fix: correct GPS epoch constant in format_G (%G)

[Koan-Bot]

What

Fixes the GPS epoch constant used by %G format specifier from 315993600 to 315964800.

Why

The GPS epoch is midnight UTC on January 6, 1980, but the constant 315993600 corresponds to 08:00 UTC — an 8-hour offset. This causes %G to return a GPS week number off by one for any timestamp falling within the first 8 hours of a GPS week boundary (~4.8% of all timestamps).

How

Single constant fix in format_G (Generic.pm line 235). The existing test data happened to avoid week boundaries, so the bug was invisible.

Testing

• New t/gps-week.t with boundary tests (epoch start, week 0→1 transition, recent dates)
• First commit adds failing tests, second commit fixes them
• Full test suite passes

🤖 Generated with Claude Code

---

Quality Report

Changes: 2 files changed, 38 insertions(+), 1 deletion(-)

Code scan: clean

Tests: skipped

Branch hygiene: clean

Generated by Kōan post-mission quality pipeline

[greptile-apps]

Greptile Summary

This PR fixes a long-standing off-by-one bug in the %G GPS week format specifier caused by an epoch constant that was 8 hours too late (315993600 = 08:00 UTC Jan 6, 1980 instead of 315964800 = midnight UTC Jan 6, 1980). The single-line fix is mathematically correct, and the new test file covers epoch identity, the week 0→1 boundary transition, a known historical reference point, and a recent-date sanity check — all verified independently.

Confidence Score: 5/5

Safe to merge — the constant change is correct and all new test assertions are mathematically verified

The fix is a single well-understood constant correction (28800 seconds = 8 hours) with independent verification. New tests cover the exact failure mode (week-boundary timestamps) and pass with the corrected value. No other code paths are affected.

No files require special attention

Important Files Changed

Filename	Overview
lib/Date/Format/Generic.pm	Single-constant fix: GPS epoch corrected from 315993600 (08:00 UTC Jan 6, 1980) to 315964800 (midnight UTC Jan 6, 1980), eliminating an 8-hour offset that caused off-by-one GPS week numbers at week boundaries
t/gps-week.t	New test file with 6 boundary-focused assertions covering GPS epoch identity, end/start of week 0→1 transition, a known historical reference, and a recent date sanity check; all expected values verified correct

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["time2str('%G', $unix_ts)"] --> B["format_G called\n($_[0]->[9] = raw Unix timestamp)"]
    B --> C["Subtract GPS epoch\n315964800 ✅\n(was 315993600 ❌)"]
    C --> D["Divide by 604800\n(seconds per week)"]
    D --> E["int() truncate\ntoward zero"]
    E --> F["GPS week number returned"]

    subgraph "Old vs New epoch"
        G["315993600\n= Jan 6, 1980 08:00 UTC ❌\n(+8 hours wrong)"]
        H["315964800\n= Jan 6, 1980 00:00 UTC ✅\n(correct GPS epoch)"]
    end

    C -.->|"was"| G
    C -.->|"now"| H

Loading

_{Reviews (1): Last reviewed commit: "fix: correct GPS epoch constant in forma..." | Re-trigger Greptile}