Eli Zaretskii wrote:
> Perhaps you are right, and it will "just work".  I will have to try
> (for now, I just looked at the sources and the changes, but didn't try
> building).

It looks to me like it should just work. You might also consider the second 
attached patch: it removes the MS-Windows implementation of mkostemp, since the 
generic code should just work as well. I've tested only the first attached 
patch, though: it is basically the same as before except it's rebased to current 
master.