fix(cdk/overlay): scrolling broken in some overlay sizes

[Maistho]

When the positioning of an overlay is calculated, the overlay pane is
moved to the top right corner, and given an unbounded size. Based on the
calculated size, a position is attempted where the entire element can
fit next to the origin element without needing to scroll.

If an overlay pane is created with a size large enough that it needs
scrolling in any valid position (top/bottom/left/right) but small enough
that it can fit in the viewport without needing to scroll, the scrolling
of that pane will break. When the pane is moved to the top left corner
for size calculation, the pane will lose it's scroll position since it
no longer has a scrollbar.

This fix introduces a maximum constraint on the size of the overlay pane
during size calculation, where the maximum available vertical and
horizontal space is used as the constraint.

This is a quick approach and there are probably some things I've missed.
I haven't written any tests for this fix, and would like some help regarding
how to best write those tests, as well as fix two tests that broke because
of these changes.

---

Edit: Added a stackblitz with a reproduction

https://stackblitz.com/edit/components-issue-yusmdo?file=src/app/example-component.html

Open the 75vh overlay, scroll a bit in it and then scroll the body. The scroll in the overlay will reset to the top.

Open the 125vh overlay, scroll a bit in it and then scroll the body. The scroll in the overlay will not reset.

[crisbeto]

Is there an example of the issue that this fix is addressing?

[Maistho]

Yes, sorry should've added a stackblitz right away

https://stackblitz.com/edit/components-issue-yusmdo?file=src/app/example-component.html

Open the 75vh overlay, scroll a bit in it and then scroll the body. The scroll in the overlay will reset to the top.

Open the 125vh overlay, scroll a bit in it and then scroll the body. The scroll in the overlay will not reset.

[wagnermaciel]

Looking into this, there are a lot of failures internally that need to be debugged. It seems like landing this PR would require a lot of attention which we can't afford at this time. If anyone still wants to see this issue resolved, please file a bug.

[angular-automatic-lock-bot]

This issue has been automatically locked due to inactivity.
Please file a new issue if you are encountering a similar or related problem.

Read more about our automatic conversation locking policy.

_{This action has been performed automatically by a bot.}