Weird intermittent ECPG test failures - Mailing list pgsql-hackers

Observe the following buildfarm failures:

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=conchuela&dt=2019-07-03%2013%3A33%3A59
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dory&dt=2019-04-30%2014%3A45%3A26
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dory&dt=2019-06-02%2015%3A15%3A26
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dory&dt=2019-06-05%2006%3A15%3A26
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dory&dt=2019-06-26%2002%3A00%3A26
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dory&dt=2019-07-03%2022%3A15%3A27
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hamerkop&dt=2019-04-10%2011%3A00%3A09
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hamerkop&dt=2019-04-14%2012%3A42%3A31
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hamerkop&dt=2019-05-01%2011%3A00%3A08
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hamerkop&dt=2019-06-29%2011%3A33%3A25

(These represent all but a couple of the ECPG-Check failures in the last
90 days, excluding a spate around 31-May that were caused by an
ill-considered code change.)

In every one of these cases, the visible symptom is that one of the ECPG
test programs produced precisely nothing on stdout.  stderr seems okay
though.  (Not all of these cases have nonempty expected stderr, but enough
do that we can say it's not a case of the program just failing entirely.)

We've been seeing this sort of thing for a *long* time, although my
recollection is that in the past it's almost always been the thread-thread
case that failed, so that I'd supposed that there was some problem with
that particular test.  It's now clear that that's not true though, and
there's seemingly some generic issue with the tests or test
infrastructure.

The other thing that I thought was invariably true was that it was a
Windows-only thing.  But here we have conchuela failing in the exact
same way, and it is, um, not Windows.

Anyone have a theory what might be going on here?

            regards, tom lane