Currently in toolkit/identity/tests/chrome/test_sandbox.xul we test that objects are suppressed by checking for channel opens. However, in bug 745030, we become slightly smarter about opening channels for <object>/<embed> tags - we will open the channel if the server might return a unblocked MIME type (e.g. if image, subdocument, OR plugin is allowable), and then abort if it returns a blocked type. This means that checking for a channel open isn't sufficient for testing whether the object finished loading.

Comment on attachment 644117 [details] [diff] [review] Update sandbox test to reflect expected plugin behavior This changes the channel checking slightly to be okay with plugin types, but then adds a new block to specifically check that no plugin tags loaded any content (via nsIObjectLoadingContent.displayedType == TYPE_NULL)

Comment on attachment 644117 [details] [diff] [review] Update sandbox test to reflect expected plugin behavior Review of attachment 644117 [details] [diff] [review]: ----------------------------------------------------------------- Thanks for doing this. r=me with fixes. ::: toolkit/identity/tests/chrome/test_sandbox.xul @@ +128,5 @@ > > if (aNothingShouldLoad) { > is(xhr.responseText, "NOTHING", "Check that nothing was loaded on the server"); > } else if (!OSX_10_5) { > + } else { I'm guessing this didn't get rebased properly. Remove the |} else {| line so that these checks stay disabled on 10.5 due to intermittent failures (bug 770063). @@ +135,4 @@ > let loadedTypes = xhr.responseText == "NOTHING" ? [] : xhr.responseText.split(","); > > + for (let ind in loadedTypes) { > + isnot(allowedTypes.indexOf(loadedTypes[ind]), -1, "Check that " + loadedTypes[ind] + " was expected to load"); You can use for…of here since you are only using |ind| to get loadedTypes[ind]: |for (let loadedType of loadedTypes) {| https://developer.mozilla.org/en/JavaScript/Reference/Statements/for...of @@ +154,5 @@ > + let nullType = Components.interfaces.nsIObjectLoadingContent.TYPE_NULL; > + let pluginTags = [ "embed", "object", "applet" ]; > + for (let ind = 0; ind < pluginTags.length; ind++) { > + let tags = doc.getElementsByTagName(pluginTags[ind]); > + for (let tagInd = 0; tagInd < tags.length; tagInd++) { I think this would be clearer with one loop using querySelectorAll and for…of: for (let tag of doc.querySelectorAll("embed, object, applet")) { tag instanceof Components.interfaces.nsIObjectLoadingContent; is(tag.displayedType, nullType, "Check that plugin did not load content"); }

for..of? querySelectorAll? That will NEVER work in IE7! Oh. Right. Updated with notes, I'll land this tomorrow with bug 745030 if nothing else burns down

https://hg.mozilla.org/integration/mozilla-inbound/rev/3d5231ba7849 https://hg.mozilla.org/integration/mozilla-inbound/rev/aef63a0ce60c

This was backed out along with bug 745030 for random oranges, though 745030 is likely to blame: https://tbpl.mozilla.org/php/getParsedLog.php?id=13845528&tree=Mozilla-Inbound&full=1

Minor tweak, carrying forward r+

Minor tweak v2!

Annnd backed out again for aggravating Bug 778179 go nuts... https://hg.mozilla.org/integration/mozilla-inbound/rev/cf46b1ea1ccc

(In reply to John Schoenick [:johns] from comment #12) > Annnd backed out again for aggravating Bug 778179 go nuts... Err, making* bug 778179 go nuts

I'll look at that tomorrow. If you can catch the crash in a debugger, that would help.

Third time's the charm, now that we think 778179 has been resolved https://hg.mozilla.org/integration/mozilla-inbound/rev/3bfbd9a8ed90