[PATCH] New condvar implementation that provides stronger ordering guarantees.

[Torvald Riegel]

On Mon, 2016-05-30 at 15:51 +0200, Florian Weimer wrote:
> On 05/27/2016 01:14 PM, Szabolcs Nagy wrote:
> > On 27/05/16 10:57, Florian Weimer wrote:
> >> The removal of the pointer member may have alignment impact elsewhere, too.
> 
> This might still be a problem.

Do we have targets other than m68k in which the alignment of a long long
int is less than the alignment of a pointer?
I had hoped that arch maintainers would notify us about such cases, but
maybe that's too easy to overlook.  Alternatively, we could add another
alignment dummy variable to the outermost union that is of type pointer.