979924 - b2g-bin wrong min-device-height on CSS media queries

Status: RESOLVED WONTFIX

(Firefox OS Graveyard :: General, defect)

Description

[Miller Medeiros [:millermedeiros]]

b2g reports the wrong device-height, it's using the host device-height instead of the window height. so media queries like `@media (min-device-height: 640px)` will be triggered when it shouldn't, which might cause errors on Travis/Marionette tests. - I had problems with the Calendar integration tests because of that.

Comment 1

[Andrew Sutherland [:asuth] (he/him)]

Looks like this is likely impacted by the docshell.device_size_is_page_size pref.
http://dxr.mozilla.org/mozilla-central/source/docshell/base/nsIDocShell.idl#950

We set this in the GAIA profile when DEBUG is set (in gaia/build/preferences.js).  We check that this is the case for DEBUG in gaia/test/integration/build.test.js.

This matters in:
http://dxr.mozilla.org/mozilla-central/source/layout/style/nsMediaFeatures.cpp#113

If the pref is true, we use the nsPresContext's dimensions.  If false, we use the device context's dimensions.  This is probably the X server:
http://dxr.mozilla.org/mozilla-central/source/gfx/src/nsDeviceContext.h#144


It seems like we/everyone probably always want this pref to be true when we are running in b2g-desktop.  Which suggests we want to put it in an "#ifndef MOZ_WIDGET_GONK" section in http://dxr.mozilla.org/mozilla-central/source/b2g/app/b2g.js

needinfo-ing Fabrice.

Comment 2

[[:fabrice] Fabrice Desré]

I'm probably an outlier there, but I don't want that in b2g desktop, not more than you want that in firefox desktop because of the potential impact on web sites.

Comment 3

[Andrew Sutherland [:asuth] (he/him)]

(In reply to Fabrice Desré [:fabrice] from comment #2)
> I'm probably an outlier there, but I don't want that in b2g desktop, not
> more than you want that in firefox desktop because of the potential impact
> on web sites.

I'm having trouble parsing your meaning. I think I my have left some things implied so will re-state.

Problem: Based on Miller's experience, media queries for device-height in our JS Marionette tests run on b2g-desktop seem to be using the dimensions of the X server.

Possible solutions:

* Don't use device-width/device-height in Gaia apps, just use width/height.

* Address this directly in JS Marionette tests by either:
** forcing the xvfb we're running under to be our canonical device resolution
** explicitly setting the pref in our JS Marionette testing profile

* Make b2g-desktop (aka b2g built with non-gonk bits) always do this.  If we assume that b2g-desktop is always used in a simulator-type context, then the size of the b2g-desktop window is the size of the simulated device.  As noted on bug 948045, the pref does fail to take into account the status-bar size, so it's also possible the most right fix is to make some other change to accomplish a similar effect.

It sounds like you are saying that we are expecting people to use b2g-desktop to browse or test websites and that they will want to use it not as a device simulator, but as a variant of Firefox.  But I could be misinterpreting.  (Are you trying to get multiprocess Firefox by running b2g-desktop? :)


cc'ing those involved in bug 948045 that introduced the pref to the gaia build step since we probably want more consistency here in general.

Comment 4

[Andrew Sutherland [:asuth] (he/him)]

Also, I was generally asking in a "Fabrice knows lots of gecko internals, what would be the right way to fix this problem?", but I was assuming it's a problem and we agreed on the desired outcome.  Totally fine to beg the question! :)

Comment 5

[[:fabrice] Fabrice Desré]

Ha ha, yes I'm probably one a the few that believe we should not set docshell.device_size_is_page_size to true by default on desktop builds.

(In reply to Andrew Sutherland (:asuth) from comment #3)
> 
> Possible solutions:
> 
> * Don't use device-width/device-height in Gaia apps, just use width/height.

Isn't that risky?

> * Address this directly in JS Marionette tests by either:
> ** forcing the xvfb we're running under to be our canonical device resolution
> ** explicitly setting the pref in our JS Marionette testing profile

That sounds reasonable to me.

> * Make b2g-desktop (aka b2g built with non-gonk bits) always do this.  If we
> assume that b2g-desktop is always used in a simulator-type context, then the
> size of the b2g-desktop window is the size of the simulated device.  As
> noted on bug 948045, the pref does fail to take into account the status-bar
> size, so it's also possible the most right fix is to make some other change
> to accomplish a similar effect.

By status bar size, do you mean the xul chrome one from the simulator or gaia's top status bar?

> It sounds like you are saying that we are expecting people to use
> b2g-desktop to browse or test websites and that they will want to use it not
> as a device simulator, but as a variant of Firefox.  But I could be
> misinterpreting.  (Are you trying to get multiprocess Firefox by running
> b2g-desktop? :)

I totally think we should move to b2g-based products everywhere, but that's another debate as you can imagine ;)

Comment 6

[BMO Automation]

Firefox OS is not being worked on