Re: Question about reproducibility tests timezone format
Soren Stoutner <soren@debian.org> writes:
> On Thursday, October 2, 2025 10:33:36 AM Mountain Standard Time Russ Allbery
> wrote:
>> The point of using GMT+12 and GMT-12 is that they keep the same
>> (incorrect, but it doesn't really matter for this purpose) time zone
>> abbreviation for reproducibility testing:
>> % env TZ=GMT+12 date
>> Thu Oct 2 05:32:09 AM GMT 2025
>> % env TZ=GMT-12 date
>> Fri Oct 3 05:32:12 AM GMT 2025
>> Otherwise, you may get spurious reproducibility failures due to the
>> changed time zone abbreviation.
>> I don't believe there is a mechanism to force the same time zone
>> abbreviation other than usig POSIX rule-based zones.
> Isn’t the exact *purpose* of the reproducible builds project to make
> sure that software compiles the same even when things like time zone
> abbreviations are different? So, if this produces failures, aren’t these
> the exact types of failures the reproducible builds project is trying to
> uncover?
I have no idea; the reproducible build team should make that call. I could
see an argument either way depending on what the definition of
reproducibility is.
If they do indeed want to fail reproducibility if the time zone
abbreviation changes, then I agree, they should use Etc/GMT+12 and
Etc/GMT-12 instead, because that will achieve that goal. So far as I know,
the two benefits to using the POSIX rule-based time zones over those Etc/*
zones are to force the abbreviation to stay the same and to not require
that tzdata be installed, and I don't *think* the latter is a valid build
target for Debian (although tzdata is only required, not essential, so
maybe that's also a consideration?).
--
Russ Allbery (rra@debian.org) <https://www.eyrie.org/~eagle/>
Reply to: