Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer.. #1044

pocmo · 2018-08-29T15:03:08Z

@mcomella @Baron-Severin I want to share the current state of the migration to use browser-engine. This version is functional and you can use it to browse the web. Like with the browser-session integration I used extension functions to work around functionality that is not available in the upstream components.

Missing pieces without a workaround in this version (list may be incomplete):

Tracking protection toggle not functional (TP is already in components 0.20, with 0.21 [Friday] we get the option to configure this globally for an engine or session and that will make this toggle simpler to implement.
We can't intercept URLs to inject HTML currently (used for firefox:home, firefox:about and error pages). ~~This functionality is scheduled to land in 0.22 (next week). However I may move this into the current sprint so that we can have this in 0.21 on Friday.~~ We moved this into the 0.21 sprint and the functionality should be available in 0.21.
- Without loading "firefox:home" and adding it to the tab's history this currently breaks some assumptions when using the back button. This should fix itself once "firefox:home" gets loaded normally.
Fullscreen API (works with YouTube because the TV website doesn't use the web API and just shows the content at full size). Scheduled to land in 0.21 on Friday.

Added follow-up items to: #1063

As a side effect:

Replaces AmazonWebView with WebView (Issue #379)
Removes the product flavor dimensions that are not needed anymore.

mcomella

With a quick skim, this looks reasonable to me. We can spend more time reviewing when we land for real (next week?). Note: I'll be PTO next week so I'll be unable to look over this until the following Monday.

You win Fire TV!

mcomella · 2018-08-30T22:34:26Z

app/src/main/java/org/mozilla/focus/Components.kt

-    override fun name(): String {
-        throw NotImplementedError()
-    }
+    val sessionUseCases = SessionUseCases(sessionManager)


nit: by lazy? Otherwise engine and sessionManager will make themselves right now.

Oh, yeah, good catch!

mcomella · 2018-08-30T22:35:07Z

Note: I'll be PTO next week so I'll be unable to look over this until the following Monday.

Unless we can land something tomorrow. :)

mcomella · 2018-08-30T23:20:23Z

Alternatively...

@pocmo How much is changing for the final PR?

Provided the changes aren't too big, I can look over the most recent WIP tomorrow, Severin can look at it next week and land it, and I can look at it to make sure we didn't miss anything when I come back. That way we can get it in sooner and test earlier. What do you think?

pocmo · 2018-08-31T08:15:45Z

@pocmo How much is changing for the final PR?

For this iteration not much.

Components 0.21 should get released today and then with that I should be able to look at the three missing pieces mentioned above. The patches for tracking protection and URL interception landed in a-c. Fullscreen API is at risk for 0.21 because the PR is not up yet. We may need to ship that with 0.22 or can ship a 0.21.1 if this gets in early next week.

Provided the changes aren't too big, I can look over the most recent WIP tomorrow, Severin can look at it next week and land it, and I can look at it to make sure we didn't miss anything when I come back.

Depending on what you and @Baron-Severin think I'd be open to land this as-is (with review comments addressed) and open separate PRs for the next pieces. The idea is to unblock others by having to wait on a big code drop and to be able to have others test and iterate on the new code. The drawback is that we temporarily have a not 100% feature complete app though (-> at least the three points mentioned above). Apart from that the app seems to be "fully" functional in my tests.

pocmo · 2018-08-31T12:11:47Z

Updated the PR. Made all checks pass (ktlint, detekt, checkstyle, lint), SessionUseCases initializes lazily now and cleaned up some of the code.

mcomella

A few things missing here or there and a few nits but otherwise lgtm. I find the SessionUseCases abstraction strange: see below.

Some other things that have changed – please file a bug or fix:

Back/forward state takes a while to update

Not sure it was related to your changes but I got into a state where a white page was loaded and the progress bar was shown without a URL (i.e. empty str or null) after going backwards and forwards a few times.

I'd be open to land this as-is (with review comments addressed) and open separate PRs for the next pieces.

Me too: please send an email to the team letting them know 1) this is landing, 2) the features that need follow-ups, and 3) that they'll need to change their build variant. Also, don't forget to file issues to track this work! :)

@Baron-Severin FYI that this may land.

app/build.gradle

app/src/main/java/org/mozilla/focus/ext/EngineView.kt

mcomella · 2018-09-01T02:24:10Z

app/src/main/java/org/mozilla/focus/ext/EngineView.kt

+ * component is powerful enough..
+ */
+@Suppress("DEPRECATION")
+fun EngineView.setupForApp(context: Context) {


Add @SuppressLint("SetJavaScriptEnabled") // We explicitly want to enable JavaScript

...src/amazonWebview/java/org/mozilla/focus/webview/FirefoxAmazonWebView.kt

app/src/main/java/org/mozilla/focus/ext/EngineView.kt

app/src/main/java/org/mozilla/focus/browser/InfoActivity.java

app/src/main/java/org/mozilla/focus/browser/BrowserFragment.kt

mcomella · 2018-09-01T02:24:10Z

app/src/main/java/org/mozilla/focus/browser/BrowserFragment.kt

@@ -194,6 +192,35 @@ class BrowserFragment : IWebViewLifecycleFragment(), Session.Observer {
        return layout
    }

+    var sessionFeature: SessionFeature? = null


nit: move to top of class

mcomella · 2018-09-01T02:24:10Z

app/src/main/java/org/mozilla/focus/browser/BrowserFragment.kt

+
+        // The SessionFeature implementation will take care of making sure that we always render the currently selected
+        // session in our engine view.
+        sessionFeature = SessionFeature(


Should be nulled in onDestroyView because it has a reference to a view

mcomella · 2018-09-01T02:24:10Z

app/src/main/java/org/mozilla/focus/browser/BrowserFragment.kt

@@ -194,6 +192,35 @@ class BrowserFragment : IWebViewLifecycleFragment(), Session.Observer {
        return layout
    }

+    var sessionFeature: SessionFeature? = null
+
+    override fun onViewCreated(view: View, savedInstanceState: Bundle?) {


nit: I don't love using onCreateView and onViewCreated in the same file: I think the relationship between them isn't intuitive (basically that onViewCreated is only for using the new WebView reference) and devs will put content in them interchangeably (even experienced devs might overlook this).

I'd prefer to make onViewCreated final in the super class and define a new abstract method like, onCreateViewWithWebView or something.

I introduced a method onWebViewCreated(webView: EngineView) that will be called from EngineViewLifecycleFragment.onViewCreated() and moved the code to onWebViewCreated in BrowserFragment.

mcomella · 2018-09-01T02:34:10Z

fwiw, I am a little concerned that we're rushing this and should maybe wait until the next release. We merged the home screen into the navigation overlay, which isn't stable yet, and we're landing this complex refactor (whose components haven't been used in a production app before) into that. For example, there are regressions in back/forward state that I couldn't reproduce consistently and I'm not sure who is to blame; debugging may get complicated when there's so much changing at once!

That being said, we can't expect this to land without bugs, the components are well tested in isolation, and it's pretty important to get the components tested in production sooner rather than later so we can add them to our other projects.

Perhaps we should land but not hesitate to back it out if it causes problems? What do you think @pocmo ? If so, let's ask QA to start testing this as soon as it lands. Note that Chenxia, Severin, and I won't be working on this next week so we'll have fewer people around if something goes wrong and after that we only have two weeks to iron out any issues.

@liuche @Baron-Severin Please share if you have any thoughts!

pocmo · 2018-09-03T13:30:06Z

From the migration I feel like the overlay and the things touching the engine are pretty well separated. We are definitely introducing risk with the migration but it doesn't feel like it gets worse by having both things (engine migration + overlay migration) happen at the same time.

whose components haven't been used in a production app before

Yeah, that's a chicken egg problem. FFFTV is the first candidate because the feature set is the smallest and the first milestone we have reached. :)

there are regressions in back/forward state that I couldn't reproduce consistently

I noticed that there is a regression because we can't intercept "firefox:home" and inject custom content. With that the page is not in the history stack and breaks some of the logic. The functionality for that landed in 0.21 - I added an item about testing back/forward after that to #1063.

Perhaps we should land but not hesitate to back it out if it causes problems? What do you think @pocmo ?

Yeah, that sounds good to me. It will get more complicated to back out later on after we added more commits on top of it though.

I'll talk to QA once this lands. :)

pocmo · 2018-09-03T13:38:19Z

@Baron-Severin From my side this is ready to land. I addressed most of @mcomella's review comments, rebased the branch and added follow-up items to #1063. I started working on the follow-up items now (using 0.21 of the components) but I'll create separate PRs for those changes once this PR lands.

severinrudie

Had a few questions, nothing major. LGTM

severinrudie · 2018-09-04T20:10:04Z

app/build.gradle

@@ -13,7 +13,7 @@ android {
    compileSdkVersion 27

    defaultConfig {
-        applicationId "org.mozilla.tv"
+        applicationId "org.mozilla.tv.firefox"


Nit: Why are we changing the app ID? Is there an issue that explains this?

Previously we had product flavors for the engine and those flavors added a suffix:

roductFlavors { amazonWebview { applicationIdSuffix ".firefox" // [..] } gecko { applicationIdSuffix ".gecko" // [..] } }

We have been shipping the amazonWebview flavor only. Now that the flavor is gone I moved the suffix to the applicationId setup directly.

severinrudie · 2018-09-04T20:10:04Z

app/src/main/java/org/mozilla/focus/MainActivity.kt

-    private fun initAmazonFactory() {
-        if (!isAmazonFactoryInit) {
-            factory = AmazonWebKitFactories.getDefaultFactory()
-            if (factory!!.isRenderProcess(this)) {


I love how much code this PR deletes.

severinrudie · 2018-09-04T20:10:04Z

app/src/main/java/org/mozilla/focus/browser/InfoActivity.kt

+        val actionBar = supportActionBar
+        if (actionBar != null) {
+            actionBar.setDisplayHomeAsUpEnabled(true)
+            actionBar.setDisplayShowHomeEnabled(true)


Nit: we could avoid capturing the actionBar variable here by using apply or similar.

That's a good idea and avoids some repetition. I'll update the PR.

… to replace WebView abstraction layer. As a side effect: * Replaces AmazonWebView with WebView (Issue #379) * Removes the product flavor dimensions that are not needed anymore.

pocmo requested a review from mozilla-mobile/firetv-codeowners as a code owner Aug 29, 2018

pocmo changed the title ~~Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer.~~ WIP: Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer. Aug 29, 2018

pocmo force-pushed the pocmo:engine-integration branch from 5c63721 to 2284d62 Aug 29, 2018

pocmo changed the title ~~WIP: Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer.~~ WIP: Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer.. Aug 29, 2018

severinrudie added the WIP label Aug 29, 2018

pocmo mentioned this pull request Aug 30, 2018

Replace webview abstraction layer with browser-engine-system component #1039

Closed

mcomella reviewed Aug 30, 2018

View changes

pocmo force-pushed the pocmo:engine-integration branch 2 times, most recently from c118132 to 8b0e857 Aug 31, 2018

mcomella approved these changes Sep 1, 2018

View changes

pocmo changed the title ~~WIP: Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer..~~ Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer.. Sep 3, 2018

pocmo removed the WIP label Sep 3, 2018

pocmo requested a review from severinrudie Sep 3, 2018

pocmo force-pushed the pocmo:engine-integration branch from 8b0e857 to 790dc50 Sep 3, 2018

pocmo force-pushed the pocmo:engine-integration branch from 790dc50 to ca9e7eb Sep 3, 2018

severinrudie approved these changes Sep 4, 2018

View changes

pocmo force-pushed the pocmo:engine-integration branch from ca9e7eb to 2c69552 Sep 5, 2018

Issue #1039: Use browser-engine-system and feature-session components…

672c927

… to replace WebView abstraction layer. As a side effect: * Replaces AmazonWebView with WebView (Issue #379) * Removes the product flavor dimensions that are not needed anymore.

pocmo force-pushed the pocmo:engine-integration branch from 2c69552 to 672c927 Sep 5, 2018

pocmo deleted the pocmo:engine-integration branch Sep 5, 2018

mozilla-mobile / firefox-tv Archived

Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer.. #1044

Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer.. #1044

pocmo commented Aug 29, 2018 •

edited

mcomella left a comment

This comment has been minimized.

This comment has been minimized.

mcomella commented Aug 30, 2018

mcomella commented Aug 30, 2018

pocmo commented Aug 31, 2018

pocmo commented Aug 31, 2018

mcomella left a comment

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

mcomella commented Sep 1, 2018 •

edited

pocmo commented Sep 3, 2018

pocmo commented Sep 3, 2018

severinrudie left a comment

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

mozilla-mobile / firefox-tv Archived

Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer.. #1044

Issue #1039: Use browser-engine-system and feature-session components to replace WebView abstraction layer.. #1044

Conversation

pocmo commented Aug 29, 2018 • edited

mcomella left a comment

This comment has been minimized.

mcomella Aug 30, 2018 Contributor

This comment has been minimized.

pocmo Aug 31, 2018 Author Contributor

mcomella commented Aug 30, 2018

mcomella commented Aug 30, 2018

pocmo commented Aug 31, 2018

pocmo commented Aug 31, 2018

mcomella left a comment

This comment has been minimized.

mcomella Sep 1, 2018 Contributor

This comment has been minimized.

mcomella Sep 1, 2018 Contributor

This comment has been minimized.

mcomella Sep 1, 2018 Contributor

This comment has been minimized.

mcomella Sep 1, 2018 Contributor

This comment has been minimized.

pocmo Sep 3, 2018 Author Contributor

mcomella commented Sep 1, 2018 • edited

pocmo commented Sep 3, 2018

pocmo commented Sep 3, 2018

severinrudie left a comment

This comment has been minimized.

severinrudie Sep 4, 2018 Contributor

This comment has been minimized.

pocmo Sep 5, 2018 Author Contributor

This comment has been minimized.

severinrudie Sep 4, 2018 Contributor

This comment has been minimized.

severinrudie Sep 4, 2018 Contributor

This comment has been minimized.

pocmo Sep 5, 2018 Author Contributor

pocmo commented Aug 29, 2018 •

edited

mcomella Aug 30, 2018
Contributor

pocmo Aug 31, 2018
Author Contributor

mcomella Sep 1, 2018
Contributor

mcomella Sep 1, 2018
Contributor

mcomella Sep 1, 2018
Contributor

mcomella Sep 1, 2018
Contributor

pocmo Sep 3, 2018
Author Contributor

mcomella commented Sep 1, 2018 •

edited

severinrudie Sep 4, 2018
Contributor

pocmo Sep 5, 2018
Author Contributor

severinrudie Sep 4, 2018
Contributor

severinrudie Sep 4, 2018
Contributor

pocmo Sep 5, 2018
Author Contributor