Message-Id: <3.0.32.19990816130552.00a39980@ren.cncware.com>
Date: Mon, 16 Aug 1999 13:05:56 -0700
To: Byron Nikolaidis <byronn@solipsys.com>
From: "Ken J. Wright" <ken@ori-ind.com>
Subject: Re: catching posting exceptions
Cc: pgsql-interfaces@postgreSQL.org
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"

At 10:10 07/30/1999 -0400, you wrote:
>
>
>"Ken J. Wright" wrote:
>
>> Attached are the logs pertaining to the posting error. Hopefully they will
>> shed some light. At the end of sql.log there is a SQLDisconnect. This is
>> the point in the program at which the dataset state goes to InActive. I can
>> not as of yet, determine where the SQLDisconnect is originating.
>>
>> Byron, could you please have a look for a possible driver problem?
>>
>
>I don't see anything.  It looks like it is working correctly.  The driver is
>returning an error from sqlexecute.  Shouldn't the application handle that?
>

Byron,

The driver is returning a SQLSTATE of "08S01" (communication link failure)
for any hstmt type error. The error in this case was from attempting to
duplicate a unique key. My application never had the chance to judge the
error, as the api (OCBDExpress) has a hard coded block that terminates the
connection if the SQLSTATE is a "08xxx". This is where my problem stems
from. I believe that both the api & the driver are not correct. The MS ODBC
reference states that the SQLSTATE is not required by the driver and should
not be trusted by an app. This is where I believe ODBCExpress is wrong.
However, 08S01 indicates a rather fatal connection type of error. This is
where I believe the driver is wrong. A failed sql statement does not
constitute a connection failure. The closest error I can find that would
work is HY000 (general error). Comments?

Ken