Thread: CVS Head: Pg_ctl bug?
I just noticed that if I don't specify a data directory either on the command line or in $PGDATA,, pg_ctl fails silently on Windows (haven't been able to test other platforms atm). Is this a bug or an oversight - an error message would seem appropriate? /D postgres@PC30:/usr/local/pgsql$ bin/pg_ctl start postgres@PC30:/usr/local/pgsql$ bin/pg_ctl -D data start postmaster starting postgres@PC30:/usr/local/pgsql$ LOG: database system was interrupted at 2005-08-29 00:34:07 GMT Standard Time LOG: checkpoint record is at 0/39F218 LOG: redo record is at 0/39F218; undo record is at 0/0; shutdown TRUE LOG: next transaction ID: 1478; next OID: 16392 LOG: next MultiXactId: 1; next MultiXactOffset: 0 LOG: database system was not properly shut down; automatic recovery in progress LOG: record with zero length at 0/39F260 LOG: redo is not required LOG: database system is ready LOG: transaction ID wrap limit is 2147484144, limited by database "postgres"
"Dave Page" <dpage@vale-housing.co.uk> writes:
> I just noticed that if I don't specify a data directory either on the
> command line or in $PGDATA,, pg_ctl fails silently on Windows (haven't
> been able to test other platforms atm). Is this a bug or an oversight -
> an error message would seem appropriate?
On my machine it does this:
$ bin/pg_ctl start
pg_ctl: no database directory specified and environment variable PGDATA unset
Try "pg_ctl --help" for more information.
$
Is write_stderr() broken on Windows?
regards, tom lane
> > I just noticed that if I don't specify a data directory=20 > either on the=20 > > command line or in $PGDATA,, pg_ctl fails silently on=20 > Windows (haven't=20 > > been able to test other platforms atm). Is this a bug or an=20 > oversight=20 > > - an error message would seem appropriate? >=20 > On my machine it does this: >=20 > $ bin/pg_ctl start > pg_ctl: no database directory specified and environment=20 > variable PGDATA unset Try "pg_ctl --help" for more information. > $ >=20 > Is write_stderr() broken on Windows? In some cases it writes to the eventlog instead of stderr. Dave, can you check if this is what's happening to you? //Magnus
=20 > -----Original Message----- > From: Magnus Hagander [mailto:mha@sollentuna.net]=20 > Sent: 29 August 2005 20:58 > To: Tom Lane; Dave Page > Cc: pgsql-bugs@postgresql.org > Subject: RE: [BUGS] CVS Head: Pg_ctl bug?=20 >=20 > > Is write_stderr() broken on Windows? >=20 > In some cases it writes to the eventlog instead of stderr.=20 > Dave, can you > check if this is what's happening to you? Bah, no coffee for 2 weeks dulls the mind :-( Yes, this is what's happening. Odd though, considering that I have: #log_destination =3D 'stderr' In my out-of-the-box config. Regards, Dave.
"Dave Page" <dpage@vale-housing.co.uk> writes:
>> In some cases it writes to the eventlog instead of stderr.
> Yes, this is what's happening. Odd though, considering that I have:
> #log_destination = 'stderr'
> In my out-of-the-box config.
pg_ctl doesn't read the config file, though. Probably we need to
make write_stderr aware that it's running in an interactive program
rather than in the postmaster, and never use the eventlog in this context.
regards, tom lane
> > > Is write_stderr() broken on Windows?
> >=20
> > In some cases it writes to the eventlog instead of stderr.=20
> > Dave, can you
> > check if this is what's happening to you?
>=20
> Bah, no coffee for 2 weeks dulls the mind :-(
>=20
> Yes, this is what's happening. Odd though, considering that I have:
>=20
> #log_destination =3D 'stderr'
>=20
> In my out-of-the-box config.
This is output from pg_ctl, right? pg_ctl doesn't look at
postgresql.conf...
pg_ctl has:
if (!isatty(fileno(stderr))) /* Running as a service */
... write to eventlog ...
That check appears to be failing. The backend has a lot more elaborate
check (in port/backend/security.c). Perhaps we need to make that kind of
advanced check in pg_ctl as well?
Though I don't see any of this code changing lately, so I don't see why
it's failing now. Are you running it under some kind of different
environment than you used to? Any other ideas on why it would claim your
console is not a tty?
//Magnus
> >> In some cases it writes to the eventlog instead of stderr.=20 >=20 > > Yes, this is what's happening. Odd though, considering that I have: > > #log_destination =3D 'stderr' > > In my out-of-the-box config. >=20 > pg_ctl doesn't read the config file, though. Probably we=20 > need to make write_stderr aware that it's running in an=20 > interactive program rather than in the postmaster, and never=20 > use the eventlog in this context. It must sitll use the eventlog when called as a service control process, otherwise any error output from pg_ctl in that scenario would go to /dev/null...=20 There is code that attempts to check this, but it's apparantly failing. //Magnus
=20 > -----Original Message----- > From: Tom Lane [mailto:tgl@sss.pgh.pa.us]=20 > Sent: 29 August 2005 21:12 > To: Dave Page > Cc: Magnus Hagander; pgsql-bugs@postgresql.org > Subject: Re: [BUGS] CVS Head: Pg_ctl bug?=20 >=20 > "Dave Page" <dpage@vale-housing.co.uk> writes: > >> In some cases it writes to the eventlog instead of stderr.=20 >=20 > > Yes, this is what's happening. Odd though, considering that I have: > > #log_destination =3D 'stderr' > > In my out-of-the-box config. >=20 > pg_ctl doesn't read the config file, though. Probably we need to > make write_stderr aware that it's running in an interactive program > rather than in the postmaster, and never use the eventlog in=20 > this context. Which makes sense given that the entire problem is that it doesn't know where to find the config file! /D
"Magnus Hagander" <mha@sollentuna.net> writes:
> Though I don't see any of this code changing lately, so I don't see why
> it's failing now. Are you running it under some kind of different
> environment than you used to?
Well, are we sure that it was working for Dave before? Maybe it's a
long-standing problem. If you didn't happen to mistype a command you'd
not know it wasn't working.
regards, tom lane
=20 > -----Original Message----- > From: Magnus Hagander [mailto:mha@sollentuna.net]=20 > Sent: 29 August 2005 21:13 > To: Dave Page; Tom Lane > Cc: pgsql-bugs@postgresql.org > Subject: RE: [BUGS] CVS Head: Pg_ctl bug?=20 >=20 > Though I don't see any of this code changing lately, so I=20 > don't see why > it's failing now. Are you running it under some kind of different > environment than you used to? Any other ideas on why it would=20 > claim your > console is not a tty? The only change is a new install of msys a month or so back, so nothing major ;-). I can't think of anything that /should/ have changed that behaviour. Regards, Dave
=20 > -----Original Message----- > From: Tom Lane [mailto:tgl@sss.pgh.pa.us]=20 > Sent: 29 August 2005 21:32 > To: Magnus Hagander > Cc: Dave Page; pgsql-bugs@postgresql.org > Subject: Re: [BUGS] CVS Head: Pg_ctl bug?=20 >=20 > "Magnus Hagander" <mha@sollentuna.net> writes: > > Though I don't see any of this code changing lately, so I=20 > don't see why > > it's failing now. Are you running it under some kind of different > > environment than you used to? >=20 > Well, are we sure that it was working for Dave before? Maybe it's a > long-standing problem. If you didn't happen to mistype a=20 > command you'd > not know it wasn't working. I couldn't say whether it was or wasn't working before. On further investigation though, it does seem to work from a Windows command prompt, just not an msys console. Regards, Dave.
Dave Page wrote: > > > > -----Original Message----- > > From: Tom Lane [mailto:tgl@sss.pgh.pa.us] > > Sent: 29 August 2005 21:32 > > To: Magnus Hagander > > Cc: Dave Page; pgsql-bugs@postgresql.org > > Subject: Re: [BUGS] CVS Head: Pg_ctl bug? > > > > "Magnus Hagander" <mha@sollentuna.net> writes: > > > Though I don't see any of this code changing lately, so I > > don't see why > > > it's failing now. Are you running it under some kind of different > > > environment than you used to? > > > > Well, are we sure that it was working for Dave before? Maybe it's a > > long-standing problem. If you didn't happen to mistype a > > command you'd > > not know it wasn't working. > > I couldn't say whether it was or wasn't working before. On further > investigation though, it does seem to work from a Windows command > prompt, just not an msys console. I assume we are OK on this item. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
=20 > -----Original Message----- > From: Bruce Momjian [mailto:pgman@candle.pha.pa.us]=20 > Sent: 23 September 2005 20:56 > To: Dave Page > Cc: Tom Lane; Magnus Hagander; pgsql-bugs@postgresql.org > Subject: Re: [BUGS] CVS Head: Pg_ctl bug? >=20 > Dave Page wrote: > >=20=20 > >=20 > > > -----Original Message----- > > > From: Tom Lane [mailto:tgl@sss.pgh.pa.us]=20 > > > Sent: 29 August 2005 21:32 > > > To: Magnus Hagander > > > Cc: Dave Page; pgsql-bugs@postgresql.org > > > Subject: Re: [BUGS] CVS Head: Pg_ctl bug?=20 > > >=20 > > > "Magnus Hagander" <mha@sollentuna.net> writes: > > > > Though I don't see any of this code changing lately, so I=20 > > > don't see why > > > > it's failing now. Are you running it under some kind of=20 > different > > > > environment than you used to? > > >=20 > > > Well, are we sure that it was working for Dave before?=20=20 > Maybe it's a > > > long-standing problem. If you didn't happen to mistype a=20 > > > command you'd > > > not know it wasn't working. > >=20 > > I couldn't say whether it was or wasn't working before. On further > > investigation though, it does seem to work from a Windows command > > prompt, just not an msys console. >=20 > I assume we are OK on this item. Yeah, I think so. Regards, Dave.