Skip to content

Notes for Addon Reviewers

Jump to bottom
marklieberman edited this page May 25, 2017 · 2 revisions

Testing

If you test the extension, the preferred OS is Windows. Everything works in OSX/Linux, but the context menu interferes with gestures when the gesture button is right button. The reason for this is described in this BugZilla entry: https://bugzilla.mozilla.org/show_bug.cgi?id=1360278

Background and Content Scripts

Program flow is generally as follows:

  1. A mouse event is generated in a tab or nested frame and bubbles up via content/mouseEvents.js in the manner described in the README under Working Principle.

  2. content/handler.js is called from mouseGestures.js when a gesture begins, updates, or ends. content/handler.js accumulates mouse movements and performs gesture detection.

  3. When a gesture is detected, content/handler.js sends a message to background/handler.js. The background script looks up the mapping for the gesture and invokes the command. A mapping is a tuple of gesture and command, for example: 'ULR' -> 'scrollToTop'.

  4. Privileged commands are implemented in background/commands.js. Some commands in this file are stubs that forward a message back to content/commands.js via content/handler.js typically because the command needs page/DOM access.

There is some additional messaging related to features like status text and wheel gestures, but the overall program generally follows this flow.

User Scripts

This is a feature similar to Greasemonkey that allows a user to execute user provided JavaScript snippet when a mouse gesture is performed.

User scripts can only be input in the options page. The feature is not enabled until the user dismisses a warning describing the risks of self-XSS attacks. The warning looks like this.

User scripts are evaluated using eval() in content/commands.js. When a user script requires access to privileged APIs, I provide a function called executeInBackground() that can be used to execute code in background/commands.js again with eval(). This is all documented in the wiki article on user scripts.

There are use cases for user scripts that use privileged APIs, for example: I often use quick UD gesture mapped to a script that downloads the image under the mouse using a filename prefixed with the domain.

Options Page

I've received some criticism for using Angular here. Angular is only used for the options page. Angular is never injected into content scripts. I know it is a large framework, but it is the framework with which I am most familiar. This allowed me to develop the extension much faster than any alternative. I think the options page is complex enough to justify using a framework.

In addition to Angular, I am using the following libraries:

  • ui-bootstrap: A pure Angular implementation of Bootstrap 3 components.
  • ACE editor: A syntax-highlighting input box for use with the user scripts feature.
  • ui-ace: Angular directive to integrate ACE editor.
Clone this wiki locally