With the following sample code (taken from the doc:
<div>
<RaisedButton onTouchTap={this.handleTouchTap} label="Click me" />
<Popover
open={this.state.open}
anchorEl={this.state.anchorEl}
anchorOrigin={{horizontal: 'left', vertical: 'bottom'}}
targetOrigin={{horizontal: 'left', vertical: 'top'}}
onRequestClose={this.handleRequestClose}
animation={PopoverAnimationVertical}
>
<Menu>
<MenuItem primaryText="Refresh" />
<MenuItem primaryText="Help & feedback" />
<MenuItem primaryText="Settings" />
<MenuItem primaryText="Sign out" />
</Menu>
</Popover>
</div>
When the popover is closed, it renders as an empty div, which breaks the layout when I try to align more than one button with popover.
Affects <DropDownMenu>, <SelectField>, and <Popover>.

I think the regression was introduced by #5444, reverting #3806, which was a cure for the same problem (o_O).
Any opinion on this one? I'd be happy to open a PR reverting the previous revert (!) to go back to the original version since that solves the issue.
As a side note, this blocks a new release of marmelab/admin-on-rest (offering React 15.4 compatibility).
Sorry for asking again, but I'm concerned by this regression and the lack of response from the maintainers. I'm sure you guys are very busy, but this is a possibly blocking regression for many projects, with a one line fix.
+1 experiencing this as well......
We worked around this bug by changing the <div> at the beginning of this issue into <span className='LastChildDisplayNone'> and adding the following CSS.
.LastChildDisplayNone > div:last-child {
display: none;
}
Though you'd have to watch your Code once this bug is officially solved, otherwise your last child disappears from your UI.
@fzaninotto @kkomaz Sorry for the slow response guys.
A lot of effort right now is being put into the next major version which has much better test and regression test coverage -- concerns in this area (regressions, consistency, etc) are major reasons for the upcoming major version.
That being said, this is definitely important for the current version. If you submit a PR, i'll 馃憤 it so it goes in the next patch release.
My workaround is a hack that should not go into a PR.
Sorry for asking again, but I'm concerned by this regression and the lack of response from the maintainers.
Sorry for the late answer, I don't receive any notification from issues. The signal vs noise is too high for what I can handle. I focus on PRs. Feel free to ping me directly next time 馃憤 .
Any opinion on this one? I'd be happy to open a PR reverting the previous revert (!) to go back to the original version since that solves the issue.
I can't remember the motivation behind that fix. Please, feel free to submit a PR adding that style back 馃憤 .
Most helpful comment
I think the regression was introduced by #5444, reverting #3806, which was a cure for the same problem (o_O).