Fix the test that fails for one hour a day, every night the journal flush runs

[springfall2008]

test_energydataservice fails for one hour a day, and the nightly journal flush is scheduled inside that hour.

The bug

today    = datetime.now().strftime("%Y-%m-%d")                    # ← the machine's clock
tomorrow = (datetime.now() + timedelta(days=1)).strftime(...)     # ← the machine's clock
tz_offset = int(my_predbat.midnight_utc.tzinfo.utcoffset(...))    # ← Predbat's clock

The fixture dates come from the host, the offset from Predbat (Europe/London in the harness). When those disagree on the date, the published rates are a day out from the day the code is modelling and rates[1440] doesn't exist — KeyError: 1440.

On a UTC runner during BST they disagree between 23:00 and 00:00 UTC.

Why it matters now

This isn't intermittent flakiness. The nightly /journal-update flush fires at local midnight, which during BST is 23:0x UTC — so it lands in the failing hour every night. #5011 is a docs-only change (one markdown file) that failed CI at 23:12 UTC for exactly this reason. Until BST ends in late October, every nightly journal PR fails the same way.

The fix

Anchor the fixture to my_predbat.midnight_utc — the same clock the next line already reads. Identical trap and identical fix to #4998 for the multi-car IOG tests last week, and the trap already has a bullet in the debug journal ("Build test dates from Predbat's clock, not the machine's").

Verification

Reproduced with TZ=Pacific/Pitcairn, which puts the host a calendar day behind Europe/London:

	before	after
--test energydataservice, shifted TZ	KeyError: 1440	passes
--test energydataservice, normal TZ	passes	passes
--quick, shifted TZ	aborts	all passed, 82.98s

The suite aborts on the first uncaught exception, so this was masking anything later in the run. With it fixed, nothing else in the quick suite fails in that window.

Worth considering separately

Nineteen other test files call datetime.now(). None is currently affected — I checked by running the whole suite in the window — but the pattern is latent, and this is the second instance in a week. A scheduled CI run in a shifted timezone would catch the next one before a maintainer does. I haven't added one here since it's a workflow change with its own cost.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

The change is narrowly scoped to a test fixture, aligns all date/offset inputs to the same clock, and I found no functional or maintainability issues in the updated code.

Pull request overview

This PR fixes a deterministic timezone/date mismatch in the test_energydataservice fixture that caused CI failures for one hour per day during BST, by aligning the fixture’s date source with Predbat’s own configured clock (my_predbat.midnight_utc).

Changes:

• Build today/tomorrow fixture dates from my_predbat.midnight_utc instead of datetime.now() to avoid host-vs-harness day disagreement.
• Add an in-test comment explaining the failure window and why anchoring to Predbat’s clock resolves it.

File summaries

File	Description
apps/predbat/tests/test_energydataservice.py	Anchors test fixture dates to Predbat’s configured local-midnight clock to eliminate the 23:00–00:00 UTC BST failure window.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.