753925 - Asana & LucidChart App has 30 seconds of white screen on startup before showing Loading Screen

Status: RESOLVED WORKSFORME

(Firefox Graveyard :: Webapp Runtime, defect, P2)

Description

[api]

Attachment: Screenshot of Blank window

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:15.0) Gecko/15.0 Firefox/15.0a1
Build ID: 20120507030511

Steps to reproduce:

- Installed Asana from Mozilla Marketplace on Firefox Nightly
- Started Asana from the Applications Folder


Actual results:

- Asana Window opens, but is entirely white for 30 seconds before the loading screen (blue background) appears



Expected results:

- Loading screen should appear immediately

Comment 1

[Jason Smith [:jsmith]]

Confirmed. This issue only happens on the desktop runtime, but does not happen on desktop firefox nightly. Nominating for k9o - this is an unacceptable amount of startup time for a web app in the runtime.

Comment 2

[Jason Smith [:jsmith]]

Per a conversation with Lawrence, marking for re-triage. We've seen this happen with more than one app (this one just so happens to reproduce consistently) - 30 seconds is unacceptable. We'll lose users over this issue.

Comment 3

[Jason Smith [:jsmith]]

Also getting a reproduction of this issue with LucidChart on Windows.

Comment 4

[Jason Smith [:jsmith]]

Needs more info to understand how widespread the problem is.

Comment 5

[Vladan Djeric (:vladan)]

(2:00:49 PM) vladan: hi Jason, i've been looking at bug 753925 (30s of white screen on app startup). i'm pretty sure it's not a performance problem. on windows, the app menus are responsive while the content area is blank.
(2:01:22 PM) vladan: I couldn't figure out exactly where the delay is coming from but i did notice in the console output that there were warnings about JavaScript errors from the webrt container  "JavaScript error: , line 0: nothing active on context" as well as a warning message about SSL handshakes timing out. I think it would make sense to turn this over to a webRT developer
(2:01:53 PM) jsmith: hmm okay. Could you put this information in the bug (bug 753925)?
(2:02:04 PM) jsmith: Also, would this issue apply to multiple apps or a select set of apps?
(2:02:13 PM) jsmith: based on what you've seen so far
(2:03:16 PM) vladan: i've only looked at the Asana app so i can't say. i think the next step would be to create a debug webapp that points to a local webserver which is hosting the same https://app.asana.com/ content
(2:03:35 PM) jsmith: okay
(2:04:13 PM) vladan: if the issue can be reproduced with the debug webapp, then we can slowly peel away complexity until we narrow down the source. i don't have the knowledge of webapps to do it myself.. i did give it a shot but ran into a bunch of issues

Comment 6

[Jason Smith [:jsmith]]

Here's an idea to move forward on this - Could someone setup a debugger with the web runtime and launch the asana application and watch what happens? What parts of the code are getting hung? Why is it getting hung? How widespread could this problem be?

Comment 7

[Jason Smith [:jsmith]]

Nominating for k9o - start-up performance of 30 seconds is unacceptable. We already know this affects at least one top app (LucidChart). Generally, the reproducible examples of this bug are LucidChart and Asana, but I have randomly stumbled into this issue with other apps (e.g. tinkercad on Win 7).

I wonder if this issue is not tied to a specific app, but rather something going on behind the scenes that affects a set of apps? Something needs to debug this to get more information.

Comment 8

[[e-mail] [:onecyrenus]]

Performance Testing is probably something you should ask Henrik about. 

Are we blocked on not knowing how to test / provide feedback, or that someone else should be testing and reporting back this data. ? 

I think it would be a useful skill to figure out how to gather performance / profiling data going forward ?

Comment 9

[Jason Smith [:jsmith]]

(In reply to [e-mail] from comment #8)
> Performance Testing is probably something you should ask Henrik about.

Actually, Vlad (already cc-ed here) would be the right person to look at this initially, and he already has (he's already taken a shot at debugging this). He works with the Snappy team. The follow-up item suggested by Vlad was to have someone from WebRT engineering debug this with me tag teaming with them, as we could jointly narrow down the root cause of the problem. Good news is that it's a reproducible problem, so this might not be hard to get to the root cause.
 
> 
> Are we blocked on not knowing how to test / provide feedback, or that
> someone else should be testing and reporting back this data. ? 

The current blocking point is narrowing down the root cause of the problem (we know that it is a problem, it's reproducible, we just don't know the root cause yet).

> 
> I think it would be a useful skill to figure out how to gather performance /
> profiling data going forward ?

Already on the radar, although might want to get bugs on file for this (which I'll do now). I'll a get a bug on file to integrate with the telemetry dashboard for app install and app launch perf initially.

Telemetry dashboard btw is here - https://metrics.mozilla.com/data/content/pentaho-cdf-dd/Render;jsessionid=90E7B51B96402FE15596C349BED2A899?solution=metrics2&path=%2Ftelemetry&file=SignIn.wcdf.

Comment 10

[Jason Smith [:jsmith]]

Also Note - We're digging into learning the profiler builds just setup recently, so that will hopefully help in diagnosing this problem (so in short, we're working on this).

Comment 11

[Jason Smith [:jsmith]]

Weirdly enough, this bug no longer reproduces on at least 3 different machines (Vlad's machine, my machine, and Juan's machine). Tested on Windows and OS X. Although that's good news it's not happening anymore. Resolving as works for me.