jpmschweitzer/settled-reach
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(ui): cap rung extents to the body + fit/center the Global canvas (T-1189, T-1192) #205

Closed
jpmschweitzer wants to merge 0 commits from atlas-rung-presentation into main
pull from: atlas-rung-presentation
merge into: :main
Conversation 2 Commits - Files Changed -
jpmschweitzer commented 2026-07-25 12:58:25 +02:00
Owner
Copy Link
Copy Source

Summary

The two client fixes for the Atlas frames Jeroen flagged as looking broken (T-1183 eyeball captures).

T-1189 — Region rung no longer requests more planet than exists. StepCanvasTransport.cap_extent_to_body() caps the viewport-fit extent at the body's own region grid, sourced from the Global rung's echoed canvas size (server-side global_cell_counts() IS the region grid — cols=regions_per_equator, rows=cols/2, no unit conversion). Matched by gridunit SPACING, not rung name (shape-generic — any future rung sharing Global's spacing is covered; finer rungs are no-ops by construction). Applied in _request_extent() before the request/cache key is built, so key discipline from T-1182/T-1183 holds. Cold-start (scroll-to-Region before the first Global echo): request uncapped, server clamps — documented, chosen over fetch-Global-first to avoid blocking the scroll gesture. Kills both the side-by-side continent repeat and the past-the-pole stripe smear.

T-1192 — Global opener fit-scales and centers. One shared letterbox mechanism (center_offset / integer_fit_scale / fit_scale with a 75%-coverage integer-vs-fractional decision) used by BOTH the Global fit and the Region cap's letterbox remainder. Integer scale preferred per D-255 texel-exactness; fractional fallback is legal because orbital rungs already force NEAREST filtering. Legend column reserved before fitting (constant pinned to the legend's own width by a cross-file test) — legend never overlaps the canvas. GJ1c at 1920x1080 → ~1.84x fractional fill; at 4K → integer 4x.

Also fixes a latent crash: NOTIFICATION_RESIZED fires mid-_ready() before children exist — null guard added.

Verification

Touched suites: transport 80/80, viewer 72/72, legend 12/12, annotation-layer 10/10; gdlint clean on all touched files. Full gate: 3,504/3,504 including the SR_LIVE e2e suites (server binary present this run). Reviews next; the before/after eyeball re-capture of the flagged frames runs on the final tip — merge waits on it.

🤖 Generated with Claude Code

## Summary The two client fixes for the Atlas frames Jeroen flagged as looking broken (T-1183 eyeball captures). **T-1189 — Region rung no longer requests more planet than exists.** `StepCanvasTransport.cap_extent_to_body()` caps the viewport-fit extent at the body's own region grid, sourced from the Global rung's echoed canvas size (server-side `global_cell_counts()` IS the region grid — cols=regions_per_equator, rows=cols/2, no unit conversion). Matched by gridunit SPACING, not rung name (shape-generic — any future rung sharing Global's spacing is covered; finer rungs are no-ops by construction). Applied in `_request_extent()` before the request/cache key is built, so key discipline from T-1182/T-1183 holds. Cold-start (scroll-to-Region before the first Global echo): request uncapped, server clamps — documented, chosen over fetch-Global-first to avoid blocking the scroll gesture. Kills both the side-by-side continent repeat and the past-the-pole stripe smear. **T-1192 — Global opener fit-scales and centers.** One shared letterbox mechanism (`center_offset` / `integer_fit_scale` / `fit_scale` with a 75%-coverage integer-vs-fractional decision) used by BOTH the Global fit and the Region cap's letterbox remainder. Integer scale preferred per D-255 texel-exactness; fractional fallback is legal because orbital rungs already force NEAREST filtering. Legend column reserved before fitting (constant pinned to the legend's own width by a cross-file test) — legend never overlaps the canvas. GJ1c at 1920x1080 → ~1.84x fractional fill; at 4K → integer 4x. **Also fixes a latent crash:** NOTIFICATION_RESIZED fires mid-_ready() before children exist — null guard added. ## Verification Touched suites: transport 80/80, viewer 72/72, legend 12/12, annotation-layer 10/10; gdlint clean on all touched files. Full gate: 3,504/3,504 including the SR_LIVE e2e suites (server binary present this run). Reviews next; the before/after eyeball re-capture of the flagged frames runs on the final tip — merge waits on it. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
jpmschweitzer added 2 commits 2026-07-25 12:58:26 +02:00
feat(ui): cap rung extents to the body + fit/center the Global canvas (T-1189, T-1192) c1a97166c5 
The two client fixes for the Atlas frames Jeroen flagged. Region rung no
longer requests more planet than exists: cap_extent_to_body() caps the
viewport-fit extent at the body's own region grid (the Global canvas
extent echo — cols=regions_per_equator, rows=cols/2), matched by gridunit
SPACING not rung name, applied before the request/cache key is built;
cold-start scroll-before-Global-echo requests uncapped and lets the
server clamp (documented). Kills both the side-by-side continent repeat
and the past-the-pole stripe smear.

Global opener now fit-scales and centers through one shared letterbox
mechanism (center_offset/integer_fit_scale/fit_scale, 75%-coverage
integer-vs-fractional decision, NEAREST already forced on orbital rungs
per D-255's escape hatch) also used for the Region cap's letterbox
remainder. Legend column reserved before fitting — pinned to the legend's
own width by a cross-constant test, never overlapping the canvas.

Also fixes a latent crash: NOTIFICATION_RESIZED fires mid-_ready() before
children exist; null guard in the resize path.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
chore(meta): update changelog 915a6790f0 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer added 2 commits 2026-07-25 13:16:31 +02:00
fix(ui): PR #205 review round — integer px-per-gridunit fit replaces fractional (T-1189, T-1192) 374b195594 
The fractional fit branch is deleted, resolving both review findings at
the root: tyre showed its comments cited D-255 for an exception the
record does not contain (the language came from the lead's ticket text,
not governance), and hoshe showed it returned sub-1x for a canvas
exceeding the viewport on one axis. Replacement: fit_scale_ratio()
chooses the largest integer pixels-per-gridunit R fitting both
legend-reserved axes, floored at 1 (over-viewport draws native and
crops like every fixed rung) — the fine-grained integer lattice (GJ1c
1080p -> 9px/gu = 1593x792, ~98% width; 4K -> 20) that makes the
fractional hatch unnecessary. center_offset() floors to whole pixels
(half-pixel centering would blur the texel grid). Doc comments cite the
real sanction (D-255 amendment 2026-07-25, this branch). Legend column
constant is now canonical in transport, read directly by the legend
(was an independently-typed literal); its test asserts real geometry.
New regression test proves _global_body_extent clears on body switch
and never caps another body's requests.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
docs(governance): D-255 amendment — rung-0 integer viewport-fit display ratio 4c44786f99 
Records the T-1192 presentation ruling: the Global opener's display
ratio is viewport-fitted per body to the largest integer px/gridunit
that fits the legend-reserved viewport (floored at 1x1, letterbox
centred), texel-exact by construction; fixed rungs keep their tuned
constants. Explicitly notes that non-integer resting-state scaling
remains unsanctioned and that a PR #205 review finding removed a
fractional-fit branch which had cited this record for an exception it
does not contain — the mis-citation episode is part of the record so it
cannot recur silently.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer commented 2026-07-25 13:16:45 +02:00
Author
Owner
Copy Link
Copy Source

Review: atlas-rung-presentation -> main (type: code)

Reviewers: Hoshe (code quality) + Tyre (architecture). Both REQUEST_CHANGES; all findings verified and addressed in 374b19559 + 4c44786f9.

Tyre (architecture): REQUEST_CHANGES

# File Issue Resolution
1 step_canvas_transport.gd (fit family docs) The fractional-fit escape hatch was justified as "D-255's own phrase/exception" — language that does not exist anywhere in the record, whose actual mandate is unqualified texel-exact. A governance-citation fabrication attached to a real interpretation call never put to the record. (Fractional+NEAREST avoids blur but is not texel-exact: uneven 1px/2px texel duplication.) Fixed at the root — the fractional branch is deleted entirely, replaced by fine-grained integer px-per-gridunit fitting (largest integer R fitting both legend-reserved axes; GJ1c@1080p → 9px/gu = 1593×792, ~98% width — the coarse footprint-multiple lattice was what made fractional look necessary). Texel-exact preserved by construction. The interpretation call that remains (rung-0's display ratio is viewport-fitted per body rather than a fixed tunable) is now recorded as a real D-255 amendment (2026-07-25) in this branch, which also names the mis-citation episode explicitly. Attribution note: the fabricated phrasing originated in the lead's ticket text — the mis-attribution is a shared failure, now structurally recorded.
2 CHANGELOG.md The pre-existing T-1182 "texel-exact / no client-side scaling" claim would be false under the fractional path Resolved by #1 — with fractional gone the claim stays true; the new entry states the Global view fills the frame at an integer texel scale
3 step_canvas_transport.gd:95 / step_canvas_legend.gd LEGEND_COLUMN_PX hand-duplicated as an independently-typed literal on the legend side Fixed — transport is canonical; the legend's RESERVED_COLUMN_PX is a direct read of StepCanvasTransport.LEGEND_COLUMN_PX; the redundant cross-pin test replaced with a real-geometry assertion (legend's actual right edge inside the reserved column)

Hoshe (code quality): REQUEST_CHANGES

# File Issue Resolution
1 step_canvas_transport.gd fit_scale Fractional branch returns sub-1x for a canvas exceeding the viewport on one axis (GJ1c 885px footprint in 348px available → 0.39x), violating the documented never-below-native invariant; reachable at ordinary window widths, untested Fixed by the same root change — fit_scale_ratio() floors at 1 (native + crop/pan, matching every fixed rung); hoshe's exact repro is now a test (348px case → R=1)
2 test_step_canvas_viewer.gd No test proves _global_body_extent resets on body switch — the exact stale-cap failure mode the mechanism must prevent Fixed — new test lands a Global extent for body A, enters body B, asserts the cap source reads ZERO and the next Region request goes out uncapped

Hoshe's verified-clean list (cache-key threading end-to-end, resize null-guard timing, degenerate-request analysis, cold-start fallback) is on record in the review transcript.

Additional in the fix round

  • center_offset() now floors to whole pixels — an odd letterbox remainder was landing the canvas on a .5px boundary, a fractional-pixel blur edge on the rung the mechanism exists to keep sharp.
  • D-255 citation sweep across all four touched sources: every remaining citation verified as a genuine quote against the record body.

Verification (fix round)

transport 92/92, viewer 76/76, legend 12/12, annotation-layer 10/10; gdlint zero warnings on all touched files; full gate re-run green on push. The before/after eyeball re-capture of the originally-flagged frames runs on this tip — merge follows it.

Verdict: CHANGES REQUESTED -> addressed in 374b19559 + 4c44786f9; merge follows the eyeball.

# Review: atlas-rung-presentation -> main (type: code) Reviewers: Hoshe (code quality) + Tyre (architecture). Both REQUEST_CHANGES; all findings verified and addressed in `374b19559` + `4c44786f9`. ### Tyre (architecture): REQUEST_CHANGES | # | File | Issue | Resolution | |---|------|-------|------------| | 1 | step_canvas_transport.gd (fit family docs) | The fractional-fit escape hatch was justified as "D-255's own phrase/exception" — language that does not exist anywhere in the record, whose actual mandate is unqualified texel-exact. A governance-citation fabrication attached to a real interpretation call never put to the record. (Fractional+NEAREST avoids blur but is not texel-exact: uneven 1px/2px texel duplication.) | **Fixed at the root** — the fractional branch is deleted entirely, replaced by fine-grained integer px-per-gridunit fitting (largest integer R fitting both legend-reserved axes; GJ1c@1080p → 9px/gu = 1593×792, ~98% width — the coarse footprint-multiple lattice was what made fractional look necessary). Texel-exact preserved by construction. The interpretation call that remains (rung-0's display ratio is viewport-fitted per body rather than a fixed tunable) is now **recorded as a real D-255 amendment (2026-07-25)** in this branch, which also names the mis-citation episode explicitly. Attribution note: the fabricated phrasing originated in the lead's ticket text — the mis-attribution is a shared failure, now structurally recorded. | | 2 | CHANGELOG.md | The pre-existing T-1182 "texel-exact / no client-side scaling" claim would be false under the fractional path | **Resolved by #1** — with fractional gone the claim stays true; the new entry states the Global view fills the frame at an integer texel scale | | 3 | step_canvas_transport.gd:95 / step_canvas_legend.gd | LEGEND_COLUMN_PX hand-duplicated as an independently-typed literal on the legend side | **Fixed** — transport is canonical; the legend's RESERVED_COLUMN_PX is a direct read of StepCanvasTransport.LEGEND_COLUMN_PX; the redundant cross-pin test replaced with a real-geometry assertion (legend's actual right edge inside the reserved column) | ### Hoshe (code quality): REQUEST_CHANGES | # | File | Issue | Resolution | |---|------|-------|------------| | 1 | step_canvas_transport.gd fit_scale | Fractional branch returns sub-1x for a canvas exceeding the viewport on one axis (GJ1c 885px footprint in 348px available → 0.39x), violating the documented never-below-native invariant; reachable at ordinary window widths, untested | **Fixed by the same root change** — fit_scale_ratio() floors at 1 (native + crop/pan, matching every fixed rung); hoshe's exact repro is now a test (348px case → R=1) | | 2 | test_step_canvas_viewer.gd | No test proves _global_body_extent resets on body switch — the exact stale-cap failure mode the mechanism must prevent | **Fixed** — new test lands a Global extent for body A, enters body B, asserts the cap source reads ZERO and the next Region request goes out uncapped | Hoshe's verified-clean list (cache-key threading end-to-end, resize null-guard timing, degenerate-request analysis, cold-start fallback) is on record in the review transcript. ### Additional in the fix round - `center_offset()` now floors to whole pixels — an odd letterbox remainder was landing the canvas on a .5px boundary, a fractional-pixel blur edge on the rung the mechanism exists to keep sharp. - D-255 citation sweep across all four touched sources: every remaining citation verified as a genuine quote against the record body. ### Verification (fix round) transport 92/92, viewer 76/76, legend 12/12, annotation-layer 10/10; gdlint zero warnings on all touched files; full gate re-run green on push. The before/after eyeball re-capture of the originally-flagged frames runs on this tip — merge follows it. ### Verdict: CHANGES REQUESTED -> addressed in `374b19559` + `4c44786f9`; merge follows the eyeball.
jpmschweitzer commented 2026-07-25 13:27:00 +02:00
Author
Owner
Copy Link
Copy Source

Presentation eyeball — PASS (merge gate)

Live run on the fixed tip 4c44786f9, real client + live server, judged against the known-bad T-1183 baseline frames.

  • Global: centered 1593×792 canvas (symmetric canvas_position=(309,144)), letterboxed, uniform 9×9 texel blocks, legend fully beside the map — vs the old ~885×440 top-left sliver with legend overlap. Integer fit re-confirmed adapting at 1280×720 (~6px/texel, still centered).
  • Region: planet exactly once, no side-by-side repeat, no pole-smear stripes; held_extent=(177,88) confirms the body-grid cap (old shape was 384×216).
  • District: no regression (unaffected by this PR; its content fixes ride the server branch).
  • Cache-key discipline proven live: the old uncapped Region entry (…384,216:0) sits untouched with last_read_at unrefreshed; the new capped request wrote a fresh …177,88:0 entry (genuine miss); Global was a legitimate hit (content identical, layout client-side).
  • Zero anomalies: no errors, no nav pops, InputSwallower had nothing to intercept.

Captures: .cache/screenshots/t1189-92-eyeball/ vs baseline .cache/screenshots/t1183-eyeball-run2/. Merging.

# Presentation eyeball — PASS (merge gate) Live run on the fixed tip `4c44786f9`, real client + live server, judged against the known-bad T-1183 baseline frames. - **Global:** centered 1593×792 canvas (symmetric `canvas_position=(309,144)`), letterboxed, uniform 9×9 texel blocks, legend fully beside the map — vs the old ~885×440 top-left sliver with legend overlap. Integer fit re-confirmed adapting at 1280×720 (~6px/texel, still centered). - **Region:** planet exactly once, no side-by-side repeat, no pole-smear stripes; `held_extent=(177,88)` confirms the body-grid cap (old shape was 384×216). - **District:** no regression (unaffected by this PR; its content fixes ride the server branch). - **Cache-key discipline proven live:** the old uncapped Region entry (`…384,216:0`) sits untouched with `last_read_at` unrefreshed; the new capped request wrote a fresh `…177,88:0` entry (genuine miss); Global was a legitimate hit (content identical, layout client-side). - Zero anomalies: no errors, no nav pops, InputSwallower had nothing to intercept. Captures: `.cache/screenshots/t1189-92-eyeball/` vs baseline `.cache/screenshots/t1183-eyeball-run2/`. Merging.
jpmschweitzer closed this pull request 2026-07-25 13:29:11 +02:00

Pull request closed

This pull request cannot be reopened because the branch was deleted.
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
jpmschweitzer (Jeroen Schweitzer)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: jpmschweitzer/settled-reach#205