coreutils `test_date::test_negative_offset` and `test_date::test_relative_weekdays` fail after parse_datetime update

[drinkcat]

• Checkout this branch that updates coreutils to latest parse_datetime: https://github.com/drinkcat/coreutils/tree/parse_datetime-0.10.0
• cargo run test_date

---- test_date::test_relative_weekdays stdout ----
bin: "/home/drinkcat/dev/coreutils/coreutils/target/debug/coreutils"
run: /home/drinkcat/dev/coreutils/coreutils/target/debug/coreutils date -d last thu --rfc-3339=seconds --utc

thread 'test_date::test_relative_weekdays' panicked at tests/by-util/test_date.rs:472:18:
Predicate for stdout as `str` evaluated to false.
stdout='2025-05-29 00:00:00+00:00
'
stderr=''

note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

---- test_date::test_negative_offset stdout ----
bin: "/home/drinkcat/dev/coreutils/coreutils/target/debug/coreutils"
run: /home/drinkcat/dev/coreutils/coreutils/target/debug/coreutils date -d -1 hour --rfc-3339=seconds

thread 'test_date::test_negative_offset' panicked at tests/by-util/test_date.rs:445:14:
Predicate for stdout as `str` evaluated to false.
stdout='2025-05-28 23:00:00+00:00
'
stderr=''



failures:
    test_date::test_negative_offset
    test_date::test_relative_weekdays

[sylvestre]

captain obvious: we should replicate these tests in parse_datetime (if possible)

[drinkcat]

Bisection:

There are only 'skip'ped commits left to test.
The first bad commit could be any of:
07d4b80ea53747ee629a2104a9a41bea16fcbf4e
4c3d2221da1be8f6e091c3842d6ef98983f37527
We cannot bisect more!

I suspect this might be it:

commit 4c3d2221da1be8f6e091c3842d6ef98983f37527
Author: Daniel Hilst <danielhilst@gmail.com>
Date:   Thu Aug 29 14:21:15 2024 -0300

    Restore the old API and add the missing combinators

@yuankunzhang this was merged as part of #137.

I was digging through the code, and what I do not understand is this:

fn at_date_inner(date: Vec<Item>, mut d: DateTime<FixedOffset>) -> Option<DateTime<FixedOffset>> {
    d = d.with_hour(0).unwrap();
    d = d.with_minute(0).unwrap();
    d = d.with_second(0).unwrap();
    d = d.with_nanosecond(0).unwrap();

Why are we resetting the time to midnight? (removing this would fix this test... but... break others)

[drinkcat]

captain obvious: we should replicate these tests in parse_datetime (if possible)

Oh yeah absolutely... This kind of cross-repo/crate issue is such a pain...

[yuankunzhang]

Addressed in #152.