1388722 - browser_ext_user_events.js fails on windows 10 inside a VM with: Expect widget not to be overflowed

Status: RESOLVED WORKSFORME

(WebExtensions :: Untriaged, defect, P3)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

there are 3 browser chrome tests that need fixing in order to get browser chrome running on windows 10 in automation- the other 2 are specific windows 10 tests, this is the only test that runs/passes on windows 8 which doesn't on windows 10.

Here is a log from a try push:
https://public-artifacts.taskcluster.net/DfgCGecPT0CSKKdl2ZHldQ/0/public/logs/live_backing.log

and some of what we see in the log:
21:16:56     INFO -  674 INFO Console message: Warning: attempting to write 17660 bytes to preference extensions.webextensions.uuids. This is bad for general performance and memory usage. Such an amount of data should rather be written to an external file. This preference will not be sent to any content processes.
21:16:56     INFO -  675 INFO Console message: 1502227016236	addons.webextension.{9034371a-f349-44b5-96fe-7228c3f4767c}	WARN	Please specify whether you want browser_style or not in your browser_action options.
21:16:56     INFO -  676 INFO Console message: 1502227016236	addons.webextension.{9034371a-f349-44b5-96fe-7228c3f4767c}	WARN	Please specify whether you want browser_style or not in your page_action options.
21:16:56     INFO -  677 INFO TEST-PASS | browser/components/extensions/test/browser/browser_ext_user_events.js | request() did not throw when called from page action click -
21:16:56     INFO -  678 INFO TEST-PASS | browser/components/extensions/test/browser/browser_ext_user_events.js | request() succeeded when called from page action click -
21:16:56     INFO -  Buffered messages finished
21:16:56    ERROR -  679 INFO TEST-UNEXPECTED-FAIL | browser/components/extensions/test/browser/browser_ext_user_events.js | Expect widget not to be overflowed -
21:16:56     INFO -  Stack trace:
21:16:56     INFO -  chrome://mochitests/content/browser/browser/components/extensions/test/browser/head.js:showBrowserAction:238
21:16:56     INFO -  680 INFO TEST-PASS | browser/components/extensions/test/browser/browser_ext_user_events.js | request() did not throw when called from browser action click -
21:16:56     INFO -  681 INFO TEST-PASS | browser/components/extensions/test/browser/browser_ext_user_events.js | request() succeeded when called from browser action click -
21:16:56     INFO -  682 INFO TEST-PASS | browser/components/extensions/test/browser/browser_ext_user_events.js | Found context menu item -
21:16:56     INFO -  683 INFO TEST-PASS | browser/components/extensions/test/browser/browser_ext_user_events.js | request() did not throw when called from context menu click -
21:16:56     INFO -  684 INFO TEST-PASS | browser/components/extensions/test/browser/browser_ext_user_events.js | request() succeeded when called from context menu click -


As you can see most of the code works.  Looking into this error message, I see code in head.js:
var showBrowserAction = async function(extension, win = window) {
  let group = getBrowserActionWidget(extension);
  let widget = group.forWindow(win);

  if (group.areaType == CustomizableUI.TYPE_TOOLBAR) {
    ok(!widget.overflowed, "Expect widget not to be overflowed");
  } else if (group.areaType == CustomizableUI.TYPE_MENU_PANEL) {
    await win.document.getElementById("nav-bar").overflowable.show();
  }
};

here is a link to the source code (which was introduced in bug 1217129):
http://searchfox.org/mozilla-central/source/browser/components/extensions/test/browser/head.js#

I would like to know if there is any reason why this would happen?  Also, I see these are extensions test case- is this web extensions, or legacy addons?

Comment 1

[Joel Maher ( :jmaher ) (UTC -8)]

:kmag, checking in here- this is the last remaining test keeping browser-chrome from running on windows10, if you have suggestions of what to do, I could hack on this, etc.

Comment 2

[Andy McKay]

Is this related to overflow code you've been working recently Gijs?

Comment 3

[:Gijs (he/him)]

(In reply to Andy McKay [:andym] from comment #2)
> Is this related to overflow code you've been working recently Gijs?

Kind of. What size is the screen on this VM? Are there screenshots of the state of the browser at this point?

Comment 4

[Joel Maher ( :jmaher ) (UTC -8)]

here is a try push (bc1,bc5 are this test case):
https://treeherder.mozilla.org/#/jobs?repo=try&revision=a130cbe134b98aa618f812e9964bf945035632f4&selectedJob=121999094

here is a screenshot:
https://public-artifacts.taskcluster.net/UoErByUlQ2q47iAnyx1dQA/0/public/test_info/mozilla-test-fail-screenshot_71vslf.png

Oddly running the test locally on my win10 development laptop with |./mach test browser/components/extensions/test/browser/browser_ext_user_events.js| results in success.

But running with |./mach test browser/components/extensions/test/browser| results in so many failures, many of them "Expect widget not to be overflowed" - so something might be odd on windows 10.

As a note, we run the test in two modes and they seem to fail in both modes (regular and oop)- not sure if we need both modes in our modern browser.

this is a VM- I need to figure out what the screen resolution is

:grenade, do you know how I can find the screen resolution on a windows10 VM?

Comment 5

[:Gijs (he/him)]

(In reply to Joel Maher ( :jmaher) (UTC-5) from comment #4)
> here is a try push (bc1,bc5 are this test case):
> https://treeherder.mozilla.org/#/
> jobs?repo=try&revision=a130cbe134b98aa618f812e9964bf945035632f4&selectedJob=1
> 21999094
> 
> here is a screenshot:
> https://public-artifacts.taskcluster.net/UoErByUlQ2q47iAnyx1dQA/0/public/
> test_info/mozilla-test-fail-screenshot_71vslf.png
> 
> Oddly running the test locally on my win10 development laptop with |./mach
> test browser/components/extensions/test/browser/browser_ext_user_events.js|
> results in success.
> 
> But running with |./mach test browser/components/extensions/test/browser|
> results in so many failures, many of them "Expect widget not to be
> overflowed" - so something might be odd on windows 10.
> 
> As a note, we run the test in two modes and they seem to fail in both modes
> (regular and oop)- not sure if we need both modes in our modern browser.
> 
> this is a VM- I need to figure out what the screen resolution is
> 
> :grenade, do you know how I can find the screen resolution on a windows10 VM?

The screenshot shows that it's tiny. :-)

The screenshot also shows that lots of random items are in the toolbar that aren't normally there. I don't know why, because this test doesn't add those items. I expect it might be due to https://dxr.mozilla.org/mozilla-central/source/browser/components/extensions/test/browser/browser_ext_themes_icons.js leaving stuff hanging around for some reason. I expect the screen resolution is the cause of this now going orange on win10, but the fundamental issue already existed. When we worked on changing the default contents of the navbar (in bug 1363485), we had to adapt the test to drop the sidebar button because otherwise the navbar was too crowded, apparently. We didn't realize then, but I bet the root cause is the same.

Comment 6

[Joel Maher ( :jmaher ) (UTC -8)]

:gijs your assertion that the browser_ext_themes-icons.js is not cleaning up and affecting the test is correct:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=43a65cc1955a86d4881862e95e389f4ca54c31e2

sadly there are new failures, is this something we could disable the one test on so we can stop the new failures from showing up and work on fixing it, or is this providing significant coverage on 64 bit windows that we don't get on 32 bit windows?

Comment 7

[:Gijs (he/him)]

(In reply to Joel Maher ( :jmaher) (UTC-5) from comment #6)
> :gijs your assertion that the browser_ext_themes-icons.js is not cleaning up
> and affecting the test is correct:
> https://treeherder.mozilla.org/#/
> jobs?repo=try&revision=43a65cc1955a86d4881862e95e389f4ca54c31e2

We should just fix the themes test. Shouldn't be hard. I'll try to have a look either late today or on Monday.

> sadly there are new failures, is this something we could disable the one
> test on so we can stop the new failures from showing up and work on fixing
> it, or is this providing significant coverage on 64 bit windows that we
> don't get on 32 bit windows?

I don't really understand this question though. What are you suggesting?

I'm not convinced the parsable_css failure is related to this test. Don't know for sure about the CUI stuff. You pushed to try from a different base rev so I don't know if it's the same issue or not. :-)

Comment 8

[Joel Maher ( :jmaher ) (UTC -8)]

thanks for looking into this bug.

My point is that if I had just disabled 3 tests last week, we could have had coverage for this and not had the new test failures.  I would like to look at what errors remain and then either get them fixed early next week or disabled temporarily so we can get win10 coverage without playing the game of waiting for fixes, then finding new failures, repeat.

I will see if parsable_css failure is fixed in a different revision :)

Comment 9

[:Gijs (he/him)]

(In reply to Joel Maher ( :jmaher) (UTC-5) from comment #8)
> thanks for looking into this bug.
> 
> My point is that if I had just disabled 3 tests last week, we could have had
> coverage for this and not had the new test failures.  I would like to look
> at what errors remain and then either get them fixed early next week or
> disabled temporarily so we can get win10 coverage without playing the game
> of waiting for fixes, then finding new failures, repeat.

This is fair. Let's do that, then, and I will be happy to try to fix the themes test once stuff is up and running. Just ping me on the relevant bug.

Comment 10

[Joel Maher ( :jmaher ) (UTC -8)]

adjusting the screen resolution seems to solve this- I am happy to call this wontfix or worksforme.