Validator for
Python Other
Pull request Compare This branch is 1206 commits behind mozilla:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

README.rst Validator

The AMO Validator is a tool designed to scan Mozilla add-on packages for problems such as security vulnerabilities, exploits, spamware and badware, and lots of other gunk. By using a combination of various techniques and detection mechanisms, the validator is capable of being both efficient as well as thorough.



Python Libraries:

  • argparse
  • cssutils
  • rdflib
  • fastchardet

Python Libraries for Testing:

  • nose
  • coverage

You can install everything you need for running and testing with

pip install -r requirements.txt


The validator may require some submodules to work. Make sure to run

git clone --recursive git://

so that you get all of the goodies inside.


A working copy of Spidermonkey (debug or non-debug is fine) is a required. The version installed must include support for the Parser API. Downloading and installing the latest copy of Tracemonkey from will include this API.:

hg clone
cd tracemonkey/js/src

HINT: if you're using brew on Mac OS X you can get autoconf2.13 with this:

brew install

Once Spidermonkey is installed, make sure the js excutable is on $PATH. If so, you're good. Otherwise you can set a custom path to the js executable as $SPIDERMONKEY_INSTALLATION in your environment.


Run the validator as follows

python addon-validator <path to xpi> [-t <expected type>] [-o <output type>] [-v] [--boring] [--selfhosted] [--determined]

The path to the XPI should point to an XPI file.

-t The type that you expect your add-on to be detected as. The list of types is listed below.
-o The type of output to generate. Types are listed below.
-v Enable verbose mode. Extra information will be displayed in verbose mode, namely notices (informational messages), Jetpack information if available, extra error info (like contexts, file data, etc.), and error descriptions. This only applies to -o text.
--selfhosted Disables messages that are specific to add-ons hosted on AMO.
--boring Disables colorful shell output.
--determined Continue validating the remaining tiers of an add-on if one tier has failed. Certain high-tiered tests may inadvertently fail when this option is enabled for badly malformed add-ons.
 Accepts a JSON string containing an object whose keys are GUIDs and values are lists of version strings. In the targetApplication and compatibility tests, the add-on's predefined <em:targetApplication> values will be overridden if its GUIDs match thoes from the JSON. E.g.: {"{ec8030f7-c20a-464f-9b0e-13a3a9e97384}": "5.*"}
 Accepts a JSON string containing an object whose keys are GUIDs and values are lists of version strings. If this list is specified, non-inlinecompatibility tests will only be run if they specifically target the applications and veresions in this parameter. E.g.: {"{ec8030f7-c20a-464f-9b0e-13a3a9e97384}": ["6.*"]}

Expected Type:

The expected type should be one of the following values:

any (default)
Accepts any extension
Accepts only extensions
Accepts only themes
Accepts only dictionaries
Accepts only language packs
Accepts only OpenSearch XML files (unpackaged)
Accepts only multi-item XPI packages

Specifying an expected type will throw an error if the validator does not detect that particular type when scanning. All addon type detection mechanisms are used to make this determination.

Output Type:

The output type may be either of the following:

text (default)
Outputs a textual summary of the addo-on analysis. Supports verbose mode.
Outputs a JSON snippet representing a full summary of the add-on analysis.


Text Output Mode:

In text output mode, output is structured in the format of one message per line. The messages are prefixed by their priority level (i.e.: "Warning: This is the message").

At the head of the text output is a block describing what the add-on type was determined to be.

JSON Output Mode:

In JSON output mode, output is formatted as a JSON snippet containing all messages. The format for the JSON output is that of the sample document below.

    "detected_type": "extension",
    "errors": 2,
    "warnings": 1,
    "notices": 1,
    "success": false,
    "compatibility_summary": {
        "errors": 1,
        "warnings": 0,
        "notices": 0
    "ending_tier": 4,
    "message_tree": {
        "module": {
            "function": {
                "error": {
                    "__messages": ["123456789"],
                    "__errors": 1,
                    "__warnings": 0,
                    "__notices": 0
                "__messages": [],
                "__errors": 1,
                "__warnings": 0,
                "__notices": 0
            "__messages": [],
            "__errors": 1,
            "__warnings": 0,
            "__notices": 0
        "__messages": [],
        "__errors": 1,
        "__warnings": 0,
        "__notices": 0
    "messages": [
            "uid": "123456789",
            "id": ["module", "function", "error"],
            "type": "error",
            "message": "This is the error message text.",
            "description": ["Description of the error message.",
                            "Additional description text"],
            "file": ["chrome/foo.jar", "bar/zap.js"],
            "line": 12,
            "column": 50,
            "context: [
                "   if(foo = bar())",
                "       an_error_is_somewhere_on_this_line.prototy.eval("whatever");",
            "compatibility_type": "error",
            "for_appversions": {
                "{ec8030f7-c20a-464f-9b0e-13a3a9e97384}": ["5.0a2", "6.0a1"]
            "tier": 2
    "metadata": {
        "name": "Best Add-on Evar",
        "version": "9000",
        "guid": ""

The message_tree element to the document above contains a series of JavaScript objects organized into a tree structure. The key of each element in the tree is the the name of each successive part of the validator that generated a particular message or set of messages (increasing in specificity as the depth of the tree increases). Each tree element also includes a series of additional nodes which provide extra information:

__errors - number - The number of errors generated in this node
__warnings - number - The number of warnings generated in this node
__notices - number - The number of messages generated in this node
__messages - list - A list of UIDs from messages in the `messages` node

JSON Notes:

File Hierarchy

When a subpackage exists, an angle bracket will delimit the subpackage name and the message text.

If no applicable file is available (i.e.: when a file is missing), the file value will be empty. If a file value is available within a subpackage, then the file attribute will be a list containing the name of the outermost subpackage's name, followed by each successive concentric subpackage's name, followed by the name of the file that the message was generated in. If no applicable file is available within a subpackage, the file attribute is identical, except the last element of the list in the file attribute is an empty string.

For instance, this tree would generate the following messages:

    |  |
    |  |-subsubpackage.xpi
    |     |
    |     |-chrome.manifest
    |     |-install.rdf
    "type": "notice",
    "message": "<em:type> not found in install.rdf",
    "description": " ... ",
    "file": "install.rdf",
    "line": 0
    "type": "error",
    "message": "Invalid chrome.manifest subject: override",
    "description": " ... ",
    "file": "chrome.manifest",
    "line": 7
    "type": "error",
    "message": "subpackage.xpi > install.rdf missing from theme",
    "description": " ... ",
    "file": ["subpackage.xpi", ""],
    "line": 0
    "type": "error",
    "message": "subpackage.xpi > subsubpackage.xpi > Invalid chrome.manifest subject: sytle",
    "description": " ... ",
    "file": ["subpackage.xpi", "subsubpackage.xpi", "chrome.manifest"],
    "line": 5
Line Numbers and Columns

Line numbers are 1-based. Column numbers are 0-based. This can be confusing from a programmatic standpoint, but makes literal sense. "Line one" would obviously refer to the first line of a file.


The context attribute of messages will either be a list or null. Null contexts represent the validator's inability to determine surrounding code. As a list, there will always be three elements. Each element represents a line surrounding the message's location.

The middle element of the context list represents the line of interest. If an element of the context list is null, that line does not exist. For instance, if an error is on the first line of a file, the context might look like:

    "This is the line with the error",
    "This is the second line of the file"

The same rule applies for the end of a file and for files with only one line.


Unit tests can be run with

fab test

or, after setting the proper python path:


However, to turn run unit tests with code coverage, the appropriate command would be:

nosetests --with-coverage --cover-package=validator --cover-skip=validator.outputhandlers.,validator.main,validator.constants,validator.constants_local --cover-inclusive --cover-tests

Note that in order to use the --cover-skip nose parameter, you must install the included patch for nose's plugin:


This file should overwrite the standard nose coverage plugin at the appropriate location:

~/.virtualenvs/[virtual environment]/lib/pythonX.X/site-packages/nose/plugins/


Some regular maintenance needs to be performed on the validator in order to make sure that the results are accurate.

App Versions

A list of Mozilla <em:targetApplication> values is stored in the validator/app_versions.json file. This must be updated to include the latest application versions. This information can be found on AMO:

JS Libraries

A list of JS library hashes is kept to allow for whitelisting. This must be regenerated with each new library version. To update:

cd extras
mkdir jslibs
# We keep a special hash for testing
echo "e96461c6c19608f528b4a3c33a032b697b999b62" >> whitelist_hashes.txt
mv whitelist_hashes.txt ../validator/testcases/hashes.txt

To add new libraries to the mix, edit extras/ and add the version number to the appropriate tuple.


In order to maintain Jetpack compatibility, the whitelist hashes need to be regenerated with each successive Jetpack version. To rebuild the hash library, simply run:

cd jetpack

That's it!