# Changelog

> **This is the PUBLIC form of this artifact. Findings are listed in full, but the details of SECURITY findings — which rule fired, in which file, on which line, and how to fix it — are deliberately withheld, and any secret-scanner results are excluded entirely. Where detail is absent here it was REMOVED FOR PUBLICATION; it is not missing from the analysis. The complete artifact is available from the repository owner.**

## Score

- CAI 55 → 61 (+6.2)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.2) — scores are not directly comparable.

## Lenses

- Code Health 65 → 65 (+0.1)
- Architecture 92 → 90 (-1.9)
- Maturity 40 → 64 (+23.6)
- Readiness 65 → 55 (-9.4)
- Security 75 → 65 (-10.1)

## Resolved (18)

- Documentation: no project overview (README.md)
- FixmeComment (lib/ferrum/frame/runtime.rb)
- FixmeComment (lib/ferrum/node.rb)
- High IaC: DS-0029 (docker/Dockerfile)
- High IaC: DS-0029 (docker/Dockerfile)
- High: security finding (details withheld)
- Inconsistent naming for blocking vs non-blocking or parameterized vs default operations. The '!' suffix usually implies a dangerous or side-effect-heavy operation in Ruby, but here it appears to be a no-argument convenience method for the same logical operation as the parameterized version. This is inconsistent with the explicit parameter version.
- Inconsistent use of '!' suffix. In Ruby, '!' often denotes a mutating or dangerous method, but here it seems to indicate a method that raises an error if the execution context is not ready, whereas the non-'!' version might return nil or handle it gracefully. This is a subtle but significant inconsistency in error handling semantics.
- Low IaC: DS-0010 (.devcontainer/Dockerfile)
- Low IaC: DS-0017 (.devcontainer/base.Dockerfile)
- Low IaC: DS-0029 (.devcontainer/Dockerfile)
- Low IaC: WD-DOCKER-0003 (.devcontainer/Dockerfile)
- Low IaC: WD-DOCKER-0003 (.devcontainer/base.Dockerfile)
- Low IaC: WD-DOCKER-0014 (.devcontainer/Dockerfile)
- Low cohesion: Frame (LCOM4 8) (lib/ferrum/frame.rb)
- Off-boarding risk: anonymized user #1
- Redundant methods for retrieving all cookies. `each()` is an iterator, while `all()` likely returns an array. While distinct, having both without clear distinction in documentation or naming convention (e.g., `all_cookies()` vs `each_cookie()`) can be confusing. More importantly, `Cookies.[](name)` retrieves a single cookie, but there is no `Cookies.find(name)` or similar, making the API inconsistent in how single items are retrieved.
- Semantic ambiguity in 'restart'. Browser.restart() likely resets the browser state (clears cookies, caches, etc.) and reuses the process, whereas Process.restart() likely kills the existing process and spawns a new one. Using the same name for distinct lifecycle operations (state reset vs. process lifecycle) is confusing.

## New (12)

- Duplicated block (6 lines × 2) (lib/ferrum/network/auth_request.rb)
- High IaC: DS-0029 (docker/Dockerfile)
- High IaC: DS-0029 (docker/Dockerfile)
- High: security finding (details withheld)
- High: security finding (details withheld)
- High: security finding (details withheld)
- Low cohesion: Frame (LCOM4 10) (lib/ferrum/frame.rb)
- Low cohesion: Options (LCOM4 4) (lib/ferrum/browser/options.rb)
- Medium: security finding (details withheld)
- Off-boarding risk: anonymized user #1
- Repeated repair: lib/ferrum/context.rb (lib/ferrum/context.rb)
- Workflow holding a long-lived secret is unscoped (.github/workflows/docs.yml)

## Changes since last survey

- 59 commits — 37 feature/other, 22 fixes

## By area

- lib/ferrum — 22 commits
- (root) — 19 commits
- .github/workflows — 3 commits
- docs/18-node.md — 2 commits
- spec/browser_spec.rb — 2 commits
- spec/network — 2 commits
- .devcontainer/Dockerfile — 1 commit
- docker/Dockerfile — 1 commit
- docs/7-network.md — 1 commit
- lib/ferrum.rb — 1 commit
- sig/ferrum — 1 commit
- spec/contexts_spec.rb — 1 commit
- spec/node_spec.rb — 1 commit
- spec/support — 1 commit
- spec/unit — 1 commit

## Notable commits

- fix: fix(client): write websocket frames from a writer thread, fixing the JRuby deadlock (#680)
- fix: fix(context): reuse the session of an attached target in attach_target (#668)
- fix: fix(context): skip workers without a session and raise a clear error from Target#worker (#669)
- fix: fix(contexts): detach nested service worker sessions so Chrome can stop them (#666)
- fix: fix(docker): retry snapshot.debian.org downloads for about 2.5 minutes in the chromium backfill
- fix: fix(headers): clear the headers and the user agent they set in Chrome right away
- fix: fix(headers): leave the user agent override alone unless given one
- fix: fix(log): serialize writes to an IO logger so lines from several threads don't mix on JRuby (#683)
- fix: fix(mouse): scroll instantly regardless of scroll-behavior
- fix: fix(network): remove the user agent override once every emulate option is nil
- fix: fix(node): give focus back after focusable? checks the node
- fix: fix(node): give moving? a delay long enough to see slow movement
- fix: fix(node): read moving?'s default delay from FERRUM_NODE_MOVING_CHECK
- fix: fix(node): return true from focusable? when focus can't go back to the previous element
- fix: fix(node): return true from moving? when the node is moving
- fix: fix(node): search within the node's own frame
- fix: fix(page): return the byte count from pdf(path:) like screenshot and mhtml
- fix: fix(page): wake navigation when a loading frame detaches (#672)
- fix: fix(runtime): keep a node first seen inside a Map or Set when it is referenced again (#693)
- fix: fix(runtime): raise JavaScriptError for any thrown value, not only Error instances (#675)
- …and 39 more

## Architecture

- Unchanged — 0 containers · 1 contexts · 0 edges
