735070 - navigator.mozApps.install() on error handler

Status: RESOLVED INVALID

(Web Apps :: General, defect)

Description

[[e-mail] [:onecyrenus]]

navigator.mozApps.install() on mozilla central will throw an exception. 

the html / js implementation does not do this, but returns to the onerror handler. 

I am unsure who would take this bug.

Comment 1

[[:fabrice] Fabrice Desré]

I don't understand this bug.

Can you give steps to reproduce, along with expected results and observed results?

Comment 2

[Ian Bicking (:ianbicking)]

Do you mean that:

calling navigator.mozApps.install() with no arguments has inconsistent behavior between implementations?

I'm a little unclear on the conflict.  The code/tests for the HTML implementation imply that it raises an exception (synchronously) if you call install() with no arguments, or with a non-string first argument (using typeof url == "string").  I don't think the IDL is very specific here, so I don't think we've clearly defined what it "should" do, but we should specify that and make sure both implementations are consistent.

Comment 3

[[e-mail] [:onecyrenus]]

Ack this might be a false positive. I will explain. 

navigator.mozApps.install() <--- calling on mozilla central would return a native exception. 

Now when i was looking at the html/js test code test_repo_api.html there was an error message associated with onerror that was getting displayed, but I didn't know if it was throwing an exception or an error message. 

spyPending(navigator.mozApps.install);

function spyPending(pending, name, noWait) {
>   var done = false;
>   pending.onsuccess = function () {
>     done = true;
>     var desc = 'onsuccess:';
>     if (name) {
>       desc = name + '.' + desc;
>     }
>     writeln(desc, JSON.stringify(pending.result));
>   };
>   pending.onerror = function () {
>     done = true;
>     var desc = 'onerror:';
>     if (name) {
>       desc = name + '.' + desc;
>     }
>     writeln(desc, JSON.stringify(pending.error));

.....

I confirmed that the html / js is throwing an exception. I still don't know how the onerror handler is getting called here, but the premise of this bug seems to be incorrect.

Comment 4

[Ian Bicking (:ianbicking)]

In the tests there's implicit code like this around every statement:

try {
  eval(statement);
} catch (e) {
  writeln('Error: ' + e);
}

So the "Error: ..." in the test is an actual exception.