Re: 8rc5 on OpenBSD

Hi,

I try a "make check" without success.

The compilation is done correctly.
But "make check" throws:
> ============== creating temporary installation ==============
> ============== initializing database system ==============
> ============== starting postmaster ==============
> running on port 65432 with pid 17114
> ============== creating database "regression" ==============
> createdb: could not connect to database template1: could not connect
> to server: No such file or directory
> Is the server running locally and accepting
> connections on Unix domain socket "/tmp/.s.PGSQL.65432"?
> pg_regress: createdb failed
> gmake[2]: *** [check] Error 2
> rm regress.o
> gmake[2]: Leaving directory
> `/admin/install/pg/postgresql-8.0.0rc5/src/test/regress'
> gmake[1]: *** [check] Error 2
> gmake[1]: Leaving directory
> `/admin/install/pg/postgresql-8.0.0rc5/src/test'
> gmake: *** [check] Error 2
> *** Error code 2
>
> Stop in /admin/install/pg/postgresql-8.0.0rc5 (line 19 of Makefile).
> admin has logged on ttyp1 from imac.

# cat src/test/regress/log/postmaster.log
> WARNING: could not create Unix-domain socket
> LOG: database system was shut down at 2005-01-12 14:25:10 CET
> LOG: checkpoint record is at 0/A3E614
> LOG: redo record is at 0/A3E614; undo record is at 0/0; shutdown TRUE
> LOG: next transaction ID: 544; next OID: 17230
> LOG: database system is ready
> LOG: received fast shutdown request
> LOG: shutting down
> LOG: database system is shut down

I recompile with the option for configure from the 7.4 ports. Nothing
better.

Any help welcomed.

Cordialement,
Jean-Gérard Pailloncy

Pailloncy Jean-Gerard <jg(at)rilk(dot)com> writes:
> # cat src/test/regress/log/postmaster.log
>> WARNING: could not create Unix-domain socket

Hmm. That's pretty odd --- the warning indicates that
StreamServerPort() failed to open a local socket, but all of the
expected failure paths will log an additional message saying why
it couldn't open the socket.

The only code path I see offhand in which no message would be logged
is if the ListenSocket[] array is already full. Is it possible that
you have 10 or more IP addresses that "localhost" would bind to?

regards, tom lane

Pailloncy Jean-Gerard <jg(at)rilk(dot)com> writes:
>>>> The only code path I see offhand in which no message would be logged
>>>> is if the ListenSocket[] array is already full. Is it possible that
>>>> you have 10 or more IP addresses that "localhost" would bind to?
>>
>>> 2 IPs + 3 alias on 2 interfaces
>>> 1 IPs + 27 alias on 1 loopback interface
>>
>> Wow. Try increasing the MAXLISTEN constant in
>> src/backend/postmaster/postmaster.c.

> Bump it to 100. Good.
> I sugess a better error message.

Done. It never occurred to anyone that that limit would actually be
reached, I suppose.

>> =======================
>> 8 of 96 tests failed.
>> =======================

Seems like you've still got some issues though. What do the detailed
diffs look like?

regards, tom lane

Pailloncy Jean-Gerard <jg(at)rilk(dot)com> writes:
> =======================
> 8 of 96 tests failed.
> =======================

The failures all seem to be due to individual tests not getting run:

> ! psql: could not fork new process for connection: Resource temporarily unavailable
> ! psql: could not fork new process for connection: Resource temporarily unavailable
> ! psql: could not fork new process for connection: Resource temporarily unavailable
> ! psql: could not fork new process for connection: Resource temporarily unavailable
> ! psql: could not send startup packet: Broken pipe
> ! psql: could not send startup packet: Broken pipe

(the other two failures are because tables these tests should have
created didn't get created)

The fork failures are easy enough to explain: you have the per-user
process limit set too low. I suspect the "broken pipe" failures
have the same root cause, but they do seem a bit odd.

regards, tom lane