1236194 - Intermittent testSessionHistory | application crashed [@ nsSocketTransport::InitiateSocket], FATAL ERROR: Non-local network connections are disabled and a connection attempt to webcompat.com (162.243.90.227) was made.

Status: RESOLVED FIXED

(Firefox for Android Graveyard :: Testing, defect)

Description

[Phil Ringnalda (:philor)]

https://treeherder.mozilla.org/logviewer.html#?job_id=2985978&repo=mozilla-central

Comment 1

[Mark Finkle (:mfinkle) (use needinfo?)]

Looks like this could be from the Webcompat add-on, which should be nightly-only:
http://mxr.mozilla.org/mozilla-central/source/mobile/android/chrome/content/WebcompatReporter.js#90

I assume URL (the object) tries to make an actual connection, maybe for DNS? Let's just switch this to use a string and stop using URL, unless there is a good reason.

Comment 2

[Mike Taylor [:miketaylr]]

Interesting, will poke at this today.

Comment 3

[Mike Taylor [:miketaylr]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=82739c35fc62

Comment 4

[Mike Taylor [:miketaylr]]

Attachment: 1236194.-Use-a-string-rather-than-URL-interface-.patch

Passing robocop locally (and verified that it still works), but I've pushed to try just in case™.

Comment 5

[Mark Finkle (:mfinkle) (use needinfo?)]

Comment on attachment 8703701 [details] [diff] [review]
1236194.-Use-a-string-rather-than-URL-interface-.patch

r++++, just for the fancy-factor!

Comment 6

[Mike Taylor [:miketaylr]]

Ahaha, thanks.

Comment 7

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/e930a107b964

Comment 8

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/e930a107b964

Comment 9

[Phil Ringnalda (:philor)]

Certainly did slow it down a lot, but even without being able to tell how many failures were ignored or starred incorrectly, it's still failed another half-dozen times since.

Comment 10

[Geoff Brown [:gbrown]]

Some of these failures have been reported in bug 1167565.

https://treeherder.mozilla.org/logviewer.html#?repo=mozilla-inbound&job_id=21097945
https://treeherder.mozilla.org/logviewer.html#?repo=fx-team&job_id=6961026
https://treeherder.mozilla.org/logviewer.html#?repo=mozilla-inbound&job_id=20825476
https://treeherder.mozilla.org/logviewer.html#?repo=mozilla-aurora&job_id=1868126

Comment 11

[Mike Taylor [:miketaylr]]

Thanks, taking a look now.

Comment 12

[Mike Taylor [:miketaylr]]

Hey Sebastian, do you have any insight as to why this might be trying to load webcompat.com, from the "Report Site Issue" menu item?

This particular intermittant happens from  testSessionHistory, which uses NavigationHelper to load 3 robocop test URLs and go back and forward in history, then reload a page.

ToolbarComponent.java defines the methods to get the back,forward and reload buttons:

https://dxr.mozilla.org/mozilla-central/source/mobile/android/tests/browser/robocop/src/org/mozilla/gecko/tests/components/ToolbarComponent.java#111-124

Is it somehow possible that findViewById is grabbing the wrong menu item? (which is...strange, given that the menu item is lazily created from JS)

https://dxr.mozilla.org/mozilla-central/source/mobile/android/chrome/content/WebcompatReporter.js#58


(lazy "fix" for this bug would add webcompat.com to build/pgo/server-location.txt, but that doesn't explain *why* this failure happens)

Comment 13

[Sebastian Kaspari (:sebastian; :pocmo)]

Hm, this is weird. Looking at the log [1] this is happening pretty late.

We first load three urls, go back two times and then when we want to navigate forward the test fails:

After opening the menu (in order to click on "forward") the test stops with the error. If the test is clicking the menu item then we should see the following two log items too (see AppComponent.pressMenuItem()):
* The menu item %s is enabled
* The menu item %s is visible

But they are missing.

We are currently fixing an issue at the exactly same spot (goForward) in bug 1247366. At first I thought they might be related but the missing click log is confusing me.

[1] https://treeherder.mozilla.org/logviewer.html#?repo=mozilla-inbound&job_id=20825476

Comment 14

[Sebastian Kaspari (:sebastian; :pocmo)]

I see this for one of my patches failing:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=1cb9adabb9e9&selectedJob=16675352

I can reproduce the test failure locally up to "TEST-UNEXPECTED-FAIL | testSessionHistory | Menu is open -". But I do not see the connection error locally. Is this expected?

And I saw this only if we are in the experiment with the longer menu, so maybe this is related to bug 1247366 after all.

Comment 15

[Mike Taylor [:miketaylr]]

> But I do not see the connection error locally. Is this expected?

I don't think the connection error is every expected. ^_^

I suppose we just wait to see if it happens again after Bug 1247366 lands, that's as good a theory as any other (that I can think of).

Comment 16

[Sebastian Kaspari (:sebastian; :pocmo)]

My comments above have been false alarms. I was looking at the failure summary (which lists this bug here) and not at the logs: This have not been issues with a network connection.

However I landed the patch in bug 1247366 today and I'm curious to see if this helped here too.

Comment 17

[Mike Taylor [:miketaylr]]

Also curious, let's see if any more of these yellows show up. Thanks Sebastian!

Comment 18

[Mike Taylor [:miketaylr]]

So, it's been 2 months. I'm gonna assume this is fixed by bug 1247366 (or something else!).