Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> On 2021-Oct-13, Andres Freund wrote:
>> Yea, let's go for your patch then. I've verified that at least locally it
>> passes under valgrind.
> Ah great, thanks. Pushed then.
Seems like this hasn't fixed the problem: skink still fails on
this test occasionally.
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=skink&dt=2021-10-22%2013%3A52%3A00https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=skink&dt=2021-11-02%2000%3A28%3A30
Both of these look like
# Failed test '000000010000000000000002 differs from 000000010000000000000002'
# at t/026_overwrite_contrecord.pl line 61.
# Looks like you failed 1 test of 3.
t/026_overwrite_contrecord.pl ........
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/3 subtests
which looks like the same thing we were seeing before.
010e52337 seems to have just decreased the probability of failure.
regards, tom lane