fix: align Date::Language::str2time with Date::Parse behavior

[Koan-Bot]

What

Fixes two behavioral differences between Date::Language::str2time and Date::Parse::str2time.

Why

Date::Language::str2time and Date::Parse::str2time should behave consistently for the same inputs. Two discrepancies caused surprising results:

1. Year inference bug: parsing "15 Apr" on April 12 gave 2026 (Date::Language) vs 2025 (Date::Parse). Date::Language only compared months, missing the day-of-month check for same-month future dates.

2. Fatal error on invalid input: Date::Language::str2time("32 Jun 2024") died with Day '32' out of range, while Date::Parse::str2time returns undef. The timegm/timelocal calls were unprotected.

How

• Added day comparison to the year-inference heuristic (matching Date::Parse's logic)
• Added input validation (month/day/hour/minute/second range checks)
• Wrapped timegm/timelocal in eval blocks to return undef on failure

Testing

New t/lang-str2time.t covers:

• Same-month future/past day year inference (parity with Date::Parse)
• Invalid dates return undef instead of dying (day=32, hour=25, minute=61)
• Basic valid date parsing and non-English round-trip

Full test suite passes (28 test files).

🤖 Generated with Claude Code

---

Quality Report

Changes: 2 files changed, 137 insertions(+), 4 deletions(-)

Code scan: clean

Tests: skipped

Branch hygiene: clean

Generated by Kōan post-mission quality pipeline

[greptile-apps]

Greptile Summary

This PR fixes two real behavioral divergences between Date::Language::str2time and Date::Parse::str2time: the missing day-of-month comparison in year inference, and unprotected timegm/timelocal calls that die on invalid input. The fixes are correct, but the eval blocks stop short of full parity with Date::Parse.

• The timegm/timelocal eval paths are missing the -1 sentinel check and the $result < 0 && $year >= 1970 overflow guard that Date::Parse::str2time uses; on older Time::Local installations these guards are load-bearing, and their absence means the two functions still diverge for edge-case inputs.

Confidence Score: 4/5

Safe to merge after addressing the missing -1 / overflow guards in the eval blocks.

One P1 finding: the eval blocks don't replicate Date::Parse's -1 sentinel and negative-result overflow checks, leaving a behavioral gap on older Time::Local versions and 32-bit platforms. All other findings are P2.

lib/Date/Language.pm — the timegm/timelocal eval paths (lines 107–121)

Important Files Changed

Filename	Overview
lib/Date/Language.pm	Year-inference and eval-wrapping fixes are correct, but the new eval blocks are missing the -1 sentinel and negative-result overflow guards present in the Date::Parse reference implementation.
t/lang-str2time.t	New test file covering year inference, invalid input handling, and a German round-trip; has an unused POSIX qw(mktime) import and skips the future-day test on days 28–31.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[str2time input] --> B[strptime parse]
    B --> C{result empty?}
    C -- yes --> D[return undef]
    C -- no --> E[fill defaults: hh/mm/ss/month/day]
    E --> F{year defined?}
    F -- no --> G{month > cur_month OR same month and day > cur_day?}
    G -- yes --> H[year = last year]
    G -- no --> I[year = current year]
    F -- yes --> J[use parsed year]
    H --> K
    I --> K
    J --> K[Validate ranges: month<=11, day 1-31, hh<=23, mm<=59, ss<=59]
    K -- invalid --> D
    K -- valid --> L{zone defined?}
    L -- yes --> M[eval timegm / catch exception]
    L -- no --> N[eval timelocal / catch exception]
    M --> O{result defined?}
    N --> O
    O -- no --> D
    O -- yes --> P[result -= zone if zoned]
    P --> Q[return result]

Loading

_{Reviews (1): Last reviewed commit: "fix: align Date::Language::str2time with..." | Re-trigger Greptile}

[greptile-apps]

Missing -1 sentinel and overflow guards from Date::Parse

The PR claims parity with Date::Parse::str2time, but the eval blocks are missing two checks that exist in the reference implementation.

1. On older Time::Local versions (pre-1.23, ~pre-2011), timegm/timelocal returns -1 to signal error instead of dying. The eval won't catch that, so -1 would be returned as a valid epoch to callers.
2. Integer overflow on 32-bit systems can cause timegm/timelocal to return a large negative value; Date::Parse rejects $result < 0 && $year >= 1970.

The corresponding guard in Date::Parse::str2time (lines 302–308 / 317–323):

return undef
  if !defined $result
     or $result == -1
        && join("",$ss,$mm,$hh,$day,$month,$year)
             ne "595923311169";
return undef if $result < 0 && $year >= 1970;

Without these, the two code paths diverge for dates on or near epoch −1, and for overflow inputs on 32-bit platforms.

[greptile-apps]

Unused import mktime

POSIX qw(mktime) is imported but mktime is never called anywhere in the test file. This can be removed.

Suggested change

	use POSIX qw(mktime);
	use POSIX;

or simply drop the use POSIX line entirely if no other POSIX symbols are needed.

[greptile-apps]

Redundant defined $month guard

$month is unconditionally assigned from $lt[4] on lines 89–90 when it was previously undefined, so by line 101 it is always defined. The defined $month check is dead code. Date::Parse::str2time omits it for the same reason.

Suggested change

	return undef
	unless(defined $month && $month <= 11 && $day >= 1 && $day <= 31
	&& $hh <= 23 && $mm <= 59 && $ss <= 59);
	return undef
	unless($month <= 11 && $day >= 1 && $day <= 31
	&& $hh <= 23 && $mm <= 59 && $ss <= 59);

[atoomic]

@Koan-Bot rebase

[Koan-Bot]

Rebase with requested adjustments

Branch koan.atoomic/fix-language-str2time was rebased onto main and review feedback was applied.

Changes applied

• All three review comments addressed:
• P1 — Missing -1 sentinel and overflow guards: Added the -1 sentinel check and $result < 0 overflow guard to both the timegm and timelocal eval blocks, matching Date::Parse::str2time exactly (including the localtime(-1) approach for the timelocal path and the different year threshold: >= 1970 for timegm, >= 1971 for timelocal).
• P2 — Unused POSIX qw(mktime) import: Removed the unused import from t/lang-str2time.t.
• P2 — Redundant defined $month guard: Removed defined $month && from the validation check since $month is always defined at that point.

Stats

2 files changed, 148 insertions(+), 4 deletions(-)

Actions performed

• Already-solved check: negative (confidence=high, reasoning=None of the recent commits on main address Date::Language::str2time year inference or invalid input )
• Rebased koan.atoomic/fix-language-str2time onto upstream/main
• Applied review feedback
• Pre-push CI check: previous run passed
• Force-pushed koan.atoomic/fix-language-str2time to origin
• CI check enqueued in ## CI (async)

CI status

CI will be checked asynchronously.

---

Automated by Kōan