|
Latest information
- 5.8.1 Added GameWith's PR / tie-up article tile (.gdb-feature_tile-item.is-pr) to the per-site hide rules; it was missed when the other GameWith ad slots were covered in 5.7.1. Verified on the live page (Chromium harness): the PR tile is hidden while every normal article tile stays (PR visible 1->0, normal 8->8), so no content is swallowed. GameWith-only rule, no effect on the corpus sites.
- 5.8.0 The EasyList fetch now also ingests its network domain rules (||host^), not only the cosmetic hide selectors it already pulled. Plain ||host^ entries — no options, exceptions, or wildcards — feed the same AD_HOST_SET index the bundled list uses; previously domain blocking relied on the bundled list alone while only the cosmetics auto-updated. Ingested hosts are cached in sessionStorage next to the cosmetic cache and restored on cache hits, so a re-visit within the 6h window keeps them instead of falling back to the bundled list. Corpus replay stays 10/10 with no content regressions; the per-corpus block increment is small because the existing pattern set already catches most ad requests — the value is automatic tracking of new hosts, which HAR replay cannot measure.
- 5.7.1 Added per-site hide rules for GameWith and Syosetu (Shousetsuka ni Narou), verified against the live DOM. GameWith's "ad-jack" takeover banner and header ad notices are covered; Syosetu's c-ad slots are covered. Game8 needed nothing new — its slots are plain Google Publisher Tag containers already caught by the global rules.
- 5.7.0 Added an auto-update channel. The script pulls a small filter delta (filters.json) from the project's GitHub Pages so new ad hosts and hide selectors can ship between gallery releases without reinstalling — closing the gap against self-updating rivals. Every entry is validated (strings only, length-bounded, no braces so a selector can't inject CSS rules), cached for 6h, and any failed or blocked fetch is ignored so the bundled lists keep working on their own.
- 5.6.0 Added safe no-op stubs for ad-display SDKs (Google Publisher Tag, AdSense, Amazon apstag, Prebid). Their requests are already blocked, so these stubs just stop the page's own code from throwing on the missing SDK or hanging on a bid callback that never fires — which also defeats SDK-presence anti-adblock checks. Queued callbacks are still executed so layout code inside them runs. Because the script runs after the page loads, this helps late / single-page-app / re-check access, not the first page load. It also prunes YouTube player-API ad metadata on the fetch path (response.json() bypasses the JSON.parse hook); strictly scoped to the youtubei endpoints, and the original response is kept untouched when there is no ad data or on any error. Finally, a hidden 'PR' / 'Sponsored' label is now treated as decoy and no longer hides content, cutting false removals.
- 5.5.7 Indexed the ad-host list so the URL check no longer walks all ~650 entries on every request (it runs on every DOM change). The block decision is unchanged — verified identical against every entry plus near-miss controls. The beforeunload blocker now suppresses a dialog only when no form field has been edited, so a page's own unsaved-changes warning is preserved instead of being silently killed.
- 5.5.6 Fixed a host-matching bug that could block or hide unrelated sites: the domain suffix check matched too broadly (a 'media.net' rule could catch 'media.network.example.com', and short names matched anywhere in the host), and a same-length hostname could pick up another site's hide rules. Stopped deleting the generic 'ad' / 'ads' keys from parsed JSON, which broke pages that use those names for non-ad data. Removed a few over-broad global hide selectors (including the .ai-viewport listeners that swallow taps) and merged duplicate per-site rule blocks that had silently disabled the fuller rule sets.
- 5.5.5 Fixed two side effects from the recent updates. Taps were getting swallowed on apexlegends-leaksnews and its subdomains (overwatch2-news, etc.) because the Advanced Ads viewport listener was being hidden. The corocoro manga viewer went blank because a path block intended for one analytics vendor was also catching the site's own event endpoint.
- 5.5.4 Added ad hosts seen on webtoon/manga sites and overseas image-board/matome pages. Scanner now bounds wait time per site so SPA pages that never settle don't stall the run.
- 5.5.3 Added ad networks used on JP manga viewers, overseas gaming media, and EC sites. Marfeel-style trackers with dynamic cloud host names are now caught by URL path.
- 5.5.2 Cross-checked against major filter lists to pick up older established networks, adult ad networks, and the mobile ad SDKs that fire from in-app WebViews.
- 5.5.1 Cross-checked against the ABP-JP list to pick up self-served ad slots on Japanese media sites. Google remarketing endpoints and Recruit's ad network also covered.
- 5.5.0 Coverage extended to IT, gaming, lifestyle, job-board, and matome sites. Added Google Conversion Manager, Adobe AAM/DTM/MC, TripleLift, Yandex and other commonly-seen long-tail networks.
- 5.4.5 Coverage extended to newspapers, business magazines, and news-curation sites. Picked up the broadsheet-side ad networks.
- 5.4.4 Coverage extended to game-strategy and matome blog sites. Picked up many mid-tier JP ad networks.
- 5.4.3 Replayed older traffic captures to grab hosts that had been missed earlier.
- 5.4.2 Replayed corpus captures to pick up hosts that slipped through. Fixed a path-match bug in the URL classifier.
- 5.4.1 Audited recent JP-site traffic captures and added the ad/tracker hosts that were missing.
- 5.4.0 Empty ad shells and IAB-sized empty frames are now removed too, to keep layout from shifting.
- 5.3.2 Added @icon metadata required by the gallery upload form.
- 4.8.1 First gallery release.
|
