIGNED(time_t) ? + ~((time_t)1 << (TYPE_BIT(time_t)-1)) + : + ~(time_t)0 + ) + : /* if time_t is a floating point number */ + ( sizeof(time_t) > sizeof(float) ? (time_t)DBL_MAX : (time_t)FLT_MAX ); + +static const time_t TIME_T_MIN = + TYPE_INTEGRAL(time_t) ? + ( TYPE_SIGNED(time_t) ? + ((time_t)1 << (TYPE_BIT(time_t)-1)) + : + 0 + ) + : + ( sizeof(time_t) > sizeof(float) ? (time_t)DBL_MIN : (time_t)FLT_MIN ); + #ifndef WILDABBR /* ** Someone might make incorrect use of a time zone abbreviation: @@ -1683,11 +1708,16 @@ increment_overflow(number, delta) int * number; int delta; { - int number0; + unsigned number0 = (unsigned)*number; + unsigned number1 = (unsigned)(number0 + delta); + + *number = (int)number1; - number0 = *number; - *number += delta; - return (*number < number0) != (delta < 0); + if (delta >= 0) { + return ((int)number1 < (int)number0); + } else { + return ((int)number1 > (int)number0); + } } static int @@ -1695,11 +1725,16 @@ long_increment_overflow(number, delta) long * number; int delta; { - long number0; + unsigned long number0 = (unsigned long)*number; + unsigned long number1 = (unsigned long)(number0 + delta); + + *number = (long)number1; - number0 = *number; - *number += delta; - return (*number < number0) != (delta < 0); + if (delta >= 0) { + return ((long)number1 < (long)number0); + } else { + return ((long)number1 > (long)number0); + } } static int @@ -1868,14 +1903,14 @@ const int do_norm_secs; } else dir = tmcomp(&mytm, &yourtm); if (dir != 0) { if (t == lo) { - ++t; - if (t <= lo) + if (t == TIME_T_MAX) return WRONG; + ++t; ++lo; } else if (t == hi) { - --t; - if (t >= hi) + if (t == TIME_T_MIN) return WRONG; + --t; --hi; } if (lo > hi) right now i'm going to work around this by building the tzcode part of the C library with -fno-strict-overflow (because i don't know how many other places are relying on this unspecified behavior), but i'll apply whatever patch goes into upstream as soon as it's available. --elliott

2 2

[PATCH] * .gitignore: Add 'date'.
by Paul Eggert Aug. 22, 2013

Aug. 22, 2013

--- .gitignore | 1 + 1 file changed, 1 insertion(+) diff --git a/.gitignore b/.gitignore index 5759d2a..18dbbcc 100644 --- a/.gitignore +++ b/.gitignore @@ -4,6 +4,7 @@ *.txt *~ ChangeLog +date leapseconds time.tab tzselect -- 1.7.11.7

1 0

[PATCH] tzselect: new options -c COORD and -n LIMIT
by Paul Eggert Aug. 22, 2013

Aug. 22, 2013

These let the user select TZ based on geographical coordinates. Combining '-c COORD' with '-t time' better insulates the user from issues of nationality. * tzselect.8 (SYNOPSIS, OPTIONS): Document the new options. * tzselect.ksh: Implement them, using the great-circle special case of the Vicenty formula for distances on ellipsoids. (LC_ALL): Set to C, since tzselect is English only. That way, we treat decimal-points in -c option operands the same in all environments. (usage): Document new options. Document existing ones better. (output_distances): New variable. --- tzselect.8 | 26 ++++++++++ tzselect.ksh | 162 +++++++++++++++++++++++++++++++++++++++++++++++++++++++---- 2 files changed, 177 insertions(+), 11 deletions(-) diff --git a/tzselect.8 b/tzselect.8 index b5e2c88..884af36 100644 --- a/tzselect.8 +++ b/tzselect.8 @@ -4,6 +4,12 @@ tzselect \- select a time zone .SH SYNOPSIS .B tzselect [ +.B \-c +.I coord +] [ +.B \-n +.I limit +] [ .B \-t .I zonetabtype ] [ @@ -21,6 +27,26 @@ The output is suitable as a value for the TZ environment variable. All interaction with the user is done via standard input and standard error. .SH OPTIONS .TP +.BI "\-c " coord +Instead of asking for continent and then country and then city, +ask for selection from time zones whose largest cities +are closest to the location with geographical coordinates +.IR coord . +Use ISO 6709 notation for +.IR coord , +for example, +.B "\-c\ \+42.391415\-071.570419" +for 42.391415\(de\|N, 71.570419\(de\|W, and +.B "\-c\ +404226\-0740319" +for 48\(de\|42\(fm\|26\(sd\|N, 74\(de\|3\(fm\|19\(sd\|W. +.TP +.BI "\-n " limit +When +.B \-c +is used, display the closest +.I limit +locations (default 10). +.TP .BI "\-t " zonetabtype Make selections from the time zone table of type .IR zonetabtype . diff --git a/tzselect.ksh b/tzselect.ksh index dc9f256..da6adb2 100644 --- a/tzselect.ksh +++ b/tzselect.ksh @@ -1,5 +1,7 @@ #!/bin/bash +export LC_ALL=C + PKGVERSION='(tzcode) ' TZVERSION=see_Makefile REPORT_BUGS_TO=tz(a)iana.org @@ -41,15 +43,43 @@ ZONETABTYPE=zone exit 1 } -usage="Usage: tzselect [--version] [--help] [-t ZONETABTYPE] +coord= +location_limit=10 + +usage="Usage: tzselect [--version] [--help] [-c COORD] [-n LIMIT] [-t ZONETABTYPE] Select a time zone interactively. -ZONETABTYPE should be one of 'time' or 'zone'. + +Options: + + -c COORD + Instead of asking for continent and then country and then city, + ask for selection from time zones whose largest cities + are closest to the location with geographical coordinates COORD. + COORD should use ISO 6709 notation, for example, '-c +4852+00220' + for Paris. + + -n LIMIT + Display at most LIMIT locations when -c is used (default $location_limit). + + -t ZONETABTYPE + Use time zone table ZONETABTYPE. ZONETABTYPE should be one of + 'time' or 'zone'. + + --version + Output version information. + + --help + Output this help. Report bugs to $REPORT_BUGS_TO." -while getopts t:-: opt +while getopts c:n:t:-: opt do case $opt$OPTARG in + c*) + coord=$OPTARG ;; + n*) + location_limit=$OPTARG ;; t*) ZONETABTYPE=$OPTARG ;; -help) @@ -90,6 +120,66 @@ case $(echo 1 | (select x in x; do break; done) 2>/dev/null) in ?*) PS3= esac +# Awk script to read a time zone table and output the same table, +# with each column preceded by its distance from 'here'. +output_distances=' + BEGIN { + FS = "\t" + while (getline <TZ_COUNTRY_TABLE) + if ($0 ~ /^[^#]/) + country[$1] = $2 + country["US"] = "US" # Otherwise the strings get too long. + } + function cvt1(coord, deg, min, ilen, sign, sec) { + ilen = length(coord) + sign = substr(coord, 1, 1) + if (coord ~ /\./) { + deg = coord + 0 + } else { + if (ilen <= 6) { + sec = 0 + } else { + sec = sign substr(coord, ilen - 1) + ilen -= 2 + } + min = sign substr(coord, ilen - 1, 2) + deg = substr(coord, 1, ilen - 2) + deg = (deg * 3600.0 + min * 60.0 + sec) / 3600.0 + } + return deg * 0.017453292519943295 + } + function convert_latitude(coord) { + match(coord, /..*[-+]/) + return cvt1(substr(coord, 1, RLENGTH - 1)) + } + function convert_longitude(coord) { + match(coord, /..*[-+]/) + return cvt1(substr(coord, RLENGTH)) + } + # Great-circle distance between points with given latitude and longitude. + # Inputs and output are in radians. This uses the great-circle special + # case of the Vicenty formula for distances on ellipsoids. + function dist(lat1, long1, lat2, long2, dlong, x, y, num, denom) { + dlong = long2 - long1 + x = cos (lat2) * sin (dlong) + y = cos (lat1) * sin (lat2) - sin (lat1) * cos (lat2) * cos (dlong) + num = sqrt (x * x + y * y) + denom = sin (lat1) * sin (lat2) + cos (lat1) * cos (lat2) * cos (dlong) + return atan2(num, denom) + } + BEGIN { + coord_lat = convert_latitude(coord) + coord_long = convert_longitude(coord) + } + /^[^#]/ { + here_lat = convert_latitude($2) + here_long = convert_longitude($2) + line = $1 "\t" $2 "\t" $3 "\t" country[$1] + if (NF == 4) + line = line " - " $4 + printf "%g\t%s\n", dist(coord_lat, coord_long, here_lat, here_long), line + } +' # Begin the main loop. We come back here if the user wants to retry. while @@ -101,10 +191,14 @@ while country= region= + case $coord in + ?*) + continent=coord;; + '') # Ask the user for continent or ocean. - echo >&2 'Please select a continent or ocean.' + echo >&2 'Please select a continent, ocean, "coord", or "TZ".' select continent in \ Africa \ @@ -117,7 +211,8 @@ while Europe \ 'Indian Ocean' \ 'Pacific Ocean' \ - 'none - I want to specify the time zone using the Posix TZ format.' + 'coord - I want to use geographical coordinates.' \ + 'TZ - I want to specify the time zone using the Posix TZ format.' do case $continent in '') @@ -130,10 +225,12 @@ while break esac done + esac + case $continent in '') exit 1;; - none) + TZ) # Ask the user for a Posix TZ string. Check that it conforms. while echo >&2 'Please enter the desired value' \ @@ -158,6 +255,45 @@ while done TZ_for_date=$TZ;; *) + case $continent in + coord) + case $coord in + '') + echo >&2 'Please enter coordinates' \ + 'in ISO 6709 notation.' + echo >&2 'For example, +4042-07403 stands for' + echo >&2 '40 degrees 42 minutes north,' \ + '74 degrees 3 minutes west.' + read coord;; + esac + distance_table=$($AWK \ + -v coord="$coord" \ + -v TZ_COUNTRY_TABLE="$TZ_COUNTRY_TABLE" \ + "$output_distances" <$TZ_ZONE_TABLE | + sort -n | + sed "${location_limit}q" + ) + regions=$(echo "$distance_table" | $AWK ' + BEGIN { FS = "\t" } + { print $NF } + ') + echo >&2 'Please select one of the following' \ + 'time zone regions,' + echo >&2 'listed roughly in increasing order' \ + "of distance from $coord". + select region in $regions + do + case $region in + '') echo >&2 'Please enter a number in range.';; + ?*) break;; + esac + done + TZ=$(echo "$distance_table" | $AWK -v region="$region" ' + BEGIN { FS="\t" } + $NF == region { print $4 } + ') + ;; + *) # Get list of names of countries in the continent or ocean. countries=$($AWK -F'\t' \ -v continent="$continent" \ @@ -185,7 +321,8 @@ while # If there's more than one country, ask the user which one. case $countries in *"$newline"*) - echo >&2 'Please select a country.' + echo >&2 'Please select a country' \ + 'whose clocks agree with yours.' select country in $countries do case $country in @@ -256,6 +393,7 @@ while } $1 == cc && $4 == region { print $3 } ' <$TZ_ZONE_TABLE) + esac # Make sure the corresponding zoneinfo file exists. TZ_for_date=$TZDIR/$TZ @@ -292,9 +430,11 @@ Universal Time is now: $UTdate." echo >&2 "" echo >&2 "The following information has been given:" echo >&2 "" - case $country+$region in - ?*+?*) echo >&2 " $country$newline $region";; - ?*+) echo >&2 " $country";; + case $country%$region%$coord in + ?*%?*%) echo >&2 " $country$newline $region";; + ?*%%) echo >&2 " $country";; + %?*%?*) echo >&2 " coord $coord$newline $region";; + %%?*) echo >&2 " coord $coord";; +) echo >&2 " TZ='$TZ'" esac echo >&2 "" @@ -313,7 +453,7 @@ Universal Time is now: $UTdate." '') exit 1;; Yes) break esac -do : +do coord= done case $SHELL in -- 1.8.1.2

3 7

[PATCH 1/3] Further removal of attempt to support floating-point time_t.
by Paul Eggert Aug. 19, 2013

Aug. 19, 2013

* localtime.c (timesub): * zdump.c (my_localtime): Don't worry about time_t being floating-point. * localtime.c (timesub): Avoid undefined behavior on integer overflow when assigning time_t to int. --- localtime.c | 5 +++-- zdump.c | 2 +- 2 files changed, 4 insertions(+), 3 deletions(-) diff --git a/localtime.c b/localtime.c index 5312ad9..f58b20a 100644 --- a/localtime.c +++ b/localtime.c @@ -1489,9 +1489,10 @@ timesub(const time_t *const timep, const int_fast32_t offset, register int leapdays; tdelta = tdays / DAYSPERLYEAR; - idelta = tdelta; - if (tdelta - idelta >= 1 || idelta - tdelta >= 1) + if (! ((! TYPE_SIGNED(time_t) || INT_MIN <= tdelta) + && tdelta <= INT_MAX)) return NULL; + idelta = tdelta; if (idelta == 0) idelta = (tdays < 0) ? -1 : 1; newy = y; diff --git a/zdump.c b/zdump.c index 36c1838..fa5f8d8 100644 --- a/zdump.c +++ b/zdump.c @@ -244,7 +244,7 @@ my_localtime(time_t *tp) tm = *tmp; t = mktime(&tm); - if (t - *tp >= 1 || *tp - t >= 1) { + if (t != *tp) { (void) fflush(stdout); (void) fprintf(stderr, "\n%s: ", progname); (void) fprintf(stderr, tformat(), *tp); -- 1.8.1.2

3 7

Re: [tz] Problem with Busingen
by Paul Eggert Aug. 19, 2013

Aug. 19, 2013

Luca Marzolla wrote: > It is possible to fix this minor inconvenience in the next release? Yes. The problem should already be fixed in the experimental version, which you can get from Github, here: https://github.com/eggert/tz It's fixed in a different way, by the way: the Busingen link got moved to a different file. See: https://github.com/eggert/tz/commit/df99923bde035a263493f3435db63db15df14d53

1 0