1104644 - Mulet crash running Gaia marionnette

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::Build, defect)

Description

[:gerard-majax]

Because of MOZ_DISABLE_NONLOCAL_CONNECTIONS

Comment 1

[:gerard-majax]

Because of tiles.

> Tue, 25 Nov 2014 10:59:10 GMT b2g-desktop stderr FATAL ERROR: Non-local network connections are disabled and a connection attempt to tiles.services.mozilla.com (54.191.174.127) was made.

Comment 2

[:gerard-majax]

Once tiles are disabled, we sill have forced update checking that kicks in and breaks, too.

Comment 3

[:gerard-majax]

Ricky, I can't find in the current build system how to add specific prefs when running Marionnette tests.

Comment 4

[:gerard-majax]

Attachment: Gaia PR

Comment 5

[:gerard-majax]

Getting crashed on marketplace.firefox.com requests now.

Comment 6

[:gerard-majax]

I can't find where this marketplace.firefox.com request is coming from: it's using https, and even dumping ssl master keys and using wireshark I could not decipher.

Comment 7

[:gerard-majax]

Comment on attachment 8528357 [details] [review]
Gaia PR

Kevin, those are needed, but maybe not enough for all cases.

Comment 8

[Kevin Grandon :kgrandon]

Comment on attachment 8528357 [details] [review]
Gaia PR

I don't know what all of these prefs do, but the commit message is insightful and makes sense. TBPL is also green, so I'm happy. Thanks!

Comment 9

[Ed Morley [:emorley]]

(In reply to Alexandre LISSY :gerard-majax from comment #6)
> I can't find where this marketplace.firefox.com request is coming from: it's
> using https, and even dumping ssl master keys and using wireshark I could
> not decipher.

I see quite a few hits here:
http://mxr.mozilla.org/mozilla-central/search?string=marketplace.firefox.com&case=1
http://mxr.mozilla.org/gaia/search?string=marketplace.firefox.com&case=1

Comment 10

[:gerard-majax]

Ok, those requests are somehow related to the manifestURL in dev_apps/marketplace.allizom.org/metadata.json and dev_apps/marketplace-dev.allizom.org/metadata.json

But why would this hit only Mulet?

Comment 11

[:gerard-majax]

Johnathan, do you think you could use a bit of brain juice to help me? :)

Comment 12

[Jonathan Griffin (:jgriffin)]

The reason this only hits Mulet is because older B2G tests (like those running on b2gdesktop builds) were grandfathered in before this requirement took effect.

Marketplace is unique among stock tes apps in that it's hosted, rather than packaged.  I don't know if we can realistically block it from loading in TBPL.  Leaving the needinfo for James Lal who might know.

Comment 13

[:gerard-majax]

(In reply to Jonathan Griffin (:jgriffin) from comment #12)
> The reason this only hits Mulet is because older B2G tests (like those
> running on b2gdesktop builds) were grandfathered in before this requirement
> took effect.

That's just false: I checked it by running locally. With B2G Desktop and the flag, it passed. With Mulet and the flag, it crashes.

Comment 14

[:gerard-majax]

Attachment: B2G vs Mulet

Comment 15

[Ed Morley [:emorley]]

Do we just need the equivalent of bug 1095466? Or more specifically:
https://github.com/mozilla-b2g/gaia/pull/26004/files#diff-2ef9d5e3555ce347b941bc231649e42bR8

Comment 16

[:gerard-majax]

So, those requests are indeed because of Webapps.jsm app update checking. This is triggered by a Gaia-side event, "force-update-check", coming from a change to the setting value of |gaia.system.checkForUpdates|.

Comment 17

[:gerard-majax]

PR updated with the added pref from bug 1094362.

Comment 18

[:gerard-majax]

Try is green: https://tbpl.mozilla.org/?tree=Gaia-Try&rev=480b74e32671

Comment 19

[:gerard-majax]

https://github.com/mozilla-b2g/gaia/commit/09da77088026f0ecb42e833222e4d009a0b4d9bf