fix(i18n): RU Fallacies title overflow via lang-ru CSS class (#316) by jsboige · Pull Request #353 · ArgumentumGames/Argumentum

jsboige · 2026-05-24T08:28:24Z

Summary

Russian Fallacies card titles overflow the card edge because Cyrillic
runs ~15-20% wider than Latin. Worst cases clip mid-word off the card
(e.g. НЕЙРОЛИНГВИСТИЧЕСКОЕ ПРОГРАММИРОВАНИЕ on both Tarot and A0/Web
faces). The FR golden master ("PROGRAMMATION NEUROLINGUISTIQUE") and PT
("PROGRAMAÇÃO NEUROLINGUÍSTICA") render fine — this is RU-specific.

Closes part of #316 (RU cards). See scope note below.

Approach — Option A (RU-only, FR golden master untouched)

The localization layer already runs StaticConversions (string replaces
on the mustache template, per destination language) after the field-name
swaps. This PR rides that hook:

AssetConverterConfig.cs — add a StaticConversions entry to the
Fallacies CardSetLocalization that injects a lang-ru marker class
onto the card body wrapper for ru only:
<div class="body"> -> <div class="body lang-ru">.
DoStaticConversions is a no-op for fr/en/pt (no entry), so their
mustache is byte-for-byte untouched.
Argumentum_Fallacies_Face_fr.json / Face_Web_fr.json — add inert
.lang-ru .title rules that only engage when the class is present:
- font-size reduction: 2.2em -> 1.9em (Tarot), 3.5em -> 3.0em (Web), ~-14%
- overflow-wrap: break-word so extreme single Cyrillic words break to a
  new line instead of clipping off-card.
Face_2_fr / Face_3_fr (1.1em titles) are left alone — they don't overflow.
Tests — 6 regression assertions in FallaciesLocalizationTests:
lang-ru is injected for ru only (fr/en/pt unchanged), and the
large-title templates carry the matching .lang-ru .title CSS.

Scope notes

RU cards only. PT Fallacies cards were visually verified against the
worst-case titles — Latin script is similar to FR widths, no overflow. Mind-map SVG
overflow (also mentioned in fix(css): Cyrillic title overflow in RU/PT mind maps and cards (v1.0.0 blocker) #316) is a separate artifact and out of scope here.
Requires a pipeline regen for ru to take effect (the class injection
happens at harvest/localization time).

Validation status

dotnet build clean (0 errors)
21/21 localization tests pass (6 new)
Visual validation PENDING — combined regen (this RU fix + the merged
FR simplification data(fallacies): simplify 27 desc_fr for younger audience #351) then ai-01 Playwright+vision pre-screen, then
@jsboige final inspection (fix(css): remove max-height constraint on Virtues card imageSection #303). Do not merge before the visual gate.

Generated with Claude Code

clusterManager-Myia

RU title overflow fix: lang-ru CSS class injected on card body for Cyrillic (15-20% wider than Latin). EN/PT/FR untouched. Test + template CSS rules guard regression. Clean.

clusterManager-Myia

[Hermes] — COMMENT_WITH_CONCERNS\n\nRisk: LOW (CSS + static conversion for i18n fix)\n\nChecklist:\n1. Security scan: CLEAN\n2. Cross-repo impact: None\n3. Tests: 2 theory methods — injection + template guard\n\nConcerns:\n- — rebase onto main before approval.\n- Font-size reduction 3.5em → 3.0em for : consider verifying this doesn't underflow on short Cyrillic titles (2-3 words).

clusterManager-Myia

[Hermes] — COMMENT_WITH_CONCERNS

Risk: LOW (CSS + static conversion for i18n fix)

Checklist:

Security scan: CLEAN
Cross-repo impact: None
Tests: 2 theory methods — injection + template guard

Concerns:

mergeable: CONFLICTING — rebase onto main before approval.
Font-size reduction 3.5em to 3.0em for .lang-ru .title: consider verifying this does not underflow on short Cyrillic titles (2-3 words).

Cyrillic Fallacies titles run ~15-20% wider than Latin equivalents and clip off-card on tarot/web formats. Inject a `lang-ru` marker class on the card body wrapper for RU only via StaticConversions, and ship inert `.lang-ru .title { font-size: ...; overflow-wrap: break-word }` rules in the overflow-prone templates (Face_fr 1.9em, Face_Web_fr 3.0em). FR/EN/PT mustache stays byte-for-byte untouched (no StaticConversion entry → DoStaticConversions is a no-op for those languages). Surgical reconstruction on top of master (branch was 35 commits behind). Build green; 6 new regression tests pass + 11 prior pass = 17/17. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

clusterManager-Myia

[Hermes] — APPROVED

SHA reviewed: b743da0 (HEAD, changed from previously reviewed e409fc1)

Security scan

✅ CLEAN — no hardcoded credentials.

CI status

✅ All checks pass (GitGuardian, Debug build, Release build).

Previous concerns status (from review on `e409fc1`)

#	Concern	Status
1	mergeable: CONFLICTING	✅ Fixed — PR is now MERGEABLE (surgical reconstruction on top of master)
2	Font-size reduction 3.5em→3.0em for .lang-ru .title	✅ Acceptable — Cyrillic titles are 15-20% wider than Latin, 3.0em is proportional. overflow-wrap: break-word escape hatch guards edge cases.

Review

Surgical reconstruction of RU title overflow fix. StaticConversions injects lang-ru body class for RU only; CSS rules are inert for EN/PT/FR. 6 new regression tests guard injection + template CSS presence. 17/17 tests pass. Architecture is clean and minimal.

APPROVE — all prior concerns resolved, clean fix with proper test coverage.

jsboige · 2026-05-29T17:43:39Z

Disposition coordinateur (jsboige 2026-05-29) : HOLD → Option B

Mise en attente. La fix RU overflow passera par Option B (auto-shrink JS dans CardPen), conforme à la décision actée #303 (Option A −14% = NO-GO, Option B prioritaire).

Cette PR applique une réduction de police −14% (2.2em→1.9em Tarot, 3.5em→3.0em Web), scopée RU-only via la classe lang-ru — c'est la variante backup "A2", pas Option B. Son propre body porte déjà "Do not merge before the visual gate".

Conservée (pas close) car réutilisable :

les 6 tests de régression FallaciesLocalizationTests (injection lang-ru ru-only, CSS .lang-ru .title) restent valides quelle que soit l'option retenue ;
le hook StaticConversions d'injection de classe lang-ru est indépendant du mécanisme de shrink → réutilisable par Option B (auto-shrink ciblé sur .lang-ru).

Repassée en draft. Réévaluation après implémentation Option B + gate visuel ai-01 (Playwright RU sur pires titres) lors de la prochaine régénération.

🤖 ai-01 coordinator

clusterManager-Myia

[Hermes] — APPROVED\n\nClean i18n fix. The three-part approach is well thought out:\n1. CSS .lang-ru .title rules reduce font-size and add overflow-wrap for Cyrillic titles\n2. StaticConversion injects lang-ru body class for RU only (EN/FR/PT untouched)\n3. Comprehensive unit tests verify both the injection and template CSS guards\n\nNo concerns. Merge when CI passes.

jsboige · 2026-05-31T10:25:01Z

Fermée comme superseded par #400.

Conformément à la disposition coordinateur (jsboige, 2026‑05‑29 : « HOLD → Option B (auto‑shrink JS dans CardPen) », décision actée #303), le débordement de titre RU (#316) est désormais corrigé côté CardPen par auto‑shrink JS plutôt que par cette classe CSS lang-ru (-14%, scope RU‑only).

Pourquoi Option B :

Indépendant de la langue et du gabarit — agit sur tout titre qui déborde (RU, mais aussi tout futur libellé long), pas seulement le cyrillique.
No‑op sur les titres qui tiennent — les titres latins ne sont pas réduits.
Un seul point d'ancrage (frame.js, avant la capture domtoimage), aucune retouche de gabarit JSON.

Validé visuellement (Playwright) : RU rogné → entièrement lisible ; latin intact. Détails et tableau de mesures dans #400.

La branche feat/fallacies-ru-title-overflow peut être supprimée.

… (#400) Long non-breaking titles — notably RU Cyrillic compounds such as "НЕЙРОЛИНГВИСТИЧЕСКОЕ ПРОГРАММИРОВАНИЕ" — overflowed the .title box and were clipped mid-word, because the inner `.title > div` uses `overflow-wrap: normal`. This implements Option B from #303/#353 (auto-shrink JS), chosen over a RU-only CSS class so the fix is language- and template-agnostic. frame.js now runs autoFitCardTitles() on each <card> in generateImages(), right after font preload and before the domtoimage capture. For each `.title` it binary-searches the largest font-size that fits horizontally (floor 8px). Titles that already fit are left untouched (no-op), so non-overflowing Latin titles are unaffected. Validated via Playwright on the real Fallacies face template + CSS: - "AD HOMINEM" (Latin): 84.5px unchanged (no-op) - "НЕЙРОЛИНГВИСТИЧЕСКОЕ ПРОГРАММИРОВАНИЕ": 84.5 -> 39.1px, now fits - "ПСЕВДОНАУЧНОСТЬ": 84.5 -> 51.7px, now fits Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

…idation dossier (#433) §2bis updated post-recalibration (#431 merged e9600c0). Replaces the stale pre-recalibration caveat (2834, global 65/85 thresholds, "po-2024 follow-up pending") with the fresh per-CardSet × per-detector breakdown from the recalibrated run (2026-06-03, 3905 imgs): - Construction-signature table: each detector flags by card construction, not defect. White-bg cards (Rules/Scenarii) → WhiteBand+blank-ratio ~96% noise; colored-bg (Fallacies-Web/Virtues) → blank-ratio clean, bottom-sat the signal. Only discriminant detector = bottom-sat, read per-CardSet. - #188 headline: WhiteBand clean on colored cards (Fallacies-Web 0, Virtues 37 scattered singletons, no 8-lang cluster) → no white-band regression on the cards #188 affected. - True exception list (bottom-sat): Rules rules_23 = #250 FIX 2 (proposed); Scenarii ~7 dense cards saturating ~43% uniformly across all 8 langs = card-intrinsic, NOT translation (new eye point); Virtues zh/ru/es/pt length-overflow cluster = #316/#353 family (PR #400 RU auto-shrink). Binary summary stays non-discriminant (3219/3905) — value is the per-CardSet exception filter, not the binary verdict. Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

clusterManager-Myia approved these changes May 24, 2026

View reviewed changes

clusterManager-Myia reviewed May 29, 2026

View reviewed changes

jsboige force-pushed the feat/fallacies-ru-title-overflow branch from e409fc1 to b743da0 Compare May 29, 2026 16:53

clusterManager-Myia approved these changes May 29, 2026

View reviewed changes

jsboige marked this pull request as draft May 29, 2026 17:43

clusterManager-Myia approved these changes May 30, 2026

View reviewed changes

jsboige mentioned this pull request May 31, 2026

fix(cardpen): auto-shrink overflowing card titles before capture (#316) #400

Merged

jsboige closed this May 31, 2026

jsboige deleted the feat/fallacies-ru-title-overflow branch June 1, 2026 21:45

jsboige mentioned this pull request Jun 3, 2026

docs(#140): recalibrated per-CardSet exception list in validation dossier #433

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(i18n): RU Fallacies title overflow via lang-ru CSS class (#316) - #353

fix(i18n): RU Fallacies title overflow via lang-ru CSS class (#316)#353
jsboige wants to merge 1 commit into
masterfrom
feat/fallacies-ru-title-overflow

jsboige commented May 24, 2026

Uh oh!

clusterManager-Myia left a comment

Uh oh!

clusterManager-Myia left a comment

Uh oh!

clusterManager-Myia left a comment

Uh oh!

clusterManager-Myia left a comment

Uh oh!

jsboige commented May 29, 2026

Uh oh!

clusterManager-Myia left a comment

Uh oh!

jsboige commented May 31, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

jsboige commented May 24, 2026

Summary

Approach — Option A (RU-only, FR golden master untouched)

Scope notes

Validation status

Uh oh!

clusterManager-Myia left a comment

Choose a reason for hiding this comment

Uh oh!

clusterManager-Myia left a comment

Choose a reason for hiding this comment

Uh oh!

clusterManager-Myia left a comment

Choose a reason for hiding this comment

Uh oh!

clusterManager-Myia left a comment

Choose a reason for hiding this comment

Security scan

CI status

Previous concerns status (from review on e409fc1)

Review

Uh oh!

jsboige commented May 29, 2026

Disposition coordinateur (jsboige 2026-05-29) : HOLD → Option B

Uh oh!

clusterManager-Myia left a comment

Choose a reason for hiding this comment

Uh oh!

jsboige commented May 31, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Previous concerns status (from review on `e409fc1`)