hjanuschka

Chrome 150 regression: stale ::after style for :nth-child(... of S):has(...)

Summary

Chrome 150 fails to update a generated ::after pseudo-element after an ancestor class starts matching a selector containing this combination:

:nth-child(3n + 2 of .ad-count):has(+ .LIFY_block)::after

hjanuschka

name	chromium-autoreview-no-ch
description	Autoreview-style Chromium code review without the ch CLI. Uses git/local files and public Chromium Gerrit REST only.
compatibility	No ch dependency. Requires Python 3 and optional local Chromium checkout. Public Gerrit CLs only unless browser/auth cookies are added manually.

Chromium Autoreview Without ch

You are doing practical issue-finding for Chromium-style code review. This is modeled after public Chromium autoreview comments, but does not require the ch

hjanuschka

SDDM greeter screenshot (chromium-test theme, full wallpaper, no overlay)

Decode with: base64 -d sddm.png.b64 > sddm.png

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

hjanuschka

kDRAFT_WasmBranchHinting cleanup

This is a small example of the kind of work I think the SOCBB metadata RFP is asking for.

It is not uploaded to Chromium yet. I am keeping it as a concrete sample while checking whether this category of work is in scope.

The mismatch

Chromium still has this WebDXFeature counter:

hjanuschka

Cross-compiling Chromium for macOS on Linux (with my own RBE)

A while back I wrote about building Chromium for Windows on a Linux box, driven entirely by my own RBE backend — no Google RBE, no Google-hosted toolchain. It worked beautifully and the obvious next question landed in my inbox more than once:

"Cool. Now do macOS."

My gut reaction was: no chance. macOS means Apple's toolchain, xcrun,

hjanuschka

Newsletter Form – Captcha Submit A/B Proof

Short verification of the current integration on https://www.alstertal-einkaufszentrum.de/newsletter-registration/unsubscribe/?subId=… (same code path as Rheinparkcenter and other ECE centers).

TL;DR

The captcha SDK works correctly. The intermittent rejection comes from a custom submit listener that is registered before KROT.interceptForm(form) and calls e.stopImmediatePropagation(). When that branch is taken, the SDK's own submit listener never runs, so no token is computed and no token is sent.

hjanuschka

Rheinparkcenter Newsletter – Captcha Token Missing Fix

Context

On https://www.rheinparkcenter-neuss.de/newsletter-registration/, the current integration sometimes logs a generated captcha solution but the final POST request does not include it.

Root cause is the custom click-based orchestration around KROT.getSolution() plus KROT.interceptForm(form, true).

Existing problem points

Current code:

• intercepts button click instead of stable form submit semantics

hjanuschka

Captcha.eu Joomla plugin (J6 compatibility build, commit d741158)

This gist contains a Base64-encoded ZIP because GitHub Gist does not accept binary uploads via API/gh CLI.

Decode command (macOS/Linux):

  base64 -d captcha_eu-j6-compat-1.0.2-root.zip.base64 > captcha_eu-j6-compat-1.0.2-root.zip

SHA256:
  98b4b226dc51d4fb02ee795fbb750ea77b497215f5010bf47cc94312bdf595b1

hjanuschka

// Source: chromium/src/third_party/blink/renderer/platform/image-decoders/jxl/jxl_image_decoder_test.cc
// Regression test for crbug.com/498155364.

// With incremental input, basic image info should not become observable before
// the embedded ICC profile is parsed.
TEST_F(JXLImageDecoderTest, DecodeCmykIccProfileIncrementalNoCrash) {
  scoped_refptr<SharedBuffer> full_data =
      ReadFileToSharedBuffer(kJxlTestDir, "conformance_cmyk_layers.jxl");
  ASSERT_TRUE(full_data);
  Vector<char> full_data_vec = full_data->CopyAs<Vector<char>>();

hjanuschka

#!/usr/bin/env bash
set -euo pipefail

# Usage:
#   ./render_i480978106_macos.sh [page_path_or_url] [output_png]
#
# Examples:
#   ./render_i480978106_macos.sh
#   ./render_i480978106_macos.sh ~/my-host/static-files/i-480978106/index.html
#   ./render_i480978106_macos.sh https://static.januschka.com/i-480978106/