Ferrum Files

Added

• Ferrum::Browser option :protocol_timeout bounds individual internal CDP bookkeeping calls, e.g. Target.createTarget, Target.attachToTarget. It's a separate setting from :timeout: its default (5s) happens to match :timeout's own default, but passing timeout: to Browser.new does not change it -- set :protocol_timeout (or the FERRUM_PROTOCOL_TIMEOUT env var) explicitly for a different value.
• Ferrum::Page#pdf/#screenshot accept a timeout: argument (default 60s) for the CDP call itself. Generating a full-page screenshot or a large PDF is a known slow outlier among CDP commands.
• Ferrum::Frame#loader_id provides a loader id when the frame navigates [#583]
• Ferrum::Frame#lifecycle_events provides a list of frame's events like init, networkIdle, firstPaint, etc. [#583]
• Ferrum::Frame#idle? whether frame was loaded [#583]
• page.accessibility API and Node#axnode for reading the CDP accessibility tree
• --no-crashpad is now included in the default Chrome flags, fully disabling the crashpad handler process [#610]
• Support for dedicated/shared Workers and Service Workers [#391], [#388]:
• Ferrum::Worker -- a lightweight connection to a Worker's single execution context, with its own #network, #evaluate/#evaluate_async/#execute, and #on(:request)/#on(:auth) for interception.
• Dedicated and shared workers spawned by a page are discovered and connected to automatically and are reachable through Browser#workers/ Context#workers.
• Service workers are discovered too, through Browser#service_workers/Context#service_workers, but aren't connected to by default. Attaching to one keeps it alive forever, so that's opt-in via Context#attach_target, then target.worker.
• Ferrum::Target#worker?, #shared_worker?, #service_worker?, and #parent_id for telling targets apart.

Fixed

• Ferrum::Client command id generation and Ferrum::Client::WebSocket's driver interactions were not thread-safe, allowing concurrent commands to collide on the same id or corrupt the frame stream; both are now serialized under a mutex [#602]
• Ferrum::Network built duplicate Exchange objects for the same request when Network.requestWillBeSent (regular event thread) and Fetch.requestPaused (priority interruption thread) raced to look up/create the exchange for a request id; the find-or-build is now atomic, fixing intermittently nil Exchange#request on blocked/intercepted requests
• Ferrum::Node#type / Ferrum::Keyboard now trigger the select-all editing shortcut (Ctrl/Cmd+A) by naming the CDP commands field, so selecting and replacing text in inputs and contenteditables works.
• Ferrum::Network::InterceptedRequest#match? no longer coerces string blacklist/whitelist patterns containing regexp metacharacters (e.g. ?, .) into regexps; string patterns are now compared literally instead. [#405], [#604]
• Ferrum::Network::AuthRequest#match? and Ferrum::Dialog#match? received the same fix as [#604] above for consistency. String patterns now compare literally: exact-match for requests/auth requests, substring-match for dialog messages. [#405]
• Target.targetCreated handler for an iframe target reusing an already-connected target raised NoMethodError on Ferrum::Browser#new(flatten: false), since it read session_id off the target's underlying Ferrum::Client, which doesn't expose it in that mode; it now reads Target#session_id directly. [#539]
• Ferrum::Node#select_file now uses the stable backendNodeId instead. [#568], [#611]
• Full-page screenshots no longer resize the window, preventing focus steal on macOS [#580]
• DOM.enable is now has includeWhitespace: "all" to keep track of new line nodes, which previously were resolved to 0, and errored with NodeNotFoundError [#596]
• DOM.requestNode call is moved to the node class and being done lazily, this reduces the number of intermediate node ids sent by backend to frontend [#596]
• Fix context can be nilable in ensure Browser#create_page [#582]
• Ferrum::Page#idling? no longer blocks on loading="lazy" iframes that Chrome never starts loading [#583]

Changed

• Ferrum::PendingConnectionsError and Ferrum::TimeoutError were swallowed even though it happens when a traffic iterator results in an empty array. [#583]
• webrick is no longer a runtime dependency. It is only required by Ferrum::Proxy, so add gem "webrick" to your Gemfile if you use it.
• Logger output for Runtime.consoleAPICalled now includes the console API type and stack trace call frames, not just the argument values [#605]
• Ferrum::Browser option :pending_connection_errors is set to false by default

Removed

• webrick runtime dependency dropped from the gemspec. Ferrum::Proxy still needs it, so add gem "webrick" to your Gemfile if you use the proxy server.