Stream long full-page PNG captures

render-repros/long-full-page-capture.html

@@ -0,0 +1,35 @@
+<!doctype html>
+<meta charset="utf-8">
+<title>Long full-page PNG capture reference</title>
+<style>
+  html, body { margin: 0; width: 1000px; }
+  body { height: 17000px; background: rgb(180, 20, 30); }
+  .middle {
+    position: absolute;
+    left: 0;
+    top: 8490px;
+    width: 1000px;
+    height: 30px;
+    background: rgb(20, 160, 40);
+  }
+  .bottom {
+    position: absolute;
+    left: 0;
+    top: 16950px;
+    width: 1000px;
+    height: 50px;
+    background: rgb(20, 40, 200);
+  }
+  .fixed {
+    position: fixed;
+    z-index: 2;
+    left: 0;
+    top: 10px;
+    width: 20px;
+    height: 20px;
+    background: rgb(240, 220, 10);
+  }
+</style>
+<div class="middle"></div>
+<div class="bottom"></div>
+<div class="fixed"></div>

render-repros/probe_long_full_page_chromium.py

@@ -0,0 +1,86 @@
+#!/usr/bin/env python3
+"""Record raw Chromium CDP semantics for the long full-page capture fixture."""
+
+import argparse
+import base64
+import io
+import json
+from pathlib import Path
+
+from PIL import Image
+from playwright.sync_api import sync_playwright
+
+
+def rgb(image, x, y):
+    return list(image.getpixel((x, y)))
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--fixture",
+        type=Path,
+        default=Path(__file__).with_name("long-full-page-capture.html"),
+    )
+    parser.add_argument("--output", type=Path)
+    parser.add_argument("--executable")
+    args = parser.parse_args()
+
+    launch = {"headless": True}
+    if args.executable:
+        launch["executable_path"] = args.executable
+
+    with sync_playwright() as playwright:
+        browser = playwright.chromium.launch(**launch)
+        context = browser.new_context(
+            viewport={"width": 1000, "height": 700},
+            device_scale_factor=1,
+            locale="en-US",
+            timezone_id="UTC",
+            color_scheme="light",
+        )
+        page = context.new_page()
+        page.goto(args.fixture.resolve().as_uri(), wait_until="load", timeout=30_000)
+        page.evaluate("window.scrollTo(0, 8000)")
+        before_scroll = page.evaluate("[window.scrollX, window.scrollY]")
+        cdp = context.new_cdp_session(page)
+        result = cdp.send(
+            "Page.captureScreenshot",
+            {"format": "png", "captureBeyondViewport": True},
+        )
+        after_scroll = page.evaluate("[window.scrollX, window.scrollY]")
+        with Image.open(io.BytesIO(base64.b64decode(result["data"]))) as opened:
+            image = opened.convert("RGBA")
+
+        report = {
+            "browser": "chromium",
+            "version": browser.version,
+            "cdp_method": "Page.captureScreenshot",
+            "parameters": {"format": "png", "captureBeyondViewport": True},
+            "viewport": [1000, 700],
+            "device_scale_factor": 1,
+            "scroll_before": before_scroll,
+            "scroll_after": after_scroll,
+            "dimensions": list(image.size),
+            "pixels": {
+                "top": rgb(image, 500, 100),
+                "middle": rgb(image, 500, 8500),
+                "bottom": rgb(image, 500, 16975),
+                "fixed_at_live_scroll": rgb(image, 10, 8015),
+                "full_page_origin": rgb(image, 10, 15),
+            },
+            "boundary_triplets": {
+                str(boundary): [rgb(image, 900, y) for y in range(boundary - 1, boundary + 2)]
+                for boundary in (4096, 8192, 12288, 16384)
+            },
+        }
+        serialized = json.dumps(report, indent=2, sort_keys=True)
+        if args.output:
+            args.output.write_text(serialized + "\n", encoding="utf-8")
+        print(serialized)
+        context.close()
+        browser.close()
+
+
+if __name__ == "__main__":
+    main()