Portrait entries (first image taller than wide) get a 4:5 container —
Instagram's proven cap that prevents single photos dominating the screen.
Landscape entries keep 4:3. Aspect-ratio moved from slide to wrap so the
strip inherits it via height:100%.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
- object-fit: contain + ::before blurred background fill on all slides
- dots moved inside photo-wrap, overlaid at bottom with shadow for contrast
- arrows hidden on touch devices via @media (hover: none)
- margin-bottom increased for breathing room below photo block
- trip.html.twig brought up to parity with dailies: same structure,
same PhotoSwipe init, same expand button, same dot overlay
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Move PhotoSwipe CSS from per-entry assets.addCss() (runs after head is
committed) to a single <link> tag at block start. Restore the expand
button as the reliable mobile tap target — it dispatches a synthetic
click on the visible <a> slide, which bubbles to PhotoSwipe's gallery
handler. Merge dot sync into the single module script.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Drops ~80 lines of fragile custom lightbox JS/HTML/CSS in favour of
PhotoSwipe v5 loaded from CDN. Slides are now <a> tags (the pswp-gallery
children) which always fire click events on iOS even inside scroll
containers — the root cause of the mobile tap issue. Pinch-to-zoom and
swipe-to-dismiss come for free. Dot sync kept as a separate vanilla JS
block.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
- Aspect ratio 3:2 → 4:3 (less aggressive crop; closer to phone native)
- Slides become <button> elements with data-full pointing to original image
- Tap/click any photo in the feed opens a full-screen lightbox showing the
uncropped original; prev/next browses all feed photos; Esc/arrows/backdrop
click close
- Dot indicator now syncs with scroll via IntersectionObserver
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Replaces the boolean toggle with a select field offering:
on — connect all consecutive entries (chronological line)
manual — force_connect entries only (user-controlled connections)
intelligent_gpx — suppress connectors where GPX covers both endpoints;
force_connect overrides (original smart logic, restored)
off — no connectors at all; force_connect also ignored
buildJourneySegments gains an optional trackpointsPerFile param used
only by intelligent_gpx mode. renderGpxJourney extracts trackpoints
only when connectMode is intelligent_gpx. dailies.html.twig falls
back from intelligent_gpx → on (mini-map has no GPX tracks to
suppress against).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
autoconnect:true now connects every consecutive entry pair in
chronological order; the old proximity check (suppress where GPX
covers the route) is removed entirely.
- buildJourneySegments: drops allTrackpoints/thresholdKm params;
logic is now force_connect || autoconnect (binary, no GPX math)
- renderGpxJourney: no longer extracts trackpoints; just renders
visual GPX layers then calls buildJourneySegments
- dailies.html.twig: removes GPX URL collection, toGeoJSON CDN load,
and the Promise.all — connectors are now synchronous
- extractTrackpoints/isNearTrack/haversineKm removed (dead code)
- blueprint help text updated
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Adds two configurable toggles to the trip blueprint (Admin2 Trip tab):
- use_gpx: show/hide GPX tracks on all maps (default: enabled)
- autoconnect: draw connector lines between markers (default: enabled)
When use_gpx is off, GPX files are not fetched or rendered on any map
(home, map, trip, dailies). The stats panel in trip.html.twig still
reads GPX_URLS directly and is unaffected.
When autoconnect is off, buildJourneySegments suppresses all
auto-connectors; only entries with force_connect:true still draw a
line — making force_connect behaviour independent of both settings.
Also refactors the inline Promise.all in trip.html.twig to use the
shared renderGpxJourney utility (reducing duplication).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
The home map was drawing an initial addJourneyLine, then trying to remove
layer 'home-journey' in the Promise.all callback — but addJourneyLine names
the layer 'home-journey-line', so removeLayer was a no-op and removeSource
failed (layer still referencing the source), leaving a ghost line on top of
the GPX tracks.
Extract the Promise.all → GPX tracks → buildJourneySegments → addJourneySegments
pattern into MapUtils.renderGpxJourney() and replace both map.html.twig and
home.html.twig with the shared call. No upfront journey line is drawn — the
function handles the no-GPX case correctly via Promise.all([]).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
user/config/media.yaml defines only 'gpx', which replaces the system
media.types instead of merging (blueprint-unaware key replacement). This
causes page.media[hero_image] to return undefined for jpg/png files.
Fallback constructs the hero URL directly from page.url + filename,
matching what shortcode plugins already do. The page.media path is still
tried first so it works correctly if the config is ever fixed.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Apply the .back-pill class to the story footer back link and add
:not(.back-pill) guard to .story-footer a rule to prevent the accent
color override. This ensures the back pill maintains its design system
styling (border, background, text color) in story footers.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WPJztrVGbwic2xTG7G9fjM
Blueprint: end_date format changed to Y-m-d H:i (same as start date) so
Admin2 uses the identical datepicker — avoids ambiguous d-m-Y input being
misread by PHP as m-d-Y.
Template guard: was comparing end_date string against page.date|date('Y-m-d')
which can never match. Now compares date-only parts of both fields:
page.header.end_date|date('Y-m-d') != page.date|date('Y-m-d')
Montalcino live page: removed test end_date '12-09-2026 00:00'.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Montalcino demo story had end_date: 2025-09-06 matching its start date,
causing a '06 – 06 Sep' range display. Removed from both the live page
and the demo source.
Template: added guard so end_date equal to the start date never renders
as a range, even if it appears in frontmatter.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Blueprint: end_date format changed from 'Y-m-d H:i' to 'Y-m-d' — the
demo frontmatter stores dates without a time component so Admin was
failing to parse it and showing the field empty.
Template: three-case smart range formatting:
- Same month & year → 01 – 03 Sep 2025
- Same year only → 01 Jan – 03 Sep 2025
- Different years → 01 Jan 2025 – 03 Feb 2026
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Blueprint: renamed 'Date' to 'Start Date', added optional 'End Date'
field (header.end_date) so it's editable in Admin.
Template: single date → 'd M Y'; range → 'd M Y – d M Y' (full year
on both sides — the old format silently dropped the start year).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Replaced IntersectionObserver (discrete threshold) with a scroll RAF loop
using getBoundingClientRect. Opacity is computed from the fraction of
.story-hero__content still visible above the viewport top — so the nav title
fades in gradually as the hero title slides off the top edge, reaching full
opacity only when the element is completely gone.
Removed CSS transition (no longer needed; per-frame JS update is smooth).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Was visible on load because story-hero__content starts below the
viewport fold (bottom:18% of 140vh hero) — IO fired immediately
as 'not intersecting'. Fixed with hasBeenVisible flag: nav title
only appears after the element has entered then exited the viewport.
Typography changed from DM Sans/500 to DM Serif Display/400 at
--text-lg (1.375rem) with -0.01em tracking — same face as the
hero title, scaled down for the 60px nav bar.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Nav title: absolutely centered in the site-header, fades in via
IntersectionObserver when .story-hero__content scrolls above the fold.
Hidden on mobile (< 640px) where there is no room.
Back-to-top: fixed bottom-right pill, appears after 80% of the hero
viewport is scrolled past, smooth-scrolls to top on click.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Overlay previously hard-hid with display:none when scrollY = 100vh, while
the title was still visible (it exits at ~115vh). Now fades in 0→0.65 over
the first 70vh then fades back out to 0 by 140vh (full hero height).
Scrolly caption changed from full-width to centered pill with
rgba(0,0,0,0.45) background — readable against any image regardless of tone.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vgmzx8VTTTmCskSpQtsLTr
Both stats.html.twig and trip.html.twig previously flattened all GPX
trackpoints into a single masterPts array before computing haversine
distance, elevation, and moving time. This caused the junction between
file N's last point and file N+1's first point to be treated as a real
segment — e.g. Florence→coast (~79 km, ~42 h) for Italy's 3-file demo
data, overstating distance and moving time significantly.
Fix: compute all metrics within each file independently and sum the
results. fileResults collection and callback consumption are unchanged.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WPJztrVGbwic2xTG7G9fjM
m038