Bug 1405245 - webconsole PageError component read repeat from props;r=nchevobbe MozReview-Commit-ID: C3Vs4Uv6gGv

Bug 1429593 - Part 2: Show that a WebExtension is managing the proxy config setting, r?mstriemer r?jaws Update the general page of about:preferences, as well as the Connection Settings panel, to show when an extension is controlling proxy settings, and allow a user to disable the extension to regain control. MozReview-Commit-ID: HKYPkg78IOK

Bug 1429593 - Part 1: Extract functions for dealing with extensions into a separate file, r?mstriemer r?jaws This moves all the functions that manage/change the UI because of extensions being in control of certain preferences to a separate file, so it can be included in SubDialogs. MozReview-Commit-ID: 7OkEn478Pus

Bug 1433361 - Switch external_host to host_ip, r=maja_zf When running tests in Firefox we don't require that the hostname is actually resolvable since we set up some internal DNS overrides. This means that we don't actually need to have web-platform.test defined in /etc/hosts. As a consequence when we check the server is responding, the check will typically fail for Firefox developers if we try to access by hostname. Previously we made use of a special dance in which external_host held the ip at the point of the check and was later replaced by host to actually run tests. But that changed, and instead we relied on the fact that we (accidentially) didn't actually fail in the case the servers couldn't be reached by host name. However in some cases things still failed because local DNS resolved web-patform.test to some ip address, causing hangs or other badness. In this patch we simply add a host_ip property to the config, and connect to that to test the servers are repopnsing rather than the hostname. We also actually fil if we didn't manage to connect to all the servers. MozReview-Commit-ID: BRIOwybGtjx

Bug 921632 - Re-enable delayed kill broad mozprocess unit test. MozReview-Commit-ID: 93KwNlw84ld

Bug 1422163 - Part 2 - Make a new test for clearing all site data with the new dialog. r=Gijs MozReview-Commit-ID: 6Lt3hPZ1YiO

Bug 1422163 - Part 1 - Make a new confirm dialog for clearing all site data that allows you to clear cache. r=Gijs MozReview-Commit-ID: G9xQXlfT9Ay

Bug 1435658: Deal with appearance changes from / to none correctly. r?mats MozReview-Commit-ID: Fl6VY0rAIiD

Bug 1169290 - Add navigator.webdriver WebIDL attribute. r?peterv,maja_zf This patch adds an enumerable, non-configurable, readonly attribute "webdriver" to the Navigator interface. The attribute is true when the -marionette flag has been passed to Firefox or the marionette.enabled preference is true. Otherwise it is false. The definition of the interface is found in the WebDriver standard: https://w3c.github.io/webdriver/webdriver-spec.html#interface The navigator.webdriver attribute is meant as an indication to web authors that a document is visited by WebDriver. It is important to stress that it is not meant as a fool-proof way to detect that a website is being visited by a browser automation tool, but as a tool for web documents to take alternate code paths. MozReview-Commit-ID: D3qXVKqZG

Bug 1169290 - Allow nsIMarionette to be initialised from C++. r?maja_zf It can be useful to let nsIMarionette be initialised from C++, and this will cause a nsIMarionette.h file with the component's contract ID to be generated. MozReview-Commit-ID: IcFU2B2Wrzu

Bug 1169290 - Make Marionette component safe to load in child process. r?maja_zf The Marionette XPCOM component is loaded once per process, but only ever initialised in the main process. A subprocess that calls nsIMarionette.running will always see it return false because loading the MarionetteComponent class resets the MarionetteComponent#server property to null, causing MarionetteComponent#running to return false. To report the correct value in child processes they need to query the main process for the running state. This patch introduces a synchronous IPC message call to the main process using the child process message manager (CPMM). Because nsIMarionette is currently never used in a subprocess it is considered acceptable to use sync IPC in this case, especially given the circumstances that Marionette instruments the browser and is not tied to any frontend Firefox UX. MozReview-Commit-ID: 93xtZN4MvWq

Bug 1169290 - Reintroduce marionette.enabled pref. r?maja_zf Some time ago it was decided to remove the marionette.enabled preference due to concerns that XPCOM addons would be able to flip arbitrary preferences, potentially causing the remote control protocol to be activiated unintentionally. Since then XPCOM based addons have been removed from Firefox in favour of WebExtensions which do not have the capability to modify preferences. There are no current plans to make WebExtensions able to change preferences: https://wiki.mozilla.org/WebExtensions/FAQ#Does_the_WebExtensions_API_provide_access_to_preferences.3F This patch reintroduces the marionette.enabled preference, which when flipped to true will enable Marionette and spin up the TCP listener. Conversely it will stop the TCP listener and disable the component when switched back to false. The primary motivation for reintroducing it is the forthcoming addition of a new navigator.webdriver WebIDL attribute. The WebIDL interfaces can be annotated with [Pref="<pref name>"] which will automatically generate C++ code for determining whether it should be present in the interface. This is the easiest and least obstrusive way for Marionette to cause the webdriver attribute to appear on the navigator global. MozReview-Commit-ID: 28wN8Z3Q2a6

Bug 1169290 - Handle -marionette flag in observe function. r?maja_zf All the other notifications are handled in the observe function, and it doesn't really make sense as a top-level function on the MarionetteComponent class. MozReview-Commit-ID: 79xWM5SA4vq

Bug 1169290 - Fire remote-active observer notification in component. r?maja_zf It is more correct to fire the remote-active observer notification in the Marionette XPCOM component after the TCP listener has bound to a port, than to fire it inside the implementation of the TCP server. MozReview-Commit-ID: GXaxkQIgE6U

Bug 1169290 - Tell running state from whether server is alive. r?maja_zf This has the advantage that we no longer have to maintain a boolean state manually to tell whether the Marionette server is running. In other words, the Marionette service is running whenever the TCP listener has been started and is alive. MozReview-Commit-ID: 59mxXBGUIOn

Bug 1169290 - Convert MarionetteComponent to a class. r?maja_zf Separate MarionetteComponent from the XPCOM component so it can be made a class, as XPCOM descriptions must override the prototype. This patch makes no behavioural changes to MarionetteComponent as such, apart from converting it to a class. It introduces a new _xpcom_factory called MarionetteFactory, which only job currently is to construct the MarionetteComponent class. We will later in this patch series need the factory to decide what type of nsIMarionette implementation to provide depending on whether Marionette is loaded in the main- or the child process. MozReview-Commit-ID: 72Zb6QEFMK

Bug 1169290 - Define log level conversion table in function scope. r?maja_zf Instead of assigning a Map object to the global scope, this patch moves the table to be initialised inside LogLevel.get. This would cause a slight detriment to performance if it was called a lot, but it isn't, so we instead free up some memory by letting us free the map sooner. MozReview-Commit-ID: 9U4q71d57A

Bug 1169290 - Improve getPrefVal readability. r?maja_zf Instead of assigning each preference value to prefValue and breaking, they can be returned immediately. This improves readability. We also have no need to expose PREF_STRING, PREF_BOOL et al. in the global scope, so these are now scoped to the getPrefVal function. MozReview-Commit-ID: 7LZBgkZ8r08

Bug 1435644: Regenerate the security/manager/ssl test certificates and keys. r=franziskus a=Aryx on a CLOSED TREE For this, I've uncommented the relevant bits in moz.build files, then: ./mach build security/manager for dir in $(rg GeneratedTestCertificate | grep security | cut -d : -f 1); do cp obj-x86_64-pc-linux-gnu/$(dirname $dir)/*.pem $(dirname $dir); done And same with GeneratedTestKey / *.key MozReview-Commit-ID: C2bkSo6YYCU

Bug 1435673 - Strengthen some MediaEngineWebRTCMicrophoneSource asserts. r?padenot MozReview-Commit-ID: HtjndNLWZI8