These methods set the *active* item, so the verb activate provides a
clearer correspondence. We could change the noun to "shown" item, but
that's awkward and having both active panes and active items is a nice
correspondence in terminology.
We don't actually need structural markup to ensure that all pane views
are absolutely positioned. We can just use the `> *` selector inside of
.pane-items.
/cc @probablycorey is there anything I'm missing here?
Previously, when you always had to specify the event name of 'value'
when calling `::subscribe` with a signal. Now, if you don't specify an
event name, 'value' is assumed.
They're turning out to cause instability in the current version of v8
or atom-shell. Perhaps we can revisit this after the upgrade to chromium
31. Once we have Object.observe, we could at least throw an exception
when someone assigns a sequence index directly via ::[] or assigns to
::length.
The .item-views div needs to be the first child of pane and contain
the absolutely positioned repaint hack div inside it, otherwise the tabs
don't get honored as flexbox items.
- Move emitting of item removal event into ::removeItemAtIndex. Pass a
destroying param instead of setting state.
- Destroy the pane if ::items is empty at the in the item removal
method, rather than with a condition. This gives the item removal event
a chance to fire first.
We were testing the behavior of returning the active pane by removing
the active class from its view. But "active" is a model-level concept
now, so this assertion makes no sense.
The introduction of the .flexbox-redraw-hack required the query to be
loosened, but I went too far. We don't want to return a pane for mini
editors that happen to be nested on another view that's inside a pane.
