Re: Does psqlODBC actually work on osx? - Mailing list pgsql-odbc

I've tracked this down to the getDSNInfo function inside dlg_specific.c
It checks only INI_SERVER and not SPEC_SERVER - while other parts of the
code check both, this leads to inconsistent behaviour.

If I use e.g. iodbctestw
"DSN=test;UID=postgres;PWD=postgres13;SERVERNAME=10.0.0.3" it bypasses
the ini reading and works as expected.

I attempted a quick patch to make getDSNInfo check both and after doing
that it works as expected, personally I think this should be fixed to
check for both.

Thanks for all the help.
- Malcolm MacLeod

> > >>> We have a client trying to connect to PostgreSQL server 9.2 from an osx
> > >>> client with our software via ODBC, he has asked for instructions to
> > >>> assist him setting up.
> > >>>
> > >>> I have attempted the setup myself using psqlODBC and no matter what I do
> > >>> configuration wise, the driver fails to connect (via iodbctestw and
> > >>> iodbc administrator) stating that the password is incorrect - I know
> > >>> this is not the case because I am using identical configuration to my
> > >>> linux machine where it works fine.
> > >> What is the exact error message you are getting?
> > >> Are you connecting from within the same network as your Linux machine?
> > >> Just trying to eliminate the possibility that it is a pg_hba.conf issue.
> > > All on same internal network.
> > > Server 10.0.0.3, working machine(s) 10.0.0.24, 10.0.0.25 etc. broken
> > > machine 10.0.0.26
> > > I've tried also setting the pg_hba.conf to 'trust' and even then it
> > > doesn't seem to work.
> > >
> > > Various config info and traces below.
> > >
> > >
> > >
> > > Snippet from configuration (Although I've played with various other
> > > options SSLmode etc. here as well)
> > >
> > > [test]
> > > Driver=psqlODBC
> > > Server=10.0.0.3
> >
> > The driver seems to ignore the above entry.
> > Could you please try Servername instead of Server?
> Hrm, you are right using ServerName seems to fix the issue, it is a
> relief that it is something so simple in the end. Thanks!
>
> However it is not as clear cut as "Server" being ignored, as mentioned
> in the other messages it does actually attempt to connect to the right
> server, postmaster on the server even acknowledges the failed connection
> attempt. If I turn ssl support on the server off and on in the
> connection string than I get a message that the server refuses SSL etc.
> etc.
>
> Further if I check the code "CopyAttributes" checks for both INI_SERVER
> (ServerName) and SPEC_SERVER (Server) and fills ci->server in
> identically regardless.
> So rather than a case of "Server" being ignored it seems instead that it
> is not ignored, but that somehow some part of the auth code is doing
> something wrong; scanning manually for servername again instead of using
> ci->server or similar?; I don't know the driver code well enough to know
> where to look but as I've already set everything up for
> compiling/testing if someone can point me to some key places this might
> happen I'd be happy to test quickly.
>
> - Malcolm MacLeod
>
>
>
>