fix: report human-readable, calendar-aware elapsed time in memleak message

[jasonb5]

Description

Fixes #5007.

The memory-leak check in CIME/SystemTests/system_tests_common.py reported elapsed model time in "days" using arithmetic that was not actually days. It subtracted two YYYYMMDD-encoded model-date stamps directly, so a 5-year run reported nonsense like 41129 days (the message mixed years ×10000, months ×100, and day-of-month).

Root cause

perf_get_memory_list stores the model date as float(YYYYMMDD) (e.g. 51231.0), and perf_check_for_memory_leak differenced the raw stamps:

elapsed_days = int(memlist[-1][0]) - int(memlist[1][0])  # 51231 - 10102 = 41129

This is cosmetic only — the pass/fail decision uses finalmem/originalmem and the tolerance ratio, both unaffected — but the message is misleading and can send developers debugging in the wrong direction.

Fix

• Decode the YYYYMMDD stamp into (year, month, day) and report elapsed time in years/months/days, dropping zero components, e.g. memleak detected, memory went from 1499.740000 to 1578.320000 in 4 years, 11 months, 29 days.
• Make the arithmetic calendar-aware: the case CALENDAR variable (NO_LEAP or GREGORIAN) is threaded through so a day-borrow uses the true length of the preceding month (including the Gregorian leap-year rule) rather than a flat 30-day approximation.
• New helper _days_in_month(year, month, calendar_type).
• Renamed the formatting helper to _format_elapsed_model_time (describes intent rather than input encoding).

Tests

• TestFormatElapsedModelTime: singular/plural units, every component combination, day/month/year borrows, float stamps, and century-scale runs.
• Calendar-specific cases: NO_LEAP February, Gregorian leap / non-leap / century / 400-year February, and case-insensitive calendar names.
• New TestDaysInMonth class covering 30/31-day months and all February edge cases.
• Updated the memleak integration test to supply CALENDAR and assert the new message.

All test_unit_system_tests.py tests pass (the 2 pre-existing test_generate_baseline failures are unrelated — they spawn a real subprocess). Doctests pass; black (22.3.0) and pylint (4.0.4) clean.

Checklist

• My code follows the style guidelines of this project (black formatting)
• I have performed a self-review of my own code
• My changes generate no new warnings
• I have added tests that exercise my feature/fix and existing tests continue to pass
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding additions and changes to the documentation

[Copilot]

Pull request overview

This PR fixes the MEMLEAK failure message to report human-readable, calendar-aware elapsed model time rather than subtracting raw YYYYMMDD stamps (which previously produced nonsensical “days” values). This improves diagnostic accuracy without changing the leak pass/fail criteria.

Changes:

• Add _days_in_month and _format_elapsed_model_time helpers to decode YYYYMMDD stamps and compute elapsed years/months/days with NO_LEAP vs GREGORIAN February handling.
• Update perf_check_for_memory_leak to use the case CALENDAR when formatting the elapsed time in the MEMLEAK message.
• Expand unit tests to cover formatting and calendar edge cases, and update the existing memleak message assertion.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
CIME/SystemTests/system_tests_common.py	Adds calendar-aware elapsed-time formatting and uses it in the memleak message.
CIME/tests/test_unit_system_tests.py	Converts tests to pytest-style classes and adds/updates unit tests for the new helpers and updated message.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.