870411 - Auto-login to WebPay for repeat payments is not working

Status: RESOLVED FIXED

(Marketplace Graveyard :: Payments/Refunds, defect, P1)

Description

[Kumar McMillan [:kumar]]

STR
- configure your device for dev/stage payments per https://webpay.readthedocs.org/en/latest/use_hosted_webpay.html
- open Marketplace dev or stage, such as Marketplace Stage at https://marketplace.allizom.org/ (installable from http://people.mozilla.com/~kmcmillan/mktdev.html)
- search for Private Yacht
- click Buy
- log in to the Marketplace via Persona (TrustedUI title says Sign In)
- proceed to the WebPay server (TrustedUI title says Marketplace Stage)

Expected: you should be automatically logged in to WebPay because these servers are on the same origin

Actual: you see a Sign In button which you have to click each time to proceed

You can also reproduce this by signing into WebPay, clicking cancel to return to Marketplace, and pressing the Buy button again to begin a new pay flow.

Note: this is *only* happening on the 1.0.1 branch. Here is a build where I confirmed it but it's not necessarily the one that broke it:

  https://git.mozilla.org/?p=releases/gecko.git;a=commitdiff;h=af474bd146240010dfb700e0c6cb70a515288c71
  https://git.mozilla.org/?p=releases/gaia.git;a=commitdiff;h=30ad443fbe5363a7381c19c6bf9d6ca49228fb8b

Comment 1

[Kumar McMillan [:kumar]]

1.1 where I confirmed it is working as expected:

  https://git.mozilla.org/?p=releases/gecko.git;a=commitdiff;h=4e5b74b6ab7ae7a9b5e07cdd63fac4965ef5398d
  https://git.mozilla.org/?p=releases/gaia.git;a=commitdiff;h=4e7d63a83508caa391c4db164c3f68422d9ca5b6

Comment 2

[Kumar McMillan [:kumar]]

I'm doing a manual bisect and I went back to 1.0.1 from more than a week to:
https://pvtbuilds.mozilla.org/pub/mozilla.org/b2g/nightly/mozilla-b2g18_v1_0_1-unagi-eng/2013/04/2013-04-29-07-02-04/unagi.zip

and it is still broken.

Build info:
  https://git.mozilla.org/?p=releases/gecko.git;a=commitdiff;h=b9338ed4da8b391aa9f916902cb347fd0b8e9192
  https://git.mozilla.org/?p=releases/gaia.git;a=commitdiff;h=cf2d4136f0ebc66039637fdbeb72ed184dfbc0f2

Comment 3

[Kumar McMillan [:kumar]]

Looks like it has been broken for a while. I still see breakage in 1.0.1 from https://pvtbuilds.mozilla.org/pub/mozilla.org/b2g/nightly/mozilla-b2g18_v1_0_1-unagi-eng/2013/04/2013-04-22-07-02-06/unagi.zip

Build info:
  https://git.mozilla.org/?p=releases/gecko.git;a=commitdiff;h=ea6a890d8bc484eeaa64f55aa4abeaad24eae683
  https://git.mozilla.org/?p=releases/gaia.git;a=commitdiff;h=b750757662209dad81942df13f9997d684e0d9b1

Comment 4

[Jed Parsons (use needinfo, please) [:jedp, :jparsons]]

kumar check it out and see if the fix works for you

we pushed the fix to native-persona.org as well as login.dev.anosrep.org (the browserid dev server, which, btw, is where the new ux is).

I think what happened here is that this used to work, and then the memshrink patch landed in b2g18, which guaranteed that this would continue to work (by ensuring that no old data could ever be cached), and in that period, the earlier state-management code started to diverge.  Another way of saying that is that I screwed up.  I really should have been more on top of that.  I'm sorry for the churn this caused, especially at a time when we're all so busy.

Comment 5

[Kumar McMillan [:kumar]]

Yep! Thanks very much for the quick fix, it is working now on today's 1.0.1 after https://bugzilla.mozilla.org/show_bug.cgi?id=870412#c6 . I tried several times to be sure. All is good.

No worries. Another problem is on our end: we were accidentally doing all testing on the v1-train instead of 1.0.1 otherwise we would have caught it sooner.

Comment 6

[Jed Parsons (use needinfo, please) [:jedp, :jparsons]]

(In reply to Kumar McMillan [:kumar] from comment #5)
> Yep! Thanks very much for the quick fix, it is working now on today's 1.0.1
> after https://bugzilla.mozilla.org/show_bug.cgi?id=870412#c6 . I tried
> several times to be sure. All is good.

Great!  Good to hear.

> No worries. Another problem is on our end: we were accidentally doing all
> testing on the v1-train instead of 1.0.1 otherwise we would have caught it
> sooner.

Well, that goes for us, too.  btw :zac was very helpful in helping me confirm the problems last night when I wasn't sure what I was seeing on hotel wifi, and he jumped in with more regression tests today.  w00t!