1048896 - Region detection (with or without a SIM) is not working properly

Status: RESOLVED FIXED

(Marketplace Graveyard :: Consumer Pages, defect, P2)

Description

[Mathieu Pillard [:mat]]

Created as a follow-up from bug 1043263 and bug 1048259. We're having some problems with region detection, causing people to see apps they shouldn't see and things like that.

:magopian, ashes 664a6
- Using production yulelog package
- No SIM
- Hits consumer-info without region (presumably returns "fr")
- Is seen in "restofworld" anyway
- No region is passed to API calls

:geekshadow, ashes fb7a2
- Using production yulelog package
- Has a french SIM
- consumer-info  without region (returns "fr" according to a test made on the same wifi network)
- Is seen in "restofworld" anyway
- No region is passed to API calls

:gerard-majax, ashes e0ba7
- Using production yulelog package
- Has a french SIM
- Hits consumer-info without region (presumably returns "fr")
- Is seen in "germany"
- No region is passed to API calls

:krupa, ashes f3cfd
- Using stage "true" package
- Has no SIM
- Hits consumer-info *with* region and carrier from old settings
- Is seen in ?
- No region is passed to API calls

I'm pretty sure I fixed the first case in https://github.com/mozilla/fireplace/commit/2d1e1802d15ecb7c4fffce061618d73baa65b7cc, which should land in production later today, but the other issues remain. The fact that we are not using the true packaged app in production could also have consequences here.

To paraphrase myself in https://bugzilla.mozilla.org/show_bug.cgi?id=1043263#c9 :
- We should refactor clean that code up
- I'm not sure we need to store the old regions anymore now that we have consumer-info
- In any case we should hit mobilenetwork detection code before doing consumer-info stuff
- We should push the packaged app to remove some of the old hacks we had

Comment 1

[Mathieu Pillard [:mat]]

Adding a dependency on bug 897156, because being a real packaged app will allow us to clean up old hacks (for instance, I think we won't need the code that deals with passing mcc/mnc using the query string once any more we are a a packaged app).

Comment 2

[Mark Striemer [:mstriemer]]

I had cleaned up this code [1] at one point but didn't end up merging it because I wasn't sure how it interacted with the yulelog version. If we're getting rid of yulelog it might be worth revisiting. Taking out the query string stuff would be a lot easier with my update applied.

[1] https://github.com/mozilla/fireplace/pull/373/files

Comment 3

[Christopher Van Wiemeersch [:cvan]]

(In reply to Mark Striemer [:mstriemer] from comment #2)
> I had cleaned up this code [1] at one point but didn't end up merging it
> because I wasn't sure how it interacted with the yulelog version. If we're
> getting rid of yulelog it might be worth revisiting. Taking out the query
> string stuff would be a lot easier with my update applied.
> 
> [1] https://github.com/mozilla/fireplace/pull/373/files

Last time I checked, querystring params are used to test that the operator shelves show up correctly: https://docs.google.com/spreadsheet/ccc?key=0Ap2bDqlwf1F9dFZ6bFV3bk9nQXVvN1hOQXM1QVdXUmc&usp=drive_web

Comment 4

[Mathieu Pillard [:mat]]

We landed the refactor so the code is a bit more healthy now. Fixing bug 1061716 and making sure we do all mobilenetwork stuff before calling consumer-info should fix all issues, hopefully.

Comment 5

[Mathieu Pillard [:mat]]

https://github.com/mozilla/fireplace/pull/663 and the PR for bug 1061716 should fix most (all?) the remaining issues, hopefully.

Comment 6

[Mathieu Pillard [:mat]]

Should be fixed in https://github.com/mozilla/fireplace/commit/e1bb494374337d9f5c569827b9612c954925794a but waiting on bug 1061716 to be closed to mark it as fixed.

Comment 7

[Victor Carciu]

I verified this for detection with SIM(regions were correctly recognized for both Grameenphone-Bangladesh and Movistar-Spain SIMs), but I am not able to verify this for NO SIM scenario.

Krupa, can you please verify this scenario in US?