I've been reading and hacking around with https://developer.mozilla.org/en/XUL_School/Intercepting_Page_Loads but can seem to do what I need.
I'm working on Chromeless, trying to prevent the main xulbrowser element from ever being navigated away from, e.g., links should not work, neither should window.location.href="http://www.example.com/".
I'm assuming I can do this via browser.webProgress.addProgressListener and then listen to onProgressChange but I can't figure out how to differentiate between a resource request and the browser changing locations (it seems that onLocationChange is too late as the document is already being unloaded).
browser.webProgress.addProgressListener({
onLocationChange: function(){},
onStatusChange: function(){},
onStateChange: function(){},
onSecurityChange: function(){},
onProgressChange: function(){
aRequest.QueryInterface(Components.interfaces.nsIHttpChannel)
if( /* need to check if the object triggering the event is the xulbrowser */ ){
aRequest.cancel(Components.results.NS_BINDING_ABORTED);
}
},
QueryInterface: xpcom.utils.generateQI([Ci.nsIWebProgressListener, Ci.nsISupportsWeakReference])
}, wo._browser.webProgress.NOTIFY_ALL);
Another option that sounds promising is the nsIContentPolicy.shouldLoad() method but I really have no clue how to "create an XPCOM component that extends nsIContentPolicy and register it to the "content-policy" category using the nsICategoryManager."
Any Ideas?
I got help on this from the mozilla's #xulrunner irc channel.
Resulting solution follows.
Note: this is a module for use in Mozilla Chromeless, the require("chrome") and require("xpcom") bits will NOT be available under normal circumstances.
const {Cc, Ci, Cu, Cm, Cr} = require("chrome");
const xpcom = require("xpcom");
/***********************************************************
class definition
***********************************************************/
var description = "Chromeless Policy XPCOM Component";
/* UID generated by http://www.famkruithof.net/uuid/uuidgen */
var classID = Components.ID("{2e946f14-72d5-42f3-95b7-4907c676cf2b}");
// I just made this up. Don't know if I'm supposed to do that.
var contractID = "#mozilla.org/chromeless-policy;1";
//class constructor
function ChromelessPolicy() {
//this.wrappedJSObject = this;
}
// class definition
var ChromelessPolicy = {
// properties required for XPCOM registration:
classDescription: description,
classID: classID,
contractID: contractID,
xpcom_categories: ["content-policy"],
// QueryInterface implementation
QueryInterface: xpcom.utils.generateQI([Ci.nsIContentPolicy,
Ci.nsIFactory, Ci.nsISupportsWeakReference]),
// ...component implementation...
shouldLoad : function(aContentType, aContentLocation, aRequestOrigin, aContext, aMimeTypeGuess, aExtra) {
let result = Ci.nsIContentPolicy.ACCEPT;
// only filter DOCUMENTs (not SUB_DOCUMENTs, like iframes)
if( aContentType === Ci.nsIContentPolicy["TYPE_DOCUMENT"]
// block http(s) protocols...
&& /^http(s):/.test(aContentLocation.spec) ){
// make sure we deny the request now
result = Ci.nsIContentPolicy.REJECT_REQUEST;
}
// continue loading...
return result;
},
createInstance: function(outer, iid) {
if (outer)
throw Cr.NS_ERROR_NO_AGGREGATION;
return this.QueryInterface(iid);
}
};
let registrar = Cm.QueryInterface(Ci.nsIComponentRegistrar);
try
{
Cm.nsIComponentRegistrar.registerFactory(classID, description, contractID, ChromelessPolicy);
}
catch (e) {
// Don't stop on errors - the factory might already be registered
Cu.reportError(e);
}
const categoryManager = Cc["#mozilla.org/categorymanager;1"].getService(Ci.nsICategoryManager);
for each (let category in ChromelessPolicy.xpcom_categories) {
categoryManager.addCategoryEntry(category, ChromelessPolicy.classDescription, ChromelessPolicy.contractID, false, true);
}
Pull Request on github for those that are interested: https://github.com/mozilla/chromeless/pull/114
Related
I'm creating a restartless (bootstrapped) addon for Thunderbird to replicate the functionality of an overlay-based addon I made years ago. The desire is to manipulate some of the text in the new email's body. In the old overlay version, it was pretty straightforward: I could just use the gMsgCompose object. How do I get the gMsgCompose object for a compose window in a restartless addon?
Here's my code right now. For a while I thought that edElem was nearly what I wanted, which is why I'm test-dumping edElem.editortype (which should be either "textmail" or "htmlmail" when I really do have the right object).
'use strict';
const Cc = Components.classes;
const Ci = Components.interfaces;
const Cu = Components.utils;
Cu.import("resource://gre/modules/Services.jsm");
function startup(data, reason)
{
Services.wm.addListener(MyAddon.windowListener);
}
function shutdown(data, reason)
{
Services.wm.removeListener(MyAddon.windowListener);
}
function install(aData, aReason) { }
function uninstall(aData, aReason) { }
if(!MyAddon) var MyAddon = {};
MyAddon.windowListener = {
onOpenWindow: function(aXULWindow) {
let aDOMWindow = aXULWindow
.QueryInterface(Ci.nsIInterfaceRequestor)
.getInterface(Ci.nsIDOMWindow);
if (!aDOMWindow) return;
aDOMWindow.addEventListener("load", function _cl_load(event){
aDOMWindow.removeEventListener("load", _cl_load, false);
MyAddon.handleOpenWindow(aDOMWindow, event.target);
}, false);
}
}; // MyAddon.windowListener
MyAddon.handleOpenWindow = function(aDOMWindow, DOMdocument) {
if(DOMdocument.documentURI != "chrome://messenger/content/messengercompose/messengercompose.xul")
return;
let edWindow = DOMdocument.documentElement;
// Just to be sure
if(edWindow.id != 'msgcomposeWindow')
return;
let edElem = DOMdocument.getElementById("content-frame");
/* I thought edElem was the element I want, but the following
* returns an empty result, which I take to mean the element isn't
* fully loaded yet.
*/
dump("XXX handleOpenWindow: edElem.editortype="+edElem.editortype+"\n");
/* Adding a listener to this element doesn't get the behaviour
* I want either. I've tried hooking several events
* in the following fashion, and the event handler code doesn't
* execute for any of them, so maybe edElem isn't what I want.
*/
[ "load",
"unload",
"compose-window-init",
"compose-window-close",
"compose-fields-ready",
"compose-send-message",
].forEach(function(eventName) {
edElem.addEventListener(eventName, function _eE_FIXME(event){
dump("XXX "+eventName+": edElem.editortype="+edElem.editortype+"\n");
}, false);
});
/* So, how do I get control of the editor,
* so I can read and play with its content?
*/
}; // MyAddon.handleOpenWindow
The only question I found on stackoverflow that's close to what I want is Get sender and recipients in Thunderbird extension upon sending message. I've tried what they do there, but all it seems to do is give me an alternate way of getting the compose window as a whole, which I can already get.
I've made an ugly stub with a timer in the onOpenWindow:
domWindow.setTimeout( function() { uploadHook( domWindow ); }, 500 );
So then (in uploadHook) I can get editor with:
var editorNode = domWindow.document.getElementById("content-frame");
// GetCurrentEditor() equivalent:
nsIEditor = editorNode.getEditor(editorNode.contentWindow)
But looking for a better solution.
I've followed every step of this walkthrough, but when I try to create a new row, I get a 403:
code: 119
message: "This user is not allowed to perform the create
operation on Messages. You can change this setting in the Data Browser."
My code:
Messages = Parse.Object.extend("Messages")
var message = new Messages();
message.set("sender", Parse.User.current());
message.set("receiver", *anotherUser*);
message.set("subject", "foo")
message.set("body", "bar")
message.save()
.then(
function(message){
console.log("success!")
},function(error){
console.log("error: ", error);
});
My CLPs are set as follows:
It looks like someone else posted the same issue in a google group. What are we missing?
I've submitted this as a bug to Parse (Facebook), and they replied:
We have managed to reproduce this issue and it appears to be a valid bug. We are assigning this to the appropriate team.
I will update this answer once the issue has been resolved. If this issue is impacting you, please subscribe to the bug, as this will help prioritize the fix.
UPDATE
Facebook replied:
Turns out that this is actually by design. To create an object, the class should have public create permissions on it
Unfortunately, with this solution, I can create a message "from" any other user (another user set as the sender). This is unacceptable and unusable IMHO.
That has been a bug since the launch of Pointer Permissions, which effectively makes them useless. My impression is they built this with the idea of letting developers secure existing schemas in one go, but of course you need it to work for future creation.
One workaround would involve combining the older Class Level Permissions and per-row ACL's while being careful to not disable your Data Browser. Let's assume you have classes "Puppy" and "Cat" and both have a field called "owner".
In your Data Browser, for each class where it makes sense to have an owner field, you set its Class Level Permissions for Puppy and Cat each to:
Public - Read: Yes or No, depends on your use case, Write: Yes
Add a Pointer Permission for "owner" - Read: Yes, Write: Yes (can skip this for now, see below)
Then in your cloud/main.js, you can use the following as a starting point (which I often call "types" below, sorry).
When Parse fixes the creation issue, you remove the Public Write Class Level permission (above), leave the Pointer Permission one, and get rid of the workaround code below.
--
var validateAndUpdateOwnerWritePerms = function(request){
var object = request.object;
var error = null;
var owner = object.get('owner');
if (!Parse.User.current()) {
error = 'User session required to create or modify object.';
} else if (!owner) {
error = 'Owner expected, but not found.';
} else if (owner && owner.id != Parse.User.current().id && !object.existed()) {
error = 'User session must match the owner field in the new object.';
}
if (request.master) {
error = null;
}
if (error) {
return error;
}
if (object.existed()) {
return null;
}
var acl = new Parse.ACL();
acl.setReadAccess(owner, true);
acl.setWriteAccess(owner, true);
object.setACL(acl);
return null;
}
// Wrapper that makes beforeSave, beforeDelete, etc. respect master-key calls.
// If you use one of those hooks directly, your tests or admin
// console may not work.
var adminWriteHook = function(cloudHook, dataType, callback) {
cloudHook(dataType, function(request, response) {
if (request.master) {
Parse.Cloud.useMasterKey();
} else {
var noUserAllowed = false;
if (cloudHook == Parse.Cloud.beforeSave &&
(dataType == Parse.Installation || dataType == Parse.User)) {
noUserAllowed = true;
}
if (!noUserAllowed && !Parse.User.current()) {
response.error('Neither user session, nor master key was found.');
return null;
}
}
return callback(request, response);
});
};
// Set hooks for permission checks to run on delete and save.
var beforeOwnedTypeWriteHook = function(type) {
var callback = function (request, response) {
var error = validateAndUpdateOwnerWritePerms(request);
if (error) {
response.error(error);
return;
}
response.success();
};
return adminWriteHook(Parse.Cloud.beforeSave, type, callback);
return adminWriteHook(Parse.Cloud.beforeDelete, type, callback);
};
beforeOwnedTypeWriteHook('Puppy');
beforeOwnedTypeWriteHook('Cat');
Unfortunately it seems that Parse Pointer Permissions do not work as you expect it on Create. The quick fix would be to allow Create permission to Public. Then to ensure that the user who is creating a record is the same as the sender. So you need to perform a manual check in the beforeSave trigger for Messages class in cloud code and if that check fails, reject the record being created.
I'm having trouble adding an observer to watch for changes in firefox's search engines. I read the nsIBrowserSearchService page on the Mozilla Developer Site which suggests to use the init() method of the Services.search object.
void init([optional] in nsIBrowserSearchInitObserver observer);
I tried that and I managed to get it to execute the function once on start up but it never calls it again when I add or remove or reorder the search engines. So I'm doing something wrong.
I have experience with observers but only with using general preferences and I generally use add them and remove them using the code below. I'm not sure how to do it with the nsIBrowserSearchService . I would like to observe the nsIBrowserSearchService in the same or a similar way but I'm not sure what I would put for
branch = Services.prefs.getBranch("preferenceNameHere");
I'm not observing Services.prefs but Services.search I assume, and that has no getBranch method as far as I can tell.
This is how I normally add and remove the observer in Chrome.js
const {Ci, Cu} = require("chrome");
const {Services} = Cu.import("resource://gre/modules/Services.jsm", {});
branch = Services.prefs.getBranch("preferenceNameHere");
branch.addObserver("", observe, false);
function observe(subject, topic, data) {
// instanceof actually also "casts" subject
if (!(subject instanceof Ci.nsIPrefBranch)) {
return;
}
//do stuff here
};
exports.onUnload = function(reason) {
// Need to remove our observer again! This isn't automatic and will leak
// otherwise.
branch.removeObserver("", observe);
if(reason == "disable" || reason == "uninstall"){
//restore changes made by addon
}
};
Can anyone advise me on how to do this so I can observe for changes and ensure that I remove the observer properly with the search object. Thanks
What you are trying above is trying to use pref observer on non-prefs, this is not possible. You have to use the regular observer service like this.
This notifies you when the user changes the engine.
Cu.import('resource://gre/modules/Services.jsm');
var observers = {
'browser-search-engine-modified': {
aTopic: 'browser-search-engine-modified',
observe: function (aSubject, aTopic, aData) {
if (aData == 'engine-current') {
console.log('current engine was changed!');
//console.log('aSubject on change:', aSubject.name, 'same as Services.search.currentEngine.name:', Services.search.currentEngine.name); //aSubject is the engine
//console.log('aTopic on change:', aTopic); //aTopic is obviously `browser-search-engine-modified`
}
},
reg: function () {
Services.obs.addObserver(observers[this.aTopic], this.aTopic, false);
},
unreg: function () {
Services.obs.removeObserver(observers[this.aTopic], this.aTopic);
}
}
};
To start listening do this:
for (var o in observers) {
observers[o].reg();
}
To stop listening do this:
for (var o in observers) {
observers[o].unreg();
}
I'm not sure what happens when user adds a new engine but doenst select it. Or if he removes a engine. Please let me know what those messages are when user does that.
I'm running a script on Facebook that requires me to get the IDs of people in my "friends" window (this might not be the most efficient way to accomplish this specific task, but since I'd like to know how to do this in general it's a good example).
This means that if I have more than a small number of friends I have to scroll down for Facebook to add them to the page.
I've added logic that scrolls the page down to the footer, but I don't know how to force my function that grabs the IDs to run after the content loads.
For now, I've resorted to using setTimeout for a few seconds - obviously, this isn't guaranteed to at the appropriate time, so I'd like to know how to do this properly:
var k;
function doit(){
k = document.getElementsByClassName("_698");
var g= Array.prototype.slice.call(k);
confirm(g.length);
// the confirm is just to make sure it's working
// (if i don't use setTimeout it'll return a smaller number
// since not all the friends were included)
}
window.addEventListener("load", function(){
document.getElementById( "pageFooter" )
.scrollIntoView();setTimeout(doit,3000);
});
Crayon Violent details how to accomplish this in his answer to JavaScript detect an AJAX event. The trick is to hook the underlying XMLHttpRequest object in order to detect when a request is sent.
I've re-written the logic there a bit to make it more suitable for your needs:
//
// Hooks XMLHttpRequest to log all AJAX requests.
// Override ajaxHook.requestCompleted() to do something specific
// in response to a given request.
//
var ajaxHook = (function()
{
// we're using a self-executing function here to avoid polluting the global
// namespace. The hook object is returned to expose just the properties
// needed by client code.
var hook = {
// by default, just logs all requests to the console.
// Can be overridden to do something more interesting.
requestCompleted: function(xmlHttp, url, method) { console.log(url); }
};
// hook open() to store URL and method
var oldOpen = XMLHttpRequest.prototype.open;
XMLHttpRequest.prototype.open = function(method, url)
{
this.hook_method = method;
this.hook_url = url;
oldOpen.apply(this, arguments);
}
// hook send() to allow hooking onreadystatechange
var oldSend = XMLHttpRequest.prototype.send;
XMLHttpRequest.prototype.send = function()
{
var xmlhttp = this;
//hook onreadystatechange event to allow processing results
var oldReadyStateChange = xmlhttp.onreadystatechange;
xmlhttp.onreadystatechange = function()
{
oldReadyStateChange.apply(xmlhttp, arguments);
if ( this.readyState === 4 ) // completed
{
hook.requestCompleted(xmlhttp,
xmlhttp.hook_url, xmlhttp.hook_method);
}
};
oldSend.apply(this, arguments);
};
return hook;
})();
With this bit of code loaded in your userscript, you can then implement your logic as follows:
var k;
function doit()
{
k = document.getElementsByClassName("_698");
var g= Array.prototype.slice.call(k);
confirm(g.length);
}
window.addEventListener("load", function()
{
ajaxHook.requestCompleted = function(xmlhttp, url, method)
{
// is this the request we're interested in?
// (Facebook appears to load friends from a URL that contains this string)
if ( /AllFriendsAppCollectionPagelet/.test(url) )
{
// Facebook defers rendering the results here,
// so we just queue up scraping them until afterwards
setTimeout(doit, 0);
}
};
// trigger loading of more friends by scrolling the bottom into view
document.getElementById( "pageFooter" )
.scrollIntoView();
});
Is there any chance to pass some data to my server through install.rdf when my Firefox add-on check server for update?
Example:
...
<em:updateURL>http://www.site.com/update.php?var=myData</em:updateURL>
...
where "myData" is saved in options.xul or in another place like simple-storage.
Yes, but it is quite nasty. The AddonManager will replace a bunch of predefined and dynamic properties in the URL:
Register a new component implementing nsIPropertyBag2 (or use an existing implementation, such as ["#mozilla.org/hash-property-bag;1"]).
Register your component in the nsICategoryManager under the "extension-update-params" category.
Since you mentioned simple-storage: restartless add-ons must also unregister their stuff when being unloaded.
There is a unit test demonstrating how this stuff works. You of course need to adapt it a bit (if alone for require("chrome").
I found one "simple solution" but I dont know if that is also good practice ...
var origLink = "http://www.site.net/update.php?var=myData";
var newsLink = "http://www.site.net/update.php?var=" + simplePref.prefs.myData;
const {Cc,Ci,Cu} = require("chrome");
var observer = {
QueryInterface: function(iid) {
if (iid.equals(Ci.nsIObserver) || iid.equals(Ci.nsISupports)) return this;
},
observe: function(subject, topic, data){
if (topic == "http-on-modify-request"){
var channel = subject.QueryInterface(Ci.nsIChannel);
if (channel.originalURI.spec == origLink) {
channel.originalURI.spec = newsLink;
}
}
}
};
var ObsService = Cc["#mozilla.org/observer-service;1"].getService(Ci.nsIObserverService);
ObsService.addObserver(observer, "http-on-modify-request", false);