1110309 - FxA user gets different room lists in different profiles

Status: RESOLVED FIXED

(Hello (Loop) :: Client, defect, P2)

Description

[Adam Roach [:abr]]

Attachment: room-lists.png

When users are logged in to Hello using FxA, they should see the same set of rooms regardless of which browser profile or device they are using. In Firefox 35, this does not appear to be the case.

The attached image shows two screen shots from two different browsers running at the same time. Note that they are the same account, but different room URLs and different room names. These should be in sync.

I'm marking this P1, since it's really a pretty serious and very user-visible flaw. Assuming this requires client work, I'd like to see this uplifted to 35 if possible.

Comment 1

[Jared Wein [:jaws] (please needinfo? me)]

The client is pretty dumb here, just asking the server for the list of rooms and displaying them. Maybe there is a cache somewhere that isn't getting invalidated?

Comment 2

[Jared Wein [:jaws] (please needinfo? me)]

It's also weird that the URLs are different. Those aren't changeable by the user, only the titles. Adam, do you know what profile created the rooms first? Are those different rooms in each profile, or are they the same rooms but with different URLs and names?

Comment 3

[Jared Wein [:jaws] (please needinfo? me)]

Also, I'm curious if bug 1102432 will fix this.

Comment 4

[Randell Jesup [:jesup] (needinfo me)]

I saw Nightly rooms sync to beta, but not the other way.  THe nightly had 17 rooms as guest when it logged into an fxa account that had 7 rooms.  Create room 18, shows up on the beta with 7.  Create room on beta (19), didn't show up on Nightly.  Create room on Nightly, also 19(?), and now beta shows two 19's with different URLs

Comment 5

[Adam Roach [:abr]]

(In reply to Jared Wein [:jaws] (please needinfo? me) from comment #2)
> It's also weird that the URLs are different. Those aren't changeable by the
> user, only the titles. Adam, do you know what profile created the rooms
> first? Are those different rooms in each profile, or are they the same rooms
> but with different URLs and names?

They're different rooms. I probably should have taken a screen shot with different room ordinalities, but the point is that these are behaving as if the same account in two different browsers aren't related to each other.

Comment 6

[Jared Wein [:jaws] (please needinfo? me)]

Follow-up from IRC discussion: Adam checked and both profiles were using the same Loop server. He then tried with a third profile and didn't get any of the rooms to come through.

The following is from the IRC logs:

2:53 PM <•abr> Okay, applied the patch from 1102432 and tried again with the Llamas/Lemurs profile. Still see those two rooms, do *not* see the room that I created in the fresh profile
2:53 PM <jaws> ummm.... that sounds quite broken
2:54 PM <•abr> jaws: Yeah. I agree.
2:55 PM <•abr> From that browser, created a new room ("Ferrets"). It does not show up in the other profile (the one with "Monkeychat 3000")
2:55 PM <•abr> https://www.dropbox.com/s/q0445mn4bsw5tun/Screenshot%202014-12-11%2013.55.33.png?dl=0
2:55 PM <•abr> https://www.dropbox.com/s/tiks8ts6ptnhq3b/Screenshot%202014-12-11%2013.55.46.png?dl=0
2:57 PM <•abr> And the fresh profile now has the same rooms as the one that created "Monkeychat 3000": https://www.dropbox.com/s/fo4j9mznmhmumdc/Screenshot%202014-12-11%2013.57.17.png?dl=0
2:58 PM <•abr> So it would appear that none of these are unaccounted rooms.
2:58 PM <•abr> They're just not syncing properly.
2:58 PM <•abr> Like I have this oddball login that is storing rooms somewhere else

Comment 7

[Adam Roach [:abr]]

For clarity, after I shut down the fresh profile browser and restarted it, two other rooms appeared. So it's like two of my profiles (the one I actually use for real work and the fresh one) are syncing, if in a somewhat delayed fashion. The third one -- which I typically use for testing -- shows me as logged in, but it's not seeing rooms created in the other two profiles, nor do those profiles see rooms created by it.

Comment 8

[Mike de Boer [:mikedeboer]]

How many Redis instances do we have in use atm and what is their setup? If possible, I'd like to take a look at their redis.conf files.

Comment 9

[Alexis Metaireau (:alexis)]

Mike, what information are you looking for specifically? We're currently using one redis instance on production, and I need to know which information you're looking for specifically.

Comment 10

[Maire Reavy [:mreavy]]

(In reply to Alexis Metaireau (:alexis) from comment #9)
> Mike, what information are you looking for specifically? We're currently
> using one redis instance on production, and I need to know which information
> you're looking for specifically.

Doing a needinfo back to Mike so he sees Alexis' question.

Comment 11

[Mike de Boer [:mikedeboer]]

(In reply to Alexis Metaireau (:alexis) from comment #9)
> Mike, what information are you looking for specifically? We're currently
> using one redis instance on production, and I need to know which information
> you're looking for specifically.

I'm looking at whether there's a possibility that something causes latency, which may shed some light on this issue. If there's only one Redis (master) instance, then we should be good.

We'll (I) have to investigate this further...

Comment 12

[Mark Banner (:standard8)]

Most of this issue really looks like bug 1102432 which will give the wrong lists after log-in until the first restart...

Adam's comment 7 implies there's possibly something else wrong - Adam, can you restart the third profile with --jsconsole on the command line, open up the panel, and see if you get any interesting errors/warnings?

Comment 13

[Maire Reavy [:mreavy]]

(In reply to Mark Banner (:standard8) from comment #12)
> Most of this issue really looks like bug 1102432 which will give the wrong
> lists after log-in until the first restart...
> 
> Adam's comment 7 implies there's possibly something else wrong - Adam, can
> you restart the third profile with --jsconsole on the command line, open up
> the panel, and see if you get any interesting errors/warnings?

Abr :  ^^^ (doing a needinfo so this pops to the top of your bug mail)

Comment 14

[Adam Roach [:abr]]

(In reply to Mark Banner (:standard8) from comment #12)
> Most of this issue really looks like bug 1102432 which will give the wrong
> lists after log-in until the first restart...
> 
> Adam's comment 7 implies there's possibly something else wrong - Adam, can
> you restart the third profile with --jsconsole on the command line, open up
> the panel, and see if you get any interesting errors/warnings?

By "third profile," I assume you mean the one that seem to be disconnected from everything else ("Llamas/Ferrets/Lemurs"). Nothing appears in the log that seems relevant. Here's the whole thing:

> OpenGL compositor Initialized Succesfully.
> Version: 2.1 NVIDIA-8.16.80 310.40.00.20f04
> Vendor: NVIDIA Corporation
> Renderer: NVIDIA GeForce GT 650M OpenGL Engine
> FBO Texture Target: TEXTURE_2D
> Could not read chrome manifest 'file:///Users/Adam/devel/mozilla/fx-team/obj-x86_64-apple-darwin12.5.0/dist/NightlyDebug.app/Contents/Resources/browser/extensions/%7B972ce4c6-7e08-4474-a285-3208198ce6fd%7D/chrome.manifest'.
> TypeError: setting a property that has only a getter <unknown>
> OpenGL compositor Initialized Succesfully.
> Version: 2.1 NVIDIA-8.16.80 310.40.00.20f04
> Vendor: NVIDIA Corporation
> Renderer: NVIDIA GeForce GT 650M OpenGL Engine
> FBO Texture Target: TEXTURE_2D
> TypeError: setting a property that has only a getter <unknown>
> ReferenceError: reference to undefined property link.enhancedImageURI DirectoryLinksProvider.jsm:399:0
> TypeError: setting a property that has only a getter <unknown>
> TypeError: setting a property that has only a getter <unknown>
> OpenGL compositor Initialized Succesfully.
> Version: 2.1 NVIDIA-8.16.80 310.40.00.20f04
> Vendor: NVIDIA Corporation
> Renderer: NVIDIA GeForce GT 650M OpenGL Engine
> FBO Texture Target: TEXTURE_2D
> failed to start socket: [Exception... "Component returned failure code: 0x80004005 (NS_ERROR_FAILURE) [nsIUDPSocket.init]"  nsresult: "0x80004005 (NS_ERROR_FAILURE)"  location: "JS frame :: resource://gre/modules/SimpleServiceDiscovery.jsm :: _search :: line 159"  data: no] SimpleServiceDiscovery.jsm:165:0
> ReferenceError: reference to undefined property json._dataType nsSearchService.js:3425:49
> ReferenceError: reference to undefined property param.purpose nsSearchService.js:959:8
> ReferenceError: reference to undefined property msg.app Webapps.jsm:1155:8
> 1418670983171	Services.HealthReport.HealthReporter	WARN	No prefs data found.
> push1.push.hello.firefox.com : server does not support RFC 5746, see CVE-2009-3555 <unknown>

By the way, it occurs to me that this may have the same root cause as the issue you discussed with Alexis here: http://logs.glob.uno/?c=mozilla%23loop#c22695 -- if we do, in fact, have some set of circumstances where you send a non-FxA token for auth while logged in (which seems to be the most likely cause of the issue you saw), then I would assume that any rooms created with that token wouldn't be properly associated with the FxA account.

Comment 15

[Maire Reavy [:mreavy]]

Mark -- Can I assign this to you to make sure we test this scenario after bug 1102432 and bug 1111579 are fixed?

(FYI to those reading this: we think this bug is fallout from bug 1102432 and bug 1111579, but we want to verify there isn't another bug here before closing this.)

Comment 16

[Maire Reavy [:mreavy]]

Lowering this to a P2 (based on our new priority definitions) because I strongly believe the bug here is solely on the server-side and if there were a client-side bug here, I don't think we'd block the release for it.  We would however work to fix it ASAP and see how far we could/should uplift it.

Comment 17

[Mark Banner (:standard8)]

I believe this is fixed now, I've not seen any more occurrences since the fixes in comment 15, and there's been plenty of time for expiry to happen.