Re: Post-mortem: final 2PC patch

Stefan Kaltenbrunner <stefan(at)kaltenbrunner(dot)cc> writes:
> anyway - here is the promised backtrace:

> #0 0x000000004489fba4 in memcpy () from /usr/lib/libc.so.34.2
> #1 0x0000000000326f9c in hash_search (hashp=0xa6e030, keyPtr=0xa5ff90,
> action=HASH_ENTER, foundPtr=0x0) at dynahash.c:653
> #2 0x00000000003434f0 in pg_tzset (name=0xa5ff90 "PST8PDT") at pgtz.c:1039
> #3 0x00000000001fbcf0 in assign_timezone (value=0xa5ff90 "PST8PDT",
> doit=1 '\001', source=PGC_S_CLIENT) at variable.c:351
> #4 0x0000000000330f28 in set_config_option (name=0xa52830 "timezone",
> value=0xa52870 "PST8PDT", context=10645504, source=PGC_S_CLIENT,
> isLocal=0 '\0',
> changeVal=1 '\001') at guc.c:3748
> #5 0x000000000029b5f0 in PostgresMain (argc=4, argv=0xa52928,
> username=0xa52740 "mastermind") at postgres.c:2759

Yeah, that's exactly what I thought it was doing: setting timezone
PST8PDT due to the client PGTZ environment variable. So the question
now is why this is an intermittent failure. Why doesn't every single
regression test crash in the same place?

regards, tom lane