[craig-o-curtis]

The landing-page sky now shows everything in its real place, not just the stars and moon, and names what you point at.

Changes

• Helsinki aurora: the curtains are farther north (750/1000 km), so the aurora sits low over the skyline. Its lower edge is about 4° up and the tallest rays reach about 17°, a third of Ísafjörður's height.
• Planets: Mercury to Saturn are placed from JPL's Keplerian elements, with real brightness, and barely twinkle.
• Sun: drawn only when it is really up and in frame. That's summer evenings at Helsinki and summer nights at Ísafjörður, often behind the mountains there. It's an ice-blue disc at true size with a soft glow, a faint, patchy 22° halo and a pillar when low. The sky stays dark.
• Moon: it gets the same soft glow as the sun, scaled by phase. It's drawn at half brightness while the sun is up.
• Solar eclipses: the moon covers the sun, showing a crescent, then the diamond ring, then an ice-blue corona at totality.
• Lunar eclipses: a soft penumbra, then a dim umbra in the theme's purple (copper in real life) with a turquoise fringe at its edge.
• Accuracy: eclipses need much tighter positions, so the orbits now run on terrestrial time (ΔT), the moon uses the full Meeus tables, and the sun gets aberration plus fitted perturbation terms.
• Tooltips: hover or tap near the sun, moon, a planet or one of 324 IAU-named stars to see its name, e.g. "Capella · Auriga". The worker does the lookup; the main thread only places the label.
• Popover: it now says it is always night here, but the stars, planets, moon and sun are where they really are. There are new credits for the planets and star names.

Verification

• Checks: pnpm lint, pnpm check and pnpm build pass on every commit.
• Against astronomy-engine:
  • Moon within 2–7″.
  • Sun–moon separation within 17″ over 40 years.
  • Planets within 5′, magnitudes within 0.2.
• Solar eclipse, 2026-08-12: total at Ísafjörður, as it was. Helsinki comes out 79.1% covered, against the reference's 79.6%.
• Lunar eclipses: 2025-09-07, 2026-03-03 and 2026-08-28 match the published umbral magnitudes (1.362, 1.151, 0.930).
• Visual checks: checked in headless Chromium (Metal) by shifting the worker clock to summer sun times and to both eclipses. The preview code is not in these commits.
• Tooltips: a pointer sweep finds 28 named stars over Ísafjörður and 4 over Helsinki right now. Helsinki has fewer because city light limits it to magnitude 4.2 and the cards cover part of the sky.

[baldurpan]

This is careful work. I checked the parts most likely to go wrong, and they hold up:

• the full Meeus 47.A and 47.B tables (60 rows each, with E ** |M| applied to the new M = ±2 rows)
• the umbra and penumbra radii
• uncovered()
• the tooltip's projection and magnitude cut, which match project() and STAR_VS

I've left two inline comments. The first one matters: a tap never shows a tooltip on a touch screen. The second is cosmetic.

[craig-o-curtis]

Thanks for the careful review, @baldurpan. Both inline points are fixed in 4513c0c and pushed: touch taps now show labels, and labels stay inside the window at the sides. Replies are on each thread. Lint, format, type check and build pass locally; CI is running.

[baldurpan]

Approved at 4513c0c. Both comments are fixed: a tap now names what is under it on touch screens, and labels stay inside the window at the sides. CI is green.