isReady may remain false in MessageListWeb

[jainkuniya]

We are using postMessage to communicate between react app and message list (i.e WebView).
postMessage is acting as a bridge between whole app and webView.
Whenever there is a update in app e.g.: typing events or new message etc we send html content to WebView via this postMessage bridge. And similarly we get scroll, touch/click events from WebView (Message List) to react native app via the same bridge postMessage.

https://github.com/zulip/zulip-mobile/blob/master/docs/architecture/react.md#the-exception-messagelistweb

One thing we need to make sure is, bridge should be successfully established before we start communication between two components. Else we will be sending messages from one side while the other side is not yet ready to receive/process messages. Thus this will cause in consistency and the updates will not be observed in the real time.

On current master, we are sending isReady message whenever webView is ready to receive/send/process messages. But the the problem is, what if when webView send isReady message at that time RN part is not ready or this event is not received by RN? We are just ignoring this case currently.
Then according to this RN will always thinks that webView is not ready and will never send any new message to webView, and keep on waiting till it hears isReady from webView.

On current master, one way hand shake is implemented (explained in above paragraph).

Which starts causing problems on Android 9. Issue like:

• Message list only works the second time #3078
• Typing indicator are never shown in the message list.
• More messages are never loaded on scrolling to top/bottom.
• White screen (or message loading) on narrowing to a narrow in which we don't have any message.
  Because on such narrow, we request new messages from server and send to webView via postMessage. So after fetching we want to send new html content to WebView, but RN thinks
  WebView is not ready, so it doesn't send. (https://chat.zulip.org/#narrow/stream/48-mobile/subject/messages.20don't.20load/near/655786)

oh, it's a mix. Sometimes loading bars, sometimes not

(https://chat.zulip.org/#narrow/stream/48-mobile/subject/messages.20don't.20load/near/655849)

• etc etc (actually whole communication seems to be broken)

Solution: Improve/enhance this hand shaking mechanism. Implement a two way hand shake.
Keep sending a message from one side till a confirmation from other side is heard/received.
I will prefer to keep this one side as RN one, because we have more control on RN one and it is easier to debug in RN.

(Sending PR in few minutes)

@zulipbot claim

[borisyankov]

Great find, @jainkuniya Also, thanks for the quick reaction.
I have some comments regarding the code itself but will do it as a review in the PR.

Let's work a bit on researching the exact cause and time this issue appeared.
Few people reported this started for them just a few days ago. That is interesting.
It led us (at least me) to believe that some code change we did recently did introduce that.
On the other hand, being Android version-specific makes sense as I have not seen this myself, and I am yet on Android 8, as Samsung is slow to update.

It is a bit strange that our component will render and the HTML, CSS and JS be ready and executing before it is even capable of receiving messages.
We need to investigate this further and be very clear why this happen and what else might be possible to happen (as this is absolutely key functionality for our app).

[gnprice]

Hmm, this is an interesting possibility! Thanks @jainkuniya for raising this.

I agree with Boris that it seems a bit surprising that this would happen, but I think well within the range of surprising things that are possible. Also definitely agree that I want to understand clearly what other things can happen.

It remains puzzling to me that this would only just start happening to some people, going from rare to 100% of the time, in the past week. We haven't sent a new release to beta or prod in several weeks, and none of the people reporting symptoms (of #3078 and your fourth bullet) are alpha testers. Also we have those reports on Android versions 7.1.2, 8.0.0, and 9.

One plausible hypothesis, really the only one I have: an update to the Chrome app, which powers the WebView. Apparently 70.0.3538 was released 2018-10-17, i.e. last week; and I'm still on the previous version, and I haven't yet seen these symptoms. They certainly roll these things out gradually, so perhaps the people seeing this are people whose devices have gotten that update.

(That'd still leave open the questions Boris mentioned, which we'll still want to understand -- but it'd give us a lead on answering them, among other things.)

[gnprice]

See updates on #3078. Most happily, we have a way to reproduce! It looks like the Chrome 70 update is indeed the trigger.

This thread is a plausible hypothesis for what's causing #3078. The next step is to nail down whether this is really what's happening, by watching closely with the debugger while reproducing #3078.