Updated: The crawlers' address lists as published on 2026-10-11; the rules page test counts the shipped ranges

[se7enxweb]

What fails

The weekly Crawler address lists workflow fails when the operators'
published lists differ from the shipped ones. Its last scheduled run did:

https://github.com/cjw-network/request-shield/actions/runs/37308913976

::error::The crawlers' address lists changed: run php bin/update-crawler-lists and commit.
 rules/crawlers.php                          | 65 ++++++++++++++++++++++++-----
 rules/crawlers/anthropic.json               |  8 ++--
 rules/crawlers/google-googlebot.json        |  4 +-

The lists have moved on since then, and an update also breaks a test:
CrawlersTest checks the rules page for the words "verified by its published
address list (26 ranges", and the Anthropic list now has 38.

The fix

php bin/update-crawler-lists, as the workflow asks, on today's main (a92a37b):

• Anthropic: 38 ranges (was 26);
• OpenAI ChatGPT-User: 234 (was 230);
• Google's two lists: unchanged entries, new creation times;
• rules/crawlers.php regenerated from them.

No crawler was added or renamed, so rules/crawlers.rules keeps its version.

tests/CrawlersTest.php now counts the ranges in the shipped
rules/crawlers/anthropic.json instead of naming 26, so the next update does
not break it. The words checked on the page stay the same.

Tested

• php bin/update-crawler-lists again on this branch just before opening the
  PR: every list "unchanged", so the data is current as of 2026-10-11.
• The fork's CI with this commit, on top of the role test fix (Fixed: The HTTP cache's role tests hold what happens without APCu (file-store and S1 legs) #5): the whole
  Tests matrix passes.
  https://github.com/se7enxweb/request-shield/actions/runs/38122084312