Bugzilla

This is a reduced test case from linkedin

`data:text/html,<span id="visible">hello world</span><span id="invisible" style="display:block;overflow:hidden; clip: rect(0 0 0 0);height: 1px; width:1px;">hello world</span>`

On the actual site, the `visible` span has `aria-hidden=true` and the visible and invisible spans are wrapped in a span container, which gets forced into the tree as a generic.
I haven't figured out why that container is forced, since it doesn't have overflow, position, or transform styling that'd make us create an acc for it, but in any case the bigger problem here seems to be the 1x1 bounds we create for the invisible div, which renders it navigable and breaks hittesting

This is a reduced test case from linkedin

`data:text/html,<div id="container"><span id="visible" aria-hidden="true">hello world</span><span id="invisible" style="display:block;overflow:hidden; clip: rect(0 0 0 0);height: 1px; width:1px;">hello world</span></div>`

STR:
1. Hittest the container

Expected:
we get the `invisible` span's child text

Actual:
we get the container itself

This is a reduced test case from linkedin

`data:text/html,<div id="container"><span id="visible" aria-hidden="true">hello world</span><span id="invisible" style="display:block;overflow:hidden; clip: rect(0 0 0 0);height: 1px; width:1px;">hello world</span></div>`

STR:
1. Hittest the container

Expected:
we get the `invisible` span's child text

Actual:
we get null

This is a reduced test case from linkedin

`data:text/html,<div id="container"><span id="visible" aria-hidden="true">hello world</span><span id="invisible" style="display:block;overflow:hidden; clip: rect(0 0 0 0);height: 1px; width:1px;position:absolute;">hello world</span></div>`

STR:
1. Hittest the container

Expected:
we get the `invisible` span's child text

Actual:
we get null