$mdUtil.disableScrollAround breaks the page when body has overflow-x:hidden.

[sebakerckhof]

Actual Behavior:

• What is the issue? *
  If a material component with disableParentScroll set to true (like mdDialog, mdSelect, ...) is opened, it calls $mdUtil.disableScrollAround. If the body has overflow-x:hidden; in its computed style, then $mdUtil.disableScrollAround will add overflow: hidden to the body (here: https://github.com/angular/material/blob/master/src/core/util/util.js#L298-L300). This breaks the page.
• What is the expected behavior?
  Same as without overflow-x: hidden

CodePen (or steps to reproduce the issue):
http://codepen.io/sebakerckhof/pen/bwQdaV
Scroll down & either open the dialog by clicking the button, or open the select element.

Angular Versions:

• Angular version: 1.5.9
• Angular Material Version: 1.1.1

Additional Information:

• Browser Type: Chrome, Firefox

Somehow adding position:fixed; width:100%; top: -Xpx makes the body.clientWidth smaller when the body has overflow-x:hidden; so that https://github.com/angular/material/blob/master/src/core/util/util.js#L298-L300 evaluates to true.
No idea why that is though.

[crisbeto]

Dialog and select don't use mdPanel yet. This is probably an issue with the $mdUtil.disableScrollAround method.

[sebakerckhof]

Ow, you're right.

[crisbeto]

Seems like this one's causing the issues: documentElement.style.overflowY = 'scroll';. If we move it a bit down, it works.

[sebakerckhof]

Seems to make sense. It probably adds a scrollbar, making the body smaller?

I could make a PR if that's encouraged. But do you know the rationale behind:
https://github.com/angular/material/blob/master/src/core/util/util.js#L298-L300
and:
https://github.com/angular/material/blob/master/src/core/util/util.js#L295
?

Just want to make sure swapping these doesn't brake anything else.

[crisbeto]

I think the rationale is that it adds an extra scrollbar to prevent the content from jumping when the body becomes non-scrollable. I'll check it up later since I ended up breaking stuff the last time I messed around with that function.

[crisbeto]

Alright, it doesn't seem like it's breaking anything. I can do the PR since I want to try and get a unit test for this as well.

[sebakerckhof]

Ok. Thanks