1340786 - toolkit/content/tests/chrome/test_largemenu.html fails when run on windows7 vm (instead of hardware)

Status: NEW

(Toolkit :: UI Widgets, defect, P5)

Description

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

an small bit from the log [0]:
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | open with scrolling bottom 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | open with scrolling overflow 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu scroll position 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu scroll position 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | open small again top 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | open small again bottom 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | open small again overflow 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu scroll position 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement (1, 1) x 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement (1, 1) y 
13:25:35     INFO - Buffered messages finished
13:25:35     INFO - TEST-UNEXPECTED-FAIL | toolkit/content/tests/chrome/test_largemenu.xul | menu movement (100, 8000) x - got 1280, expected 100
13:25:35     INFO - SimpleTest.is@chrome://mochikit/content/tests/SimpleTest/SimpleTest.js:271:5
13:25:35     INFO - is@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:164:24
13:25:35     INFO - testPopupMovement@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:343:3
13:25:35     INFO - popupShown@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:98:12
13:25:35     INFO - onpopupshown@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:1:1
13:25:35     INFO - Not taking screenshot here: see the one that was previously logged
13:25:35     INFO - TEST-UNEXPECTED-FAIL | toolkit/content/tests/chrome/test_largemenu.xul | menu movement (100, 8000) y - got 864, expected 688
13:25:35     INFO - SimpleTest.is@chrome://mochikit/content/tests/SimpleTest/SimpleTest.js:271:5
13:25:35     INFO - is@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:164:24
13:25:35     INFO - testPopupMovement@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:344:3
13:25:35     INFO - popupShown@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:98:12
13:25:35     INFO - onpopupshown@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:1:1
13:25:35     INFO - Not taking screenshot here: see the one that was previously logged
13:25:35     INFO - TEST-UNEXPECTED-FAIL | toolkit/content/tests/chrome/test_largemenu.xul | menu movement (6000, 100) x - got 2805, expected 1205
13:25:35     INFO - SimpleTest.is@chrome://mochikit/content/tests/SimpleTest/SimpleTest.js:271:5
13:25:35     INFO - is@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:164:24
13:25:35     INFO - testPopupMovement@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:357:3
13:25:35     INFO - popupShown@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:98:12
13:25:35     INFO - onpopupshown@chrome://mochitests/content/chrome/toolkit/content/tests/chrome/window_largemenu.xul:1:1
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement (6000, 100) y 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement left is empty after moving 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement top is empty after moving 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement set left and top x 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement set left and top y 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement move after set left and top x 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement move after set left and top y 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement left is set after moving 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement top is set after moving 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement move after set left and top x to -1 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement move after set left and top y to -1 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement left is not set after moving to -1 
13:25:35     INFO - TEST-PASS | toolkit/content/tests/chrome/test_largemenu.xul | menu movement top is not set after moving to -1 
13:25:


Fixing this bug will allow us to move over all of mochitest-chrome to a VM and reducing the load we have on hardware!


https://archive.mozilla.org/pub/firefox/try-builds/[e-mail]-657209ff283794a666a2302a78a9368ca2ab0674/try-win32-debug/try_win7_vm_gfx-debug_test-mochitest-chrome-3-bm128-tests1-windows-build5.txt.gz

Comment 1

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

Neil, can you help us figure this out?  This happens both on opt and debug in the same ways.

Comment 2

[Neil Deakin]

The numbers suggest something odd about the screen configuration. There is only one screen configured, right?

Do you have a screenshot of the failure?

Comment 3

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

oh, sorry I didn't think to do that at first.  here is a try push:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=657209ff283794a666a2302a78a9368ca2ab0674&filter-searchStr=win%20vm%20mochitest-chrome&exclusion_profile=false&selectedJob=78440932

and a screenshot from opt:
http://mozilla-releng-blobs.s3.amazonaws.com/blobs/try/sha512/b5e377f3d4d44a1361701ce555774c241f3f5d46a17daf15d411412a66294d1a1b195909b19040e6095c09bc78e3356974f12205318b30a9f2d7bcc97c8ee8ab

as far as I know there is only one screen (headless virtual screen that is- yay for hacking on windows)

Comment 4

[Neil Deakin]

The screenshot suggests some kind of rendering issue; the popup is drawn multiple times yet there is only of them. The popup should be drawn near the bottom of the screen.

Comment 5

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

any tips on how to debug this?  would some more logging in the test case help?  Any system settings that we could do that could help (windows registry settings, etc.)?

From looking at the logs it appears we have 1280x1024 resolution both on hardware and vm- maybe that is not what it really is, although finding screenshots for win7-hw failures is hard- I did find one and it looks to be similar sized screenshot:
http://mozilla-releng-blobs.s3.amazonaws.com/blobs/mozilla-inbound/sha512/50f30a6dc342d6ea8ff90808ed45f8489e9863d784f1460daf0c110044f96220f6a5537da700b9e0319746e13ec005e8827386631379b88fe3b11502902b7ec8

Comment 6

[Armen [:armenzg]]

There's a difference between the two screenshots.

One of the versions of Windows is running without using all graphic fanciness (no transparencies). Can we assume that has nothing to do? (since my assumption is that we have failures in both machines with graphics and without).

Neil: anything we could try to investigate a bit more?

Comment 7

[Neil Deakin]

Does this fail the same everytime?

The part of the test that is failing is trying to move a menu to a position off the bottom of the screen, but it should be pushed up to be visible along the bottom edge of the screen.

You could check to see what the various coordinates and sizes of the screen are as those are used to compute the values in the test that is failing.

Comment 8

[Neil Deakin]

Attachment: Extra debugging

This adds that extra debugging info.

Comment 9

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

Neil, this is the only test preventing us from moving mochitest-chrome to a VM; is it ok to skip this on windows?  maybe we move this to the 'clipboard' suite which always runs on hardware?

Comment 10

[Neil Deakin]

Did you try running with the debugging patch from comment 8?

Otherwise, I would suggest trying to run this in the clipboard set if that works, as this test checks some platform-specific behaviour that would go untested otherwise.

Comment 11

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

my try push is here:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=c8aecad1255a4164f9ecf8a169f3c4b381fe3c87&filter-tier=1&filter-tier=2&filter-tier=3&filter-searchStr=c3&exclusion_profile=false

there is a failure on a gpu instance and a green on a non-gpu instance (target platform).  I did some retriggers, lets see what shakes out!

Comment 12

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

ok, so looking at the c3 failures they are all on gpu instances (not our target), Neil, is there data you can harvest from this patch?  Should I push without your patch and see if things are 'magically' fixed?

Comment 13

[Neil Deakin]

The results show that the screen information is correct yet the popup is indeed appearing horizontally offscreen.

It does sound like the is specific to the machine configuration. I can produce a much more detailed debugging patch if desired with lots of extra logging if we want to investigate further. Do I need to do anything special to push my own try-builds for testing?

Comment 14

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

ok, here is a try push with no extra patches:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=65b74b3926cac3bb10e95908e7a2f1ced668fc15&filter-tier=1&filter-tier=2&filter-tier=3

try pushes are as simple as:
try: -b o -p win32 -u mochitest-chrome -t none

that will run win7 hardware, and win7 vm- but the win7-vm jobs are excluded, so you need to get that showing on treeherder with a toggle near the top of the webpage.


you can see that all the regular instances are passing.  Should we debug why the gpu instances fail?  I think we can move forward and port these to windows7-vm; I would recommend closing this bug as worksforme, but happy to help debug more.

Comment 15

[Neil Deakin]

Debugging shows that the failing machines have two screens configured:

The first is at position 0,0 with a size of 1280,1024
The second is to the right of the first at 1280,0 and ends at 1600,1200. That has a size of 320x1200 which is an unusual configuration.

Passing machines have only the first.

Comment 16

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

oh, interesting find, Q: do you know why the gpu instances in aws have a second screen available to the win7 session?  Possibly this could hurt us in some areas, but maybe help us in others.

Comment 17

[Q]

I don't know why it would be inconsistent with some machines having it and some not. However, there are two video devices a cirrus card and an nvidia card on a GPU deivce. There may be some tweaking to the fakemon script to make sure that the second screen is disabled.

Q

Comment 18

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

Greg, another one for you to look at when you get bored :)

Comment 19

[Chris AtLee [:catlee]]

(In reply to Q from comment #17)
> I don't know why it would be inconsistent with some machines having it and
> some not. However, there are two video devices a cirrus card and an nvidia
> card on a GPU deivce. There may be some tweaking to the fakemon script to
> make sure that the second screen is disabled.
> 
> Q

Q, can you take a look to make sure that the fakemon script is doing the right thing here?

Comment 20

[Q]

Apologies folks got my wires crossed here and thought this was being taken care of in Rob's TC work. However, Looking back at old bugs I can see this is a buildbot request. Are we still seeing this failure ?

Comment 21

[Firefox Bug Husbandry Bot]

https://wiki.mozilla.org/Bug_Triage/Projects/Bug_Handling/Bug_Husbandry#Move_fix-optionals