files/en-us/mozilla/add-ons/webextensions/chrome_incompatibilities/index.md
The WebExtension APIs aim to provide compatibility across all the main browsers, so extensions should run on any browser with minimal changes.
However, there are significant differences between Chrome (and Chromium-based browsers), Firefox, and Safari. In particular:
Support for WebExtension APIs differs across browsers. See Browser support for JavaScript APIs for details.
Support for manifest.json keys differs across browsers. See the "Browser compatibility" section on the manifest.json page for more details.
Extension API namespace:
browser namespace. The chrome namespace is also supported for compatibility with Chrome.chrome namespace. (cf. Chrome bug 798169)Asynchronous APIs:
The rest of this page details these and other incompatibilities.
In Firefox and Safari: The APIs are accessed using the browser namespace.
browser.browserAction.setIcon({ path: "path/to/icon.png" });
In Chrome: The APIs are accessed using the chrome namespace.
chrome.browserAction.setIcon({ path: "path/to/icon.png" });
In Firefox and Safari (all versions), and Chrome (starting from Manifest Version 3): Asynchronous APIs use promises to return values.
function logCookie(c) {
console.log(c);
}
function logError(e) {
console.error(e);
}
let setCookie = browser.cookies.set({
url: "https://developer.mozilla.org/",
});
setCookie.then(logCookie, logError);
In Chrome: In Manifest V2, asynchronous APIs use callbacks to return values and {{WebExtAPIRef("runtime.lastError")}} to communicate errors. In Manifest V3, callbacks are supported for backward compatibility, along with support for promises on most appropriate methods.
function logCookie(c) {
if (chrome.runtime.lastError) {
console.error(chrome.runtime.lastError);
} else {
console.log(c);
}
}
chrome.cookies.set({ url: "https://developer.mozilla.org/" }, logCookie);
As a porting aid, the Firefox implementation of WebExtensions supports chrome using callbacks and browser using promises. This means that many Chrome extensions work in Firefox without changes.
[!NOTE] The
browsernamespace is supported by Firefox and Safari. Chrome does not offer thebrowsernamespace, until Chrome bug 798169 is resolved.
If you choose to write your extension to use browser and promises, Firefox provides a polyfill that should enable it to run in Chrome: https://github.com/mozilla/webextension-polyfill.
The Browser support for JavaScript APIs page includes compatibility tables for all APIs that have any support in Firefox. Where there are caveats regarding support for an API method, property, type, or event, this is indicated in these tables with an asterisk "*". Selecting the asterisk expands the table to display a note explaining the caveat.
The tables are generated from compatibility data stored as JSON files in GitHub.
The rest of this section describes the main compatibility issues you may need to consider when building a cross-browser extension. Also, remember to check the browser compatibility tables, as they may contain additional compatibility information.
For notifications.create(), with type "basic":
iconUrl is optional.iconUrl is required.When the user clicks on a notification:
If you call notifications.create() more than once in rapid succession:
notifications.create() callback function is not a sufficient delay to prevent this.Firefox and Chrome include a Proxy API. However, the design of these two APIs is incompatible.
proxy.ProxyConfig object. Depending on Chrome's proxy settings, the settings may contain proxy.ProxyRules or a proxy.PacScript. Proxies are set using the proxy.settings property.
See chrome.proxy for more information on the API.Firefox and Chrome provide incompatible APIs for working with a sidebar.
sidebar_action manifest key and manipulated with the {{WebExtAPIRef("sidebarAction")}} API.side_panel manifest key. The sidePanel API then enables panels to be manipulated.When using tabs.executeScript() or tabs.insertCSS():
To work cross-browser, you can specify the path as an absolute URL, starting at the extension's root, like this:
/path/to/script.js
When calling tabs.remove():
tabs.remove() promise is fulfilled after the beforeunload event.beforeunload.In Firefox:
Requests can be redirected only if their original URL uses the http: or https: scheme.
The activeTab permission does not allow for intercepting network requests in the current tab. (See bug 1617479)
Events are not fired for system requests (for example, extension upgrades or search bar suggestions).
If an extension wants to redirect a public (e.g., HTTPS) URL to an extension page, the extension's manifest.json file must contain a web_accessible_resources key with the URL of the extension page.
[!NOTE] Any website may link or redirect to that URL, and extensions should treat any input (POST data, for example) as if it came from an untrusted source, as a normal web page should.
Some of the browser.webRequest.* APIs allow for returning Promises that resolves webRequest.BlockingResponse asynchronously.
In Chrome: Only webRequest.onAuthRequired supports asynchronous webRequest.BlockingResponse by supplying 'asyncBlocking', through a callback instead of a Promise.
onFocusChanged of the {{WebExtAPIRef("windows")}} API triggers multiple times for a focus change.declarativeContent.RequestContentScript API (which is rarely used and is unavailable in stable releases of Chrome).moz-extension://«random-UUID»/«path». This randomness can prevent you from doing things, such as adding your extension's URL to another domain's CSP policy.web_accessible_resources, it is accessible as chrome-extension://«your-extension-id»/«path». The extension ID is fixed for an extension.web_accessible_resources, this property is unsupported. Firefox extensions can fix their extension ID through the browser_specific_settings.gecko.id manifest key (see browser_specific_settings.gecko)."key" property to pin the extension ID across different machines. This is mainly useful when working with web_accessible_resources.fetch()) to a relative URL (like /api), it is sent to https://example.com/api.window (Firefox bug 1208775). More specifically, the global scope (globalThis) is composed of standard JavaScript features as usual, plus window as the prototype of the global scope. Most DOM APIs are inherited from the page through window, through Xray vision to shield the content script from modifications by the web page. A content script may encounter JavaScript objects from its global scope or Xray-wrapped versions from the web page.window, and the available DOM APIs are generally independent of the web page (other than sharing the underlying DOM). Content scripts cannot directly access JavaScript objects from the web page.element.onclick = xxx overwrites the page's or other extensions' event handlers.To work around this inconsistency, use {{domxref("EventTarget.addEventListener", "addEventListener()")}} to register event listeners. See Firefox bug 1965975 for more information.
window.eval runs code in the context of the page. See Using eval in content scripts.window.eval always runs code in the context of the content script, not in the context of the page.this.{variableName} in one script and then attempting to access them using window.{variableName} in another. This is a limitation created by the sandbox environment in Firefox. This limitation may be removed; see Firefox bug 1208775.In Firefox: Content scripts remain injected in a web page after the user has navigated away. However, window object properties are destroyed. For example, if a content script sets window.prop1 = "prop" and the user then navigates away and returns to the page window.prop1 is undefined. This issue is tracked in Firefox bug 1525400.
To mimic the behavior of Chrome, listen for the pageshow and pagehide events. Then simulate the injection or destruction of the content script.
In Chrome: Content scripts are destroyed when the user navigates away from a web page. If the user clicks the back button to return to the page through history, the content script is injected into the web page.
See {{WebExtAPIRef("tabs.ZoomSettingsScope")}}.
The main manifest.json page includes a table describing browser support for manifest.json keys. Where there are caveats around support for a given key, this is indicated in the table with an asterisk "*". Selecting the asterisk expands the table to display a note explaining the caveat.
The tables are generated from compatibility data stored as JSON files in GitHub.
On Linux and Mac: Chrome passes one argument to the native app, which is the origin of the extension that started it, in the form of chrome-extension://«extensionID/» (trailing slash required). This enables the app to identify the extension.
On Windows: Chrome passes two arguments:
allowed_extensions.allowed_origins.CreateProcess, instead of ShellExecute, to launch the additional process with the CREATE_BREAKAWAY_FROM_JOB flag.Some extension APIs allow an extension to send data from one part of the extension to another, such as {{WebExtAPIRef("runtime.sendMessage()")}}, {{WebExtAPIRef("tabs.sendMessage()")}}, {{WebExtAPIRef("runtime.onMessage")}}, the postMessage() method of {{WebExtAPIRef("runtime.port")}}, and {{WebExtAPIRef("tabs.executeScript()")}}.
The Structured clone algorithm supports more types than the JSON serialization algorithm. A notable exception are (DOM) objects with a toJSON method. DOM objects are not cloneable nor JSON-serializable by default, but with a toJSON() method, these can be JSON-serialized (but still not cloned with the structured cloning algorithm). Examples of JSON-serializable objects that are not structured cloneable include instances of {{domxref("URL")}} and {{domxref("PerformanceEntry")}}.
Extensions that rely on the toJSON() method of the JSON serialization algorithm can use {{jsxref("JSON.stringify()")}} followed by {{jsxref("JSON.parse()")}} to ensure that a message can be exchanged because a parsed JSON value is always structurally cloneable.