FYI the string for the location bar is "Search or type URL" which doesn't sound good.

But isn't "Search or type URL" the right actions we'd like to inform the users about? Why doesn't it sound good?

The a11y label for the magnifying glass/Search button itself can simply be "Search". Once the Search button is tapped, focus goes to the omnibox. Omnibox announcement is a little more complex. Here's what YouTube does once the cursor is in the search box:
 
VoiceOver announces "Search field", the state "is editing", and then the hint text in the visual UI, which in our case is "search or type URL", and then other details which are provided by VO, "character mode, insertion point at start"

We could follow the same pattern. I think the only actual label we'd add is "Search"; all else, like the component type "field", is presumably handled by VoiceOver? I'm not a VO expert.

Re: c#2, VoiceOver should automatically read the hint text aloud, so we don't want or need to reiterate the hint text in the a11y label.

What is read when you focus the omnibox by tapping the bottom toolbar's button (with VoiceOver) "Address and search bar, textfield, is editing, <Current URL>".

So if the bottom button just as the "Search" label we would have:
- "Search, button"
- User taps the button
- "Address and..."

I think it is fine and I will implement it. I am just concern about having just "Search" as label for the button, because the user might think it is "Find in Page" (same as having a magnifying glass icon).
WDYT?

Should we make it:

- "Search the web, button"
- User taps the button
- "Address and..."

I went with "search". I followed the same reasoning as the one for the magnifying glass: the user understand it is to search something, and on the first tap, you understand what it does.

SGTM. Thank you.

The NextAction date has arrived: 2018-05-23