[HosseinKarami]

Summary

Version 1.3.3. Fixes from the fifth full audit of the large WordPress multisite, with 1.3.2 and the full budget. The site did not change, but some issues and questions came and went from run to run. Also, the report and summary.json gave different counts. The list had 8 items, and the review found 1 more. The details are in CHANGELOG.md.

The tests

• The focus test counts only the points of a control in the viewport. It reads a control that looks hidden a second time, when the position is stable. On the multisite, 5 statuses per run came from one read during a scroll or a transition.
• The watch of the content that changes by itself skips the body and the elements that hold the main content. On the multisite, all 37 outcomes of this test were on the body. The item and the HTML of a changing element have no text, so its ID and its answer stay.

The results from run to run

• A new name of a state no longer makes a false "fixed": the new ID continues the old one, with its status, origin, source, solution, and decision. A refresh corrects such a pair of 1.3.2.
• An old example path with no trailing slash is checked on the page with the slash.
• The pages of earlier results that the crawl did not find get one request each. A page that is gone closes its results. A page that still answers is named in the report.
• The pages of the open questions of the sample tests join the sample, with the reason "open question". The random part still changes in each run, on purpose.
• An axe-core issue needs the sample only in a state.
• When two old questions give one new ID, the answer stays.
• A change of an ID matches only an issue that the run found.

The counts and the refresh

• One function counts the issues for summary.json, history.jsonl, and the report. A new count, replaced, holds the new issues that replace a fixed one.
• A refresh writes the same summary.json.

Test plan

• npm test, with a new test file, false-changes (21 checks).
• The new test on the 1.3.2 code: 17 of its 21 checks fail there. The other 4 are setup checks and one check that passes on both versions.
• The seeded faults: 74 of 74 found.
• A refresh with this branch on a copy of the multisite results of 1.3.2. It changes 6 issue files and no question file. The 2 state issues continue with their origin and solution, their old IDs are "rule changed", and the 2 issues with the old path are resolved. The report columns equal summary.json. A second refresh changes nothing.
• A full audit of the multisite with 1.3.3.