Peter Eisentraut <peter_e(at)gmx(dot)net> writes:
> I tried that and various combinations of switches and environments but
> could not reproduce this.
I find it striking that both complainants reported that the issue went
away on a fresh build attempt. This seems to suggest that even on
configurations that can produce the problem, it only happens one time
in N, where we have little idea how large N is.
Marcel, could you try just doing "make distclean, configure, make"
ten or a hundred times overnight, and see if you get more failures?
regards, tom lane