Parsing zone files generated with zic -b slim makes data be bogus

[LarsGullik]

iana-tzdb changed the default for zic from -b fat to -b slim in 2020b. This seems to trigger some
problems with the tzif parser.

In particular the begin/end info in sys_info is not correct anymore:

auto sys = sys_days(2020_y/April/26_d);
auto * europe_oslo = locate_zone("Europe/Oslo");
auto sys_info = europe_oslo->get_info(sys);
ASSERT_EQUALS(sys_days{2020_y/March/29_d} + 1h, sys_info.begin);

Fails with:
expected: 2020-03-29 01:00:00
actual: 1996-03-31 01:00:00

[HowardHinnant]

Thanks for the report. The binary db reader is going to need an overhaul to accommodate slim. Do you know if any platform is actually shipping slim?

[LarsGullik]

On Mon, 26 Oct 2020 at 00:20, Howard Hinnant ***@***.***> wrote: Thanks for the report. The binary db reader is going to need an overhaul to accommodate slim. Do you know if any platform is actually shipping slim?

I am not aware of any, except for the one that we make ourselves. We are changing to explicit '-b fat' to work around this issue.

…

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#614 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADXUFODYOYZRM43LYCDOLZ3SMSXCXANCNFSM4S6UC4AQ> .

-- Lgb

[LarsGullik]

A google search shows me that some projects that have their own copy of the zoneinfo files are already switching to using the slim format.

[charno]

zic has changed it's standard behavour. From the Changelog (https://data.iana.org/time-zones/releases/tzcode2021d.tar.gz, file NEWS)

Release 2020b - 2020-10-06 18:35:04 -0700
[...]
zic now defaults to '-b slim' instead of to '-b fat'.

[joemarshall]

@HowardHinnant We've hit this bug in Apache Arrow also. On emscripten / webassembly python, we don't have embedded tzdata so we use the tzdata package as a source for zone files. The python tzdata package uses -b slim format now; so timezone handling breaks in arrow.