Thread: Issues with the latest 6.5 source

I refreshed my 6.5 source last night (last time was early April). This had
some patches to case and coalesce that I needed.  This is working great
except for the following two areas:

1)    The \g option in pgsql is failing.  I have been using this technique
for months to create a script file from and SQL statement and then execute
the file.  Here is the steps that worked before I pulled down code last
night:
        \g sip_revoke_all.sql        unexpected character Z following 'I'
        Once the above error occurs, the script will hang
until I manually break (ctrl-c).

2)    My Access97 application using ODBC can't connect to the database.
Nothing has changed other than rebuilding the source.  In attempts to fix
this, I and destroyed my databases, rm $PGDATA, and performed initdb. I have
verified that pg_hba.conf is correct before and after recreating my
databases.  The Psqlodbc.log  log file is below.  I can log onto the
database using psql on my linux workstation just fine.  

Has anything changed in the code recently (within the past 3 weeks) that
could have had some effect these two conditions?  I am at a complete stand
still until I can get my ODBC issue resolved.  I can work around item 1.

Thanks Michael

Psqlodbc.log file:

conn=159330904, SQLDriverConnect( in)='DSN=PostgreSQL;', fDriverCompletion=1
DSN info:
DSN='PostgreSQL',server='192.168.97.2',port='5432',dbase='mp',user='postgres
',passwd=''
readonly='0',protocol='6.4',showoid='0',fakeoidindex='0',showsystable='0'         conn_settings=''
translation_dll='',translation_option=''
Global Options: Version='06.40.0004', fetch=100, socket=4096,
unknown_sizes=0, max_varchar_size=254, max_longvarchar_size=4094               disable_optimizer=1, ksqo=1,
unique_index=1,
use_declarefetch=1               text_as_longvarchar=1, unknowns_as_longvarchar=0,
bools_as_char=1               extra_systable_prefixes='dd_;', conn_settings=''
conn=159330904, query=' '
CONN ERROR: func=SQLDriverConnect, desc='Error from CC_Connect', errnum=105,
errmsg='The database does not exist on the server
or user authentication failed.'           ------------------------------------------------------------
henv=159330888,conn=159330904, status=0, num_stmts=16           sock=159337240, stmts=159345496, lobj_type=-999
 ---------------- Socket Info -------------------------------           socket=360, reverse=0, errornumber=0,
errormsg='(null)'          buffer_in=159337288, buffer_out=159341392           buffer_filled_in=2, buffer_filled_out=0,
buffer_read_in=2

Michael J Davis <michael.j.davis@tvguide.com> writes:
> 1)    The \g option in pgsql is failing.

I'll look at that this evening --- I was fooling with psql's command
parsing a couple days ago, might have busted it.

> 2)    My Access97 application using ODBC can't connect to the database.

Dunno about this one.
        regards, tom lane

Michael J Davis <michael.j.davis@tvguide.com> writes:
> 1)    The \g option in pgsql is failing.

Fixed.  I had managed to break the empty-query response protocol on
Sunday --- odd that none of the regression tests detected this.  Sigh.

> 2)    My Access97 application using ODBC can't connect to the database.

I now have a strong suspicion that this is caused by the same goof.
ODBC might issue an empty query during startup.  (That used to be a
necessary part of the startup protocol; it isn't anymore, but ODBC
very possibly hasn't been changed.)

If you don't have CVS access and don't want to wait for tonight's
snapshot, here is the patch:

*** src/backend/tcop/dest.c~    Wed Apr 28 18:15:07 1999
--- src/backend/tcop/dest.c    Wed Apr 28 18:15:45 1999
***************
*** 336,342 ****              *        tell the fe that we saw an empty query string              * ----------------
         */
 
!             pq_putbytes("I", 1);             break;          case Local:
--- 336,342 ----              *        tell the fe that we saw an empty query string              * ----------------
         */
 
!             pq_putbytes("I", 2); /* note we send I and \0 */             break;          case Local:


Let us know if that helps...
        regards, tom lane