920293 - Provide a download-start event to Gaia

Status: RESOLVED WONTFIX

(Firefox OS Graveyard :: General, defect)

Description

[[:fabrice] Fabrice Desré]

This mozChromeEvent will be triggered in the following situations:
- Downloading content from an unsupported mime type.
- When the "download" attribute exists on a <a> tag.
- When the content-disposition header is set.

The event sends the url of the resource to download, along with the manifest URL of the application that started the download, and a boolean that let gaia know if this comes from a browser frame.

Gaia must then use the manifest URL and the browser flag as parameters to a XHR constructor to make sure that we use the appropriate data jar for this request.

Comment 1

[[:fabrice] Fabrice Desré]

Attachment: wip

Ben, as we discussed on irc, please check how this will fail in the worker xhr and advise a solution (I just need to get the appId from the manifest url using the xpcom AppsService).

Comment 2

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Comment on attachment 809522 [details] [diff] [review]
wip

Review of attachment 809522 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/workers/XMLHttpRequest.cpp
@@ +1489,5 @@
> +    // Get the appId from the manifestURL, if any.
> +    uint32_t appId = nsIScriptSecurityManager::NO_APP_ID;
> +    if (!aParams.mMozManifestURL.IsEmpty()) {
> +      nsCOMPtr<nsIAppsService>
> +        appsService = do_GetService(APPS_SERVICE_CONTRACTID);

This won't work properly since this service is implemented in JS. You'll need to proxy to the main thread for this information.

I think basically you can just store the manifestURL that is passed here and then pass it along to the proxy. Then use the appservice once you've bounced to the main thread.

Comment 3

[[:fabrice] Fabrice Desré]

Attachment: wip 2

Comment 4

[[:fabrice] Fabrice Desré]

Attachment: tests

First try at writing tests. They don't pass and I'm not sure where the issue is yet. We set a load context on the xhr channel and I can see the GetAppId() being called and returning the expected ID, but we still don't get the previously set cookie back.

Comment 5

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Comment on attachment 816894 [details] [diff] [review]
wip 2

Review of attachment 816894 [details] [diff] [review]:
-----------------------------------------------------------------

This looks right from the worker side, I think. Maybe I found your bug:

::: dom/workers/XMLHttpRequest.cpp
@@ +137,3 @@
>    : mWorkerPrivate(nullptr), mXMLHttpRequestPrivate(aXHRPrivate),
>      mMozAnon(aMozAnon), mMozSystem(aMozSystem),
> +    mMozManifestURL(EmptyString()),

You probably want to use the manifest url that you passed in, right?

@@ +1444,5 @@
>    mWorkerPrivate(aWorkerPrivate),
>    mResponseType(XMLHttpRequestResponseType::Text), mTimeout(0),
>    mJSObjectRooted(false), mBackgroundRequest(false),
> +  mWithCredentials(false), mCanceled(false), mMozAnon(false), mMozSystem(false),
> +  mMozManifestURL(EmptyString()), mMozIsInBrowserElement(false)

No need to initialize mMozManifestURL, it will be empty automatically.

Comment 6

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Comment on attachment 816896 [details] [diff] [review]
tests

This test looks ok to me. You'll want to split the actual test code into a js file so that you can run it in a worker, I guess.

Comment 7

[BMO Automation]

Firefox OS is not being worked on