Jaime Casanova <jaime(at)2ndquadrant(dot)com> writes:
> I was running some tests, and noted $SUBJECT.
> It fails consistently on current HEAD (without any patch).
> ...
> Attached regressions.diff, basically the fails give an error of
> "select failed: Interrupted system call"
Hm, looks to me like the select() error path at isolationtester.c:717
ought to consider EINTR as a retryable case. Curious though that we
have not seen this in the buildfarm ... unless maybe this explains the
failures-with-no-output that have been seen on some of the Windows
critters? But you'd think we'd see the "select failed:" message if so.
regards, tom lane