text-0: the text oracle method - #68
Conversation
The first entry in fixtures/web-first/fonts/: Ahem from web-platform-tests @ 986881aa, public domain, sha256-pinned. A font here is an identity, not a family name — the text-oracle brief owns the method that consumes it.
How text enters the byte-exact Chromium gate without weakening it: - the gate ladder (deterministic-font byte-exact active; geometry oracle named for real fonts later; thresholded raster parked pending FLIP) - the probe record against Chromium 149: default macOS smoothing dilates ink beyond true coverage; -webkit-font-smoothing:none goes bilevel; over integer-aligned Ahem boxes every raster policy is byte-identical - the admitted numeric domain (integer x/y, font-size % 5, integer anchor starts) with refusal-over-snapping - the hermetic font environment (empty by default, content-hash identity) - the corpus-growth law: probes never committed, cells are a gate not coverage, text contained in its own suite, breadth stays untracked Decides no code; the D-M text stage stays open.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Warning Review limit reached
Next review available in: 9 minutes You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (1)
WalkthroughThe PR documents a proposed deterministic SVG text oracle, including grading rungs, Chromium probe findings, supported inputs, refusal rules, hermetic fonts, corpus governance, and text-1 architecture decisions. ChangesDeterministic SVG text gating
Estimated code review effort: 2 (Simple) | ~10 minutes Possibly related issues
Possibly related PRs
🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (1)
docs/wg/consolidation/text-oracle.md (1)
202-205: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winClarify the
rframeresource-reference rule.Line [203] says that
rframerefuses resource references. Line [205] then says that a glyph run referencing a font key “is one” without naming the noun. State whether the font-key reference is the refused resource reference and explain how outlines-first lowering avoids it.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@docs/wg/consolidation/text-oracle.md` around lines 202 - 205, Clarify the paragraph around the “lowering shape for text-1” section by explicitly stating that the glyph run’s font-key reference is the resource reference refused by rframe, then explain that outlines-first lowering removes or resolves that reference before the shaped-text fact enters rframe.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/wg/consolidation/text-oracle.md`:
- Around line 116-129: Expand the “The admitted numeric domain” contract before
ratifying Rung A: define whether transforms and viewBox scaling are admitted
only when final device-space geometry remains integral, and explicitly admit or
refuse direction="rtl", text-anchor variants, and tspan layout behavior. If
these inputs are not covered, narrow Rung A’s “whole text resolution pipeline”
claim to the stated admitted subset rather than implying broader coverage.
- Around line 156-167: Update the text rendering input contract around the
--font FAMILY=PATH interface and fixtures/web-first/fonts/README.md so every
declared font includes its expected SHA-256. Verify each loaded font’s bytes
against that declared hash before rendering, and fail immediately on mismatch
rather than proceeding with the bake.
- Around line 64-70: The text-oracle documentation generalizes beyond the single
macOS Chromium probe without evidence for other supported rasterizers. Add
matching Linux and Windows re-bake evidence using the same bake posture, or
revise the documented oracle scope and admission claims to limit support to
macOS arm64; update the relevant generalization section alongside the existing
Identity evidence.
---
Nitpick comments:
In `@docs/wg/consolidation/text-oracle.md`:
- Around line 202-205: Clarify the paragraph around the “lowering shape for
text-1” section by explicitly stating that the glyph run’s font-key reference is
the resource reference refused by rframe, then explain that outlines-first
lowering removes or resolves that reference before the shaped-text fact enters
rframe.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 42230ff2-8ee8-4493-ae1d-4d38cee0015c
⛔ Files ignored due to path filters (1)
fixtures/web-first/fonts/ahem.ttfis excluded by!**/*.ttf
📒 Files selected for processing (3)
docs/wg/consolidation/index.mddocs/wg/consolidation/text-oracle.mdfixtures/web-first/fonts/README.md
…oracle brief A scratch spike (rustybuzz + pinned Ahem -> outline path facts -> n0::glyphless on the reftest posture) rendered the em-box, space-advance, and middle-anchor runs byte-identical to the round-2 Chromium captures, deterministic across double renders. The coincidence argument is now measured on both sides of the gate.
- scope the measured oracle identity to macOS arm64 Chromium 149; cross- rasterizer coincidence is a named argument the verify-only harness keeps safe until a re-bake measures it - bind the admitted numeric domain to final device-space geometry (CTM restriction named); construct admission belongs to the slice's statement of record - make the font hash executable: hash-bearing --font declarations verified before rendering, typed refusal on mismatch - name the resource reference the outlines-first lowering resolves away
Merged as #68. The gate ladder, admitted numeric domain, bake posture, hermetic font environment, corpus-growth law, outlines-first lowering, and the resolver crate now bind the text arc's rungs. The D-M shaped-text stage stays open — text-2 is the spike that closes it.
Merged as #68. The gate ladder, admitted numeric domain, bake posture, hermetic font environment, corpus-growth law, outlines-first lowering, and the resolver crate now bind the text arc's rungs. The D-M shaped-text stage stays open — text-2 is the spike that closes it.
1 participant
The first milestone of the text arc, per the shape agreed for it: a PR rather than a direct rung, because its content is owner decisions, not measurements.
What this is
The method for gating text on the SVG engine of record (
websem → rframe → n0) — the statement is docs/wg/consolidation/text-oracle.md. It decides no code, leaves the D-M text stage open, and produces no score (FLIP stays unratified).The probe headline (Chromium 149, the exact bake posture)
-webkit-font-smoothing: none, every probe went bilevel and exact: alphas {0,255}, solid boxes at precisely the predicted Ahem coordinates, byte-stable across repeats.antialiasedproduced byte-identical PNGs tononeover integer-aligned boxes — the deep fact the method rests on: over the admitted numeric domain, every raster policy coincides, so the posture declaration leaks no quality flag into the engine." X X "→ ink exactly[10,30) ∪ [50,70).Decisions to ratify (§ Decisions for the owner)
-webkit-font-smoothing: noneas bake posture carried by text fixtures.--font FAMILY=PATH, content-hash identity.fixtures/web-first/text/; breadth untracked infixtures/local/.textlayout): backend-free rustybuzz + ttf-parser, refuses font discovery/paint/clocks/estimates.Also in this PR
fixtures/web-first/fonts/— Ahem pinned by content hash (b719ecb3…94b8448, public domain, WPT @986881aa), with the directory's README stating the identity law.Once ratified, text-1 (the
<text>proving slice) returns to the direct-to-main rung shape.