TDEC-a2 (v2): decoration geometry and the decoration capture, restacked onto the text stack by thejackshelton · Pull Request #109 · compiled-run/dragoncss

thejackshelton · 2026-10-03T10:43:35Z

Stacked, lands as the text-stack train after train 3. This branch is a restack onto INL1a (#88 B1, #89 B2, #90 C1, #91 C2), made by intent: C2 plus this branch's own change. Its base is review/tdec-a1-v2, pushed at tdec-a1-v2's head dbe22e4, so the diff is this branch's alone. The text stack lands as one train after train 3; master is merged in during the train's catch-up. Host-only: no device run yet (pending-device).

What changed

TDEC-a2 (T148 part a), restacked from tdec-a (5a3c7b890): decoration geometry in layout, parity:decoration-capture, and the decorated web-only cases.
Conflicts:
- faults.ts: propagatedIntoOutOfFlow leaves master's NO_FAULTS (the plant is now text-decoration.ts's own fault type).
- decoration-capture.ts skips replaced leaves (REPL-a).
regen.ts gains a decoration-capture step writing packages/parity/expected-decorations/**, and .gitattributes merges those outputs as generated. This is LAND-1: pnpm regen (cached, fixed-point regeneration chain) and a merge policy for generated outputs #58's rule for a branch that adds a generated output.
docs/ports.json:
- layout/text-decoration.ts ports text_decoration_offset.cc, decoration_line_painter.cc, text_fragment_painter.cc, text_painter.cc and SkScan_Antihair.cpp antifilldot8. All are BSD and new.
- It also adds the cited text_decoration_info.cc and simple_font_data.cc ranges.
Regenerated outputs:
- The new decoration step's PNGs come out byte-identical to the committed ones.
- The 12 decorated cases gain their font captures.
- The web profile gains the text-decoration rows.

Tests and pins changed, with reasons

regen.test.ts passes, with a producer for every tracked capture.
No other pin moved.

What passed (host-side, this head)

/tmp/heavy-lease.sh pnpm regen reaches a fixed point; its outputs are in their own commit, whose message names the command.
pnpm typecheck passes.
pnpm test: everything passes except the 12 device-record tests that need a fresh device run (lanes, lanes-records, lanes-concurrent, device-failures, p6a-promotion): device step pending. Load timeouts in the full run (load up to 83): 24 tests timed out, all at 120 s except one at 900 s. In the reruns, 16 files under the heavy lease, then the slowest one at a time: everything passes, inline.test.ts at the default timeout once load dropped (12/12). CI risk: inline.test.ts 'labels inline boxes' took up to 560 s at load 50, so it may time out on CI. No timeout was loosened.

🤖 Generated with Claude Code

…xt px is proven in, and in the border shorthands css/values.ts: a math function in border, border-<side>, border-block(-start/-end) and border-inline(-start/-end) is the <line-width>, typed and refused as the border-*-width longhands are (css-backgrounds-3 §3.1, css-logical-1 §6.3); shorthands/border.ts and logical.ts assign it to the width. Fixtures (values group, ltr and rtl): values-calc-contexts-block, -flex, -positioned, -display-none, -root, -root-min-max put each of calc(), min(), max() and clamp() in every box property in the block, flex-row/column, relative, absolute (cb-ltr and cb-rtl), display-none and root contexts, and gaps in every flex line mode and on a non-flex container, plus px, rem and % insets, flex-basis and gaps on a static block; values-calc-border-shorthand; values-calc-length-max, the CSS_LENGTH_MAX golden (T118J): Chrome collapses float(33554429) with -33554430 to -2 device px at every DPR. values-reject-order-fraction pins the order refusal. Retargets: values-reject-border-shorthand moves from calc(1px + 0.5vw), now accepted, to calc(1px * 2px / 1px) solid #000, still refused (typed arithmetic), refused on border-left-width. values.test.ts blinkNumber writes printf %g (3.35544e+07), which the new computed values need.

…ture, then pnpm run profile:rows and pnpm run parity:capture three times to the fixed point, then pnpm run profile:rows; pnpm run parity:dpr-capture; pnpm run layout:vectors; pnpm run layout:dpr-vectors; pnpm run layout:break-vectors; pnpm run parity:break-capture; pnpm run parity:pixel-capture; pnpm run native:gen; node --conditions=dragon-internal examples/music-player/tools/check.ts (pnpm run north-star:check lacks the condition, T124); pnpm run parity:glyph-b3 -- --write-bottom-pins; node --conditions=dragon-internal packages/parity/src/cli/media-sweep.ts (then --check: 18/18, 286 samples); pnpm run wpt:run -- --target web and pnpm run wpt:update-expectations -- --target web (no change); pnpm run tw:sweep. Profiles: 1922 rows added on each target (web and ios exact, android unsupported), none removed or changed. Tailwind 4.3.3: web and ios 2214 -> 4121 supported, none the other way, 0 mismatch

…nes -- --run-host) for the CALC-2 cases; device lanes not run (device step pending, landing-queue rule)

…h its own height treated as auto (T128) css-flexbox-1 §4.5 content size suggestion; ported from Blink third_party/blink/renderer/core/layout/flex/flex_layout_algorithm.cc at 145.0.7632.6 (BSD, The Chromium Authors; sha256 b0ea68654c62d2ca1d73d801ffca395901aa37e51569281266fc9ef87a9e892c) lines 914-923 and 1117-1120 (LayoutResult::IntrinsicBlockSize). A percentage height child of such an item is refused as percent-height-flex, since Blink resolves it against the item's set height.

…ontainer shorter than its items, px and calc() heights, empty and non-empty content, ltr and rtl; Chrome captures from pnpm run parity:capture and pnpm run parity:dpr-capture

…ize (flex_layout_algorithm.cc lines 1078-1084 at 145.0.7632.6): an item that cannot shrink takes its specified size without measuring its content, which gives the same used size and avoids 4 percent-height-flex refusals

…enerated), rt-hit.ts (Chrome 145 hit test with the T063J inline rule), the hit table, hit capture and report, the states fixture group

…eld-r1: faults.ts keeps both field sets; lanes.ts takes master's SUITE_LINE; grammar.generated.ts regenerated (pnpm run grammar:gen); out/lanes.json taken from master and regenerated later

…ure values-calc-flex-column-auto-min): pnpm run grammar:gen; pnpm run parity:capture and pnpm run profile:rows twice (fixed point); pnpm run parity:dpr-capture; pnpm run layout:vectors; pnpm run layout:dpr-vectors; pnpm run layout:break-vectors; pnpm run parity:break-capture; pnpm run parity:pixel-capture; pnpm run native:gen; node --conditions=dragon-internal examples/music-player/tools/check.ts; pnpm run tw:sweep; pnpm run wpt:run -- --target web; pnpm run wpt:update-expectations -- --target web; pnpm run parity:glyph-b3 -- --write-bottom-pins; node --conditions=dragon-internal packages/parity/src/cli/media-sweep.ts (then --check); pnpm run parity:lanes -- --run-host. Every existing vector and capture is byte-identical; emitted CSS changes only in the compilation-digest header; the new fixture adds its vectors, captures, pixel and break records; corpus.json and corpus-dpr.json carry the new engine digests

…unk); hit-capture identity manifest writer and hit-capture-identity.test.ts

…efused loudly by the hit table, the line ink gate from Skia's run-local glyph bounds, layers in order-modified paint order, layer clips; tap steps on the host runtime; rt-hit, pointer-events, hit report and identity tests; hit-pointer-events covers an absolute pointer-events: none track (the north-star context)

… run parity:hit-capture -- --identity-base 9b60a9e, then grammar:gen, parity:capture, profile:rows, parity:capture, profile:rows, parity:dpr-capture, layout:vectors, layout:dpr-vectors, layout:break-vectors, parity:break-capture, parity:pixel-capture, native:gen, node --conditions=dragon-internal examples/music-player/tools/check.ts, tw:sweep, parity:glyph-b3 -- --write-bottom-pins, node --conditions=dragon-internal packages/parity/src/cli/media-sweep.ts, parity:hit-capture twice; ua:capture earlier changed nothing more). Every capture and emitted file of 9b60a9e differs only by the pointer-events key (hit-capture-identity.test.ts); profiles gain only the pointer-events rows; expected-media changes only valuesCompared (one more longhand per element); RtHit.swift and RtHit.kt are new translated roots

…ed grid, pinned by the grid's sha256 (hit-capture/2): 15 MB to 2 MB

…y:hit-capture, run twice: the second run changes nothing): 468 cases, 703666 points

…rder, and px lengths clamped to the CSS length range Chrome places an rtl in-flow block child at line-left = ((Pb + border-padding) + additional) + margin-left with additional = ((available - inline size) - InlineSum()), so a margin-left near the CSS length minimum saturates and lands the box at x = 1.984375 px where CSS2 (and the engine) ignored it (block_layout_algorithm.cc 2799-2804, 3283-3304 at 145.0.7632.6, BSD). The engine now threads the container's BFC line offset and follows Blink's order for in-flow children (ComputeChildData), new formatting contexts (LayoutNewFormattingContext 2054-2169, no floats), ResolveInlineAutoMargins (length_utils.cc 1555-1574), the stretch size (46-65) and SlowToPhysical. px lengths are clamped after zoom like calc() results already were (measured: width 1e9px is 33554428 px at DPR 1 and 2). Without saturation the result equals the old closed form (pinned by a 2000-case test). New fixture values-length-saturation.

…it.ts (translated, the same function on host and device); P1 corpus hit suite (one line per layout vector at DPR 1, 2, 3 and 2.625: rt-hit harness op, corpus.ts suite hunk, targets.ts declares p1/hit from the vector counts); parity:hit-capture -- --vectors writes the hit facts

…ilder is rt-hit.ts hitTableOf

…ors, then pnpm run native:gen): RtHit.swift and RtHit.kt gain hitTableOf, hitGrid and hitRuns; the harnesses gain the rt-hit operation; corpus.json declares hit 1872; native:swift and native:kotlin give hit 1872/1872 and every other suite unchanged

…-table builder

… inline placement): pnpm run parity:capture, profile:rows, parity:capture, profile:rows (fixed point); pnpm run parity:dpr-capture; pnpm run layout:vectors; pnpm run layout:dpr-vectors; pnpm run layout:break-vectors; pnpm run parity:break-capture; pnpm run parity:pixel-capture; pnpm run native:gen; node --conditions=dragon-internal examples/music-player/tools/check.ts (no change); node --conditions=dragon-internal packages/parity/src/cli/media-sweep.ts (then --check: 18/18, 286 samples; no change); pnpm run parity:glyph-b3 -- --write-bottom-pins (no change); pnpm run wpt:run -- --target web and wpt:update-expectations -- --target web (no change); pnpm run tw:sweep (no change). No existing vector, capture or expectation moves: every changed tracked file is a new fixture's files, an emitted-CSS compilation header (bodies identical), 88 profile rows that gain values-length-saturation proof cases (no status changes), the generated Swift and Kotlin engines, the pixel manifest, and the corpus-dpr lock (+2 vectors-m2, +6 vectors-dpr/engine-dpr/snap-values lines; on identical corpus lines the old engine differs only on the 16 lines built from values-length-saturation-rtl (1 vectors-m2, 3 vectors-dpr, 4 engine-dpr, 8 engine-calc); corpus.json (the milestone-1 corpus with its 20000 random trees) is unchanged).

…nes --run-host) for the values-length-saturation cases; device lanes not run (device step pending, landing-queue rule)

…ve-swift and native-kotlin: hit 1872/1872 appended; intent kept: exact suite sizes, full passes); device-vectors pins p1/hit at 1872 and fails a short run; the identity test checks every manifest hash's shape

…parity:lanes --run-host) on the SELD-R1b head after merging origin/master: layout-vectors-host passes on ios and android with p1/hit 1872/1872 and p1/rt 54588/54588; device lanes not run until the device step

… keeps the engine's Ctx and Piece names (the native support code reads Ctx)

… run native:gen): the engine's Ctx and Piece keep their names; only RtHit and the harnesses change

…once, hitAt answers a point (hitTest is the two); hitRuns and the host compare prepare once per case. The Swift hit suite drops from 65 s to 19 s, so it no longer reaches the 180 s suite timeout when both host lanes run at once

…en); native:swift and native:kotlin keep hit 1872/1872

…t): seams and grid registry order gain pointer-events after grid (and INHERITED gains it); values stays the group before states; rt stays the P1 suite before hit; css-escapes twin count 13778 -> 13799 and the glyph-clearance counts per DPR grow with the eight hit-* cases. targets.ts derives the hit suite's DPR directories from DPRS (no DPR literal). The visiblePainted reject expects Dragon's lowercased message

…to seld-r1: lanes.ts takes master's SUITE_CAUSE; native-swift and native-kotlin pins take master's cause check with rt and hit appended; generated files taken from seld-r1a and regenerated in the next commit

…in: grammar:gen, ua:capture, parity:capture and profile:rows twice, parity:dpr-capture, layout:vectors, layout:dpr-vectors, layout:break-vectors, parity:break-capture, parity:pixel-capture, parity:hit-capture --vectors, native:gen, music-player check.ts, tw:sweep, parity:glyph-b3 --write-bottom-pins, media-sweep.ts, parity:hit-capture, wpt:run and wpt:update-expectations --target web, parity:lanes --run-host) and parity:hit-capture --identity-base 77c641e; north-star errors per target equal master's (android equals ios) and pointer-events has no diagnostic

…nline-tags-faces is refused on ios and android until phase R, ready on web, and the engine lane lays it out; the original no-error compile holds under phase C's lowering (nativeRealFaces); the tags-change-the-faces check reads the engine projection, the faces the engine is given

…h no error under phase C's native lowering, off by default until TXT1a-2 phase R; text-weight.test.ts: the cases before the text-weight group are 533 (520 on the merged C2, TXT1a-2's 11 and T133's 2)

…rts at 533 and font-shorthand.test.ts's cases before its group are 545 (the merged C2's 520, TXT1a-2's 11, T133's 2 and TXT-W1's 12)

…Projection, the engine lane's projection: native refuses the real faces of the decorated cases until TXT1a-2 phase R (review of #104), and the engine lowers them in engine mode

…heavy-lease.sh pnpm regen (fixed point after 2 passes). The generated Swift and Kotlin harnesses list INL1a's break faults once, and lanes.json host lanes pass again

# Conflicts: # packages/translate/harness/harness.ts

…-lanes-v2 merged forward): /tmp/heavy-lease.sh pnpm regen (fixed point after 3 passes). The translated engines and harnesses (shaping and order faults), corpus digests and lanes.json host lanes

thejackshelton added 30 commits September 30, 2026 17:04

packages/parity/out/lanes.json: host-only rewrite (pnpm run parity:la…

192623e

…nes -- --run-host) for the CALC-2 cases; device lanes not run (device step pending, landing-queue rule)

Fixture values-calc-flex-column-auto-min (T128): column flex, fixed c…

d917537

…ontainer shorter than its items, px and calc() heights, empty and non-empty content, ltr and rtl; Chrome captures from pnpm run parity:capture and pnpm run parity:dpr-capture

SELD-R1b WIP: pointer-events longhand (grammar:gen and ua:capture reg…

8e22838

…enerated), rt-hit.ts (Chrome 145 hit test with the T063J inline rule), the hit table, hit capture and report, the states fixture group

Merge origin/master (9b60a9e: #42 lane-speed, #41 WM-0, board) into s…

5663612

…eld-r1: faults.ts keeps both field sets; lanes.ts takes master's SUITE_LINE; grammar.generated.ts regenerated (pnpm run grammar:gen); out/lanes.json taken from master and regenerated later

SELD-R1b: rt-hit.ts becomes a translated root (additive engineRoots h…

0f11b2b

…unk); hit-capture identity manifest writer and hit-capture-identity.test.ts

Hit captures store Chrome's answers run-length encoded over the deriv…

ccd4ac9

…ed grid, pinned by the grid's sha256 (hit-capture/2): 15 MB to 2 MB

Rewrite packages/parity/expected-hit in hit-capture/2 (pnpm run parit…

9f320fb

…y:hit-capture, run twice: the second run changes nothing): 468 cases, 703666 points

SELD-R1b: dragon keeps only the hit facts (HitFact); the hit-table bu…

8729305

…ilder is rt-hit.ts hitTableOf

SELD-R1b tests: the hit report and state tests use the translated hit…

7c9b50e

…-table builder

packages/parity/out/lanes.json: host-only rewrite (pnpm run parity:la…

2bcc33d

…nes --run-host) for the values-length-saturation cases; device lanes not run (device step pending, landing-queue rule)

Retarget the pinned P1 suite lists for the hit suite (translate, nati…

2291b13

…ve-swift and native-kotlin: hit 1872/1872 appended; intent kept: exact suite sizes, full passes); device-vectors pins p1/hit at 1872 and fails a short run; the identity test checks every manifest hash's shape

Rewrite packages/parity/out/lanes.json (/tmp/heavy-lease.sh pnpm run …

336f148

…parity:lanes --run-host) on the SELD-R1b head after merging origin/master: layout-vectors-host passes on ios and android with p1/hit 1872/1872 and p1/rt 54588/54588; device lanes not run until the device step

rt-hit: name its private types HitCtx and HitPiece, so the translator…

d632df0

… keeps the engine's Ctx and Piece names (the native support code reads Ctx)

Regenerate the translated engines after the rt-hit type renames (pnpm…

3bdddfe

… run native:gen): the engine's Ctx and Piece keep their names; only RtHit and the harnesses change

Regenerate the translated engines after prepareHit (pnpm run native:g…

4422b8d

…en); native:swift and native:kotlin keep hit 1872/1872

Merge seld-r1a (R1a on anim-a2-translate on origin/master 77c641e) in…

63ec07d

…to seld-r1: lanes.ts takes master's SUITE_CAUSE; native-swift and native-kotlin pins take master's cause check with rt and hit appended; generated files taken from seld-r1a and regenerated in the next commit

thejackshelton added 26 commits October 3, 2026 21:38

Merge branch 'inl1a-tags-v2' into txt-w1-v2

8d9b541

inl1a-tags.test.ts: text-weight-nested-bold (Lato faces) compiles wit…

9e4ab67

…h no error under phase C's native lowering, off by default until TXT1a-2 phase R; text-weight.test.ts: the cases before the text-weight group are 533 (520 on the merged C2, TXT1a-2's 11 and T133's 2)

Merge txt-w1-v2 into txt-w2-v2; text-weight.test.ts's group slice sta…

dd0bad8

…rts at 533 and font-shorthand.test.ts's cases before its group are 545 (the merged C2's 520, TXT1a-2's 11, T133's 2 and TXT-W1's 12)

Merge branch 'tdec-a1-v2' into tdec-a2-v2

b1a7dad

Merge branch 'txt-w2-v2' into tdec-a1-v2

4ca2682

decoration-capture.ts lays the stripped copy out through engineLayout…

0f85373

…Projection, the engine lane's projection: native refuses the real faces of the decorated cases until TXT1a-2 phase R (review of #104), and the engine lowers them in engine mode

Regenerate outputs for inl1a-lowering after the FAULT_KEYS fix: /tmp/…

7613949

…heavy-lease.sh pnpm regen (fixed point after 2 passes). The generated Swift and Kotlin harnesses list INL1a's break faults once, and lanes.json host lanes pass again

Merge branch 'inl1a-lowering' into txt1a-1a-v2

891a2ce

# Conflicts: # packages/translate/harness/harness.ts

Merge branch 'txt1a-1a-v2' into txt1a-1b-v2

be47487

Merge branch 'txt1a-1b-v2' into txt1a-1-v2

bb5786d

Merge branch 'txt1a-1-v2' into txt1a-2-v2

f951b06

Merge branch 'txt1a-2-v2' into inl1a-tags-v2

a6eebd7

Merge branch 'inl1a-tags-v2' into txt-w1-v2

cbf08f4

Merge branch 'txt-w1-v2' into txt-w2-v2

1fe87b5

Merge branch 'txt-w2-v2' into tdec-a1-v2

032570a

Merge branch 'tdec-a1-v2' into tdec-a2-v2

e420b32

Regenerate outputs for txt1a-1a-v2 after merging inl1a-lowering (seld…

1f8e562

…-lanes-v2 merged forward): /tmp/heavy-lease.sh pnpm regen (fixed point after 3 passes). The translated engines and harnesses (shaping and order faults), corpus digests and lanes.json host lanes

Merge branch 'txt1a-1a-v2' into txt1a-1b-v2

0484a77

Merge branch 'txt1a-1b-v2' into txt1a-1-v2

787c4ef

Merge branch 'txt1a-1-v2' into txt1a-2-v2

fb7734c

Merge branch 'txt1a-2-v2' into inl1a-tags-v2

560cb71

Merge branch 'inl1a-tags-v2' into txt-w1-v2

892233b

Merge branch 'txt-w1-v2' into txt-w2-v2

06d8ec7

Merge branch 'txt-w2-v2' into tdec-a1-v2

2cbdca5

Merge branch 'tdec-a1-v2' into tdec-a2-v2

8384270

Navigation Menu

Repository navigation

TDEC-a2 (v2): decoration geometry and the decoration capture, restacked onto the text stack - #109

TDEC-a2 (v2): decoration geometry and the decoration capture, restacked onto the text stack#109
thejackshelton wants to merge 192 commits into
review/tdec-a1-v2from
tdec-a2-v2

thejackshelton commented Oct 3, 2026

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

thejackshelton commented Oct 3, 2026

What changed

Tests and pins changed, with reasons

What passed (host-side, this head)

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant