<ghaverla@freenet.edmonton.ab.ca> writes:
> Well, I captured a run using strace. 301 lines I believe, I've
> edited it down to 120 or so. It's included below.
Hm. There is nothing apparently wrong in the trace (though it's too
bad we cannot see whether the password from the file is actually sent
in the subsequent connection-request packet ... does your strace have
an option to print more of the data involved in a send call?)
> geteuid32() = 31 (postgres)
> socket(PF_FILE, SOCK_STREAM, 0) = 3
> connect(3, {sa_family=AF_FILE, path="/var/run/.nscd_socket"}, 110)
> = 0
> writev(3, [{"\2\0\0\0\1\0\0\0\3\0\0\0", 12}, {"31\0", 3}], 2) = 15
> read(3, "\2\0\0\0\1\0\0\0\t\0\0\0\2\0\0\0\37\0\0\0
> \0\0\0\t\0\0"..., 36) = 36
> read(3, "postgres\0x\0postgres\0/var/lib/pos"..., 46) = 46
> close(3) = 0
This part of the trace implies that you're using nscd for
authentication, which raises a few red flags for me since I've
seen bugs in nscd before. Can you try turning off nscd and see
if the behavior changes?
regards, tom lane