test(make-pdf): emoji render gate (pdffonts + pixel proof)

pdftotext is a false oracle for emoji: Skia preserves the Unicode in the text cluster even when the glyph drew as .notdef tofu, so extraction passes on a broken render. The gate instead asserts (1) pdffonts shows an emoji family embedded and (2) pdftoppm rasterizes the page to color (measured ~1650 saturated pixels vs ~0 for tofu). pdfimages is not used: macOS embeds color emoji as Type 3 fonts, so it lists nothing even on a correct render. Adds resolvePopplerTool() (DRY resolver, returns null for clean skips) and a fixture exercising FE0F variation-selector emoji. Skips cleanly when poppler tools or a color-emoji font are unavailable. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-17 15:20:11 +02:00 · 2026-05-29 07:04:46 -07:00
parent 065f7ed9f6
commit afaa40b796
3 changed files with 194 additions and 0 deletions
@@ -114,6 +114,34 @@ export function resolvePdftotext(env: NodeJS.ProcessEnv = process.env): Pdftotex
  ].join("\n"));
 }

+/**
+ * Locate a poppler companion tool (pdffonts, pdfimages, pdftoppm) used by the
+ * emoji render gate. Mirrors resolvePdftotext's resolution order:
+ *   1. $GSTACK_<TOOL>_BIN env override (e.g. GSTACK_PDFFONTS_BIN)
+ *   2. PATH via Bun.which
+ *   3. standard POSIX locations (Homebrew + distro)
+ *
+ * Returns null (does NOT throw) when the tool is missing — the emoji gate skips
+ * cleanly rather than failing on a box without full poppler-utils.
+ */
+export function resolvePopplerTool(
+  tool: "pdffonts" | "pdfimages" | "pdftoppm",
+  env: NodeJS.ProcessEnv = process.env,
+): string | null {
+  const override = resolveOverride(env[`GSTACK_${tool.toUpperCase()}_BIN`], env);
+  if (override) return override;
+
+  const PATH = env.PATH ?? env.Path ?? "";
+  const onPath = Bun.which(tool, { PATH });
+  if (onPath) return onPath;
+
+  for (const dir of ["/opt/homebrew/bin", "/usr/local/bin", "/usr/bin"]) {
+    const candidate = findExecutable(path.join(dir, tool));
+    if (candidate) return candidate;
+  }
+  return null;
+}
+
 function isExecutable(p: string): boolean {
  try {
    fs.accessSync(p, fs.constants.X_OK);
@@ -0,0 +1,154 @@
+/**
+ * Emoji render gate — proves emoji code points render as real color glyphs in
+ * the output PDF instead of .notdef tofu boxes (▯). This is the regression gate
+ * for fix/make-pdf-emoji-tofu.
+ *
+ * Why not just check pdftotext? Because text extraction is a FALSE oracle for
+ * emoji: Skia preserves the Unicode in the text cluster even when the displayed
+ * glyph is .notdef, so pdftotext can report the emoji survived on a render that
+ * actually drew tofu. Verified empirically on macOS — pdftotext extracts 😀
+ * regardless of whether a color font was available.
+ *
+ * Two assertions that DO distinguish a real render from tofu:
+ *   1. pdffonts shows an emoji family embedded in the PDF (the cascade selected
+ *      a real emoji font — AppleColorEmoji as Type 3 on macOS, NotoColorEmoji
+ *      on Linux). Missing-fallback => no emoji font embedded.
+ *   2. pdftoppm rasterizes the page and we count saturated (colored) pixels.
+ *      A color-emoji render has hundreds (measured: ~1650 at 100dpi); a tofu
+ *      render is a monochrome black outline on white (~0 saturated). Tolerant
+ *      threshold, not an exact-pixel fixture diff, to dodge cross-platform AA
+ *      and font-version variance.
+ *
+ * Note: pdfimages -list is intentionally NOT used — macOS embeds color emoji as
+ * Type 3 fonts, so pdfimages lists nothing even on a correct render.
+ *
+ * Gating: runs only when the compiled binary + browse + pdffonts + pdftoppm are
+ * available AND a color-emoji font is installed for Chromium to fall back to.
+ * Skipped cleanly otherwise (local dev before ./setup installs the font).
+ */
+
+import { describe, expect, test } from "bun:test";
+import { execFileSync } from "node:child_process";
+import * as fs from "node:fs";
+import * as path from "node:path";
+
+import { resolvePopplerTool } from "../../src/pdftotext";
+
+const FIXTURE = path.resolve(__dirname, "../fixtures/emoji-gate.md");
+const ROOT = path.resolve(__dirname, "../../..");
+const PDF_BIN = path.join(ROOT, "make-pdf/dist/pdf");
+const BROWSE_BIN = path.join(ROOT, "browse/dist/browse");
+
+// Saturated-pixel floor. Measured ~1650 at 100dpi for the fixture's color
+// emoji; a tofu render yields ~0. 200 sits well clear of both.
+const SATURATED_PIXEL_FLOOR = 200;
+// A pixel is "colored" when its max-min channel spread exceeds this. Black text,
+// gray rules, and white background all stay near 0; color emoji spike high.
+const SATURATION_DELTA = 40;
+
+/** Is a color-emoji font available for Chromium to fall back to? */
+function emojiFontAvailable(): boolean {
+  if (process.platform === "darwin") {
+    return fs.existsSync("/System/Library/Fonts/Apple Color Emoji.ttc");
+  }
+  if (process.platform === "linux") {
+    const fcMatch = Bun.which("fc-match");
+    if (!fcMatch) return false;
+    try {
+      const out = execFileSync(
+        fcMatch,
+        ["-f", "%{color}\n", ":lang=und-zsye:charset=1F600"],
+        { encoding: "utf8" },
+      );
+      return /true/i.test(out);
+    } catch {
+      return false;
+    }
+  }
+  return false;
+}
+
+function prerequisitesAvailable(): { ok: true } | { ok: false; reason: string } {
+  if (!fs.existsSync(PDF_BIN)) return { ok: false, reason: `make-pdf binary missing (${PDF_BIN}). Run bun run build.` };
+  if (!fs.existsSync(BROWSE_BIN)) return { ok: false, reason: `browse binary missing (${BROWSE_BIN}).` };
+  if (!fs.existsSync(FIXTURE)) return { ok: false, reason: `fixture missing (${FIXTURE}).` };
+  if (!resolvePopplerTool("pdffonts")) return { ok: false, reason: "pdffonts not found (install poppler-utils)." };
+  if (!resolvePopplerTool("pdftoppm")) return { ok: false, reason: "pdftoppm not found (install poppler-utils)." };
+  if (!emojiFontAvailable()) return { ok: false, reason: "no color-emoji font installed; run ./setup (Linux) or install one." };
+  return { ok: true };
+}
+
+/** Count pixels in a P6 (binary) PPM whose RGB channel spread exceeds delta. */
+function countSaturatedPixels(ppmPath: string, delta: number): number {
+  const b = fs.readFileSync(ppmPath);
+  let i = 0;
+  const token = (): string => {
+    while (b[i] === 0x20 || b[i] === 0x0a || b[i] === 0x09 || b[i] === 0x0d) i++;
+    const s = i;
+    while (i < b.length && b[i] !== 0x20 && b[i] !== 0x0a && b[i] !== 0x09 && b[i] !== 0x0d) i++;
+    return b.slice(s, i).toString("ascii");
+  };
+  const magic = token();
+  if (magic !== "P6") throw new Error(`expected P6 PPM, got ${magic}`);
+  const w = Number(token());
+  const h = Number(token());
+  token(); // maxval
+  i++; // single whitespace byte after the maxval precedes the pixel block
+  let sat = 0;
+  const total = w * h;
+  for (let p = 0; p < total; p++) {
+    const o = i + p * 3;
+    const r = b[o], g = b[o + 1], bl = b[o + 2];
+    if (Math.max(r, g, bl) - Math.min(r, g, bl) > delta) sat++;
+  }
+  return sat;
+}
+
+describe("emoji render gate", () => {
+  const avail = prerequisitesAvailable();
+
+  test.skipIf(!avail.ok)("emoji render as color glyphs, not tofu", () => {
+    if (!avail.ok) return; // type narrowing
+    const outputPdf = `/tmp/make-pdf-emoji-gate-${process.pid}.pdf`;
+    const ppmPrefix = `/tmp/make-pdf-emoji-gate-${process.pid}`;
+    const ppmPath = `${ppmPrefix}.ppm`;
+    try {
+      execFileSync(PDF_BIN, ["generate", FIXTURE, outputPdf, "--quiet"], {
+        encoding: "utf8",
+        env: { ...process.env, BROWSE_BIN },
+        stdio: ["ignore", "pipe", "pipe"],
+      });
+      expect(fs.existsSync(outputPdf)).toBe(true);
+
+      // 1. An emoji family must be embedded — the cascade found a real emoji
+      //    font instead of falling through to .notdef.
+      const pdffonts = resolvePopplerTool("pdffonts")!;
+      const fontList = execFileSync(pdffonts, [outputPdf], { encoding: "utf8" });
+      if (!/emoji/i.test(fontList)) {
+        process.stderr.write(`\n--- pdffonts ---\n${fontList}\n--- END ---\n`);
+      }
+      expect(/emoji/i.test(fontList)).toBe(true);
+
+      // 2. The page must actually rasterize to color, not a monochrome tofu box.
+      const pdftoppm = resolvePopplerTool("pdftoppm")!;
+      execFileSync(pdftoppm, ["-r", "100", "-singlefile", outputPdf, ppmPrefix], {
+        stdio: ["ignore", "pipe", "pipe"],
+      });
+      expect(fs.existsSync(ppmPath)).toBe(true);
+      const saturated = countSaturatedPixels(ppmPath, SATURATION_DELTA);
+      if (saturated < SATURATED_PIXEL_FLOOR) {
+        process.stderr.write(`\n[emoji-gate] saturated pixels: ${saturated} (floor ${SATURATED_PIXEL_FLOOR})\n`);
+      }
+      expect(saturated).toBeGreaterThanOrEqual(SATURATED_PIXEL_FLOOR);
+    } finally {
+      try { fs.unlinkSync(outputPdf); } catch { /* ignore */ }
+      try { fs.unlinkSync(ppmPath); } catch { /* ignore */ }
+    }
+  }, 30000);
+
+  if (!avail.ok) {
+    test("prerequisites check", () => {
+      console.warn(`[skip] ${avail.reason}`);
+    });
+  }
+});
@@ -0,0 +1,12 @@
+# Emoji rendering gate 😀
+
+This fixture exists to prove that emoji code points render as real color
+glyphs in the output PDF, not as `.notdef` tofu boxes (▯).
+
+Color emoji on one line: 😀 ❤️ 🚀 ✅ 💡
+
+A variation-selector sequence (FE0F) renders color: ❤️ — the bare code point
+❤ is text-style. Both must come from a font in the cascade, never tofu.
+
+Non-emoji Unicode (unchanged, regression guard): em dash —, times ×, arrow →,
+bullet •, ellipsis …