This is part of an AI fuzzing project, where an AI wrote a fuzzer and then triages bugs from the fuzzer and creates fixes. See #77716 for the tracking issue.

What?

This ports existing RTC e2e tests to use WebSockets and adds an additional test.

Why?

I tried adding WebSockets support to the #77716 fuzzer because the world's largest deployment of Gutenberg RTC uses WebSockets. On adding WS support, the fuzzer immediately found what appear to be a large number of distinct bugs. Users are reporting issues and some of the bugs the fuzzer turned up have symptoms that are very similar to issues users have reported. It's likely that some of the bugs that are turned up are issues users are seeing (although we can't tell for sure with the instrumentation we have).

I then tried porting (that is, had my AI agent port) the existing HTTP e2e tests to WS and found that two failed. On applying all outstanding fixes from #77716, this fixed 1.95 of the tests (one test would fail 1 in 20 times in what seems to be a real failure that's non-determinstic for timing reasons).

I'd like to a submit a PR to fix the bug causing that failure, but I think it makes sense to have all of these tests available to run and this is a large enough diff that, although I don't know what the standards/norms are for Gutenberg, my inclination would be to split it out into a separate diff and then have that PR fix this PR.

These tests do not attempt to run in CI because I'm presuming there's some kind of "Chesterton's Fence" reason we don't have WS RTC support in CI today. If that's an incorrect assumption, this PR could be modified to try to run WS tests in CI. I don't know if WS is blocked in CI, but Storybook CI seems to run a TCP server and talk to it and WS runs over TCP, so it seems like it should be do-able?

Testing Instructions

npm run test:e2e:rtc-websocket

Setup, if necessary:

  npm install
  composer install
  npm run wp-env status
  npm run wp-env start   # only if wp-env is not already running
  npm run build

Run non-determistically failing test many times:

  npm run test:e2e:rtc-websocket -- --project=chromium \
    test/e2e/specs/editor/collaboration/websocket/collaboration-same-user-title-reload-loss.spec.ts \
    test/e2e/specs/editor/collaboration/websocket/collaboration-stress.spec.ts \
    --grep "keeps an unsaved same-user title|two users concurrently move list items" \
    --repeat-each=50

As noted above, this PR is test only and doesn't include the fix for the e2e WS tests, so these tests are expected to fail.

Use of AI Tools

As noted above, this is an AI generated PR for an AI fuzzing project.