Thread: Re.:ODBC changes about parameter handling

Re.:ODBC changes about parameter handling

From
"Johann Zuschlag"
Date:
Hi,

I just started testing the driver 7.01.0004 with the patch of Hiroki Kataoka.

Works wonderful so far. I only found some funny behaviour:

1. When leaving my application it crashes with a page fault in one of it's dll's.
    Funny thing: If I switch on CommLog in the driver or tracing in the
    MS driver manager everything is fine. So there are no logs. :-)
    In case of the crash the backend reports:

    pg_recvbuf: recv() failed: Connection reset by peer

    Any ideas?

2. When I switch on cursors (Use Declare/Fetch) and my application does a
    certain query my application crashes with a stack overflow in psqlODBC.
    I only get the address 0x00000FD. The debugger just asked for some
    assembler sources. So no I can't give any details on that.
    Without cursors everything works fine but slow of course.

    I can send some short logs if it helps. Hiroshi?

regards

Johann



Johann Zuschlag
zuschlag2@online.de



Re: Re.:ODBC changes about parameter handling

From
"Johann Zuschlag"
Date:
Hi,

On Sat, 31 Mar 2001 19:15:17 +0200, Johann Zuschlag wrote:

>2. When I switch on cursors (Use Declare/Fetch) and my application does a
>    certain query my application crashes with a stack overflow in psqlODBC.
>    I only get the address 0x00000FD. The debugger just asked for some
>    assembler sources. So no I can't give any details on that.
>    Without cursors everything works fine but slow of course.
>
>    I can send some short logs if it helps. Hiroshi?

I just found out if I increase the Cache Size of the cursor control,
it just works fine.

100 -> 2 tables don't work
200 -> all ok
50 - > most table won't open
20 -> application won't even start :-)

120 seems also fine for me.

Is that behaviour correct? I think the cache size shouldn't be
to large, since network traffic will be increased and the performance
will go down.

regards

Johann




Johann Zuschlag
zuschlag2@online.de



RE: ODBC changes about parameter handling

From
"Hiroki Kataoka"
Date:
Hi Johann,

Send logs to me.  It is good that you send logs to Hiroshi as well.

=====
Hiroki Kataoka

> -----Original Message-----
> From: pgsql-odbc-owner@postgresql.org
> [mailto:pgsql-odbc-owner@postgresql.org]On Behalf Of Johann Zuschlag
> Sent: Sunday, April 01, 2001 2:15 AM
> To: pgsql-odbc@postgresql.org
> Subject: Re.:[ODBC] ODBC changes about parameter handling
>
>
> Hi,
>
> I just started testing the driver 7.01.0004 with the patch of
> Hiroki Kataoka.
>
> Works wonderful so far. I only found some funny behaviour:
>
> 1. When leaving my application it crashes with a page fault in
> one of it's dll's.
>     Funny thing: If I switch on CommLog in the driver or tracing in the
>     MS driver manager everything is fine. So there are no logs. :-)
>     In case of the crash the backend reports:
>
>     pg_recvbuf: recv() failed: Connection reset by peer
>
>     Any ideas?
>
> 2. When I switch on cursors (Use Declare/Fetch) and my application does a
>     certain query my application crashes with a stack overflow in
> psqlODBC.
>     I only get the address 0x00000FD. The debugger just asked for some
>     assembler sources. So no I can't give any details on that.
>     Without cursors everything works fine but slow of course.
>
>     I can send some short logs if it helps. Hiroshi?
>
> regards
>
> Johann
>
>
>
> Johann Zuschlag
> zuschlag2@online.de


Attachment

RE: Re.:ODBC changes about parameter handling

From
"Hiroshi Inoue"
Date:
> -----Original Message-----
> From: Johann Zuschlag
>
> Hi,
>
> On Sat, 31 Mar 2001 19:15:17 +0200, Johann Zuschlag wrote:
>
> >2. When I switch on cursors (Use Declare/Fetch) and my
> application does a
> >    certain query my application crashes with a stack overflow
> in psqlODBC.
> >    I only get the address 0x00000FD. The debugger just asked for some
> >    assembler sources. So no I can't give any details on that.
> >    Without cursors everything works fine but slow of course.
> >
> >    I can send some short logs if it helps. Hiroshi?
>
> I just found out if I increase the Cache Size of the cursor control,
> it just works fine.
>
> 100 -> 2 tables don't work
> 200 -> all ok
> 50 - > most table won't open
> 20 -> application won't even start :-)
>
> 120 seems also fine for me.
>

I couldn't reproduce it here.
What's your environment ?

regards,
Hiroshi Inoue

RE: Re.:ODBC changes about parameter handling

From
"Johann Zuschlag"
Date:
On Sun, 1 Apr 2001 23:45:28 +0900, Hiroshi Inoue wrote:

>
>I couldn't reproduce it here.
>What's your environment ?
>
I knew that one was going to be a difficult one.
I use:

Win 98 SE
MS driver manager 3.510.3711.0

Maybe my application is the reason. But it's
just sending simple SELECT's.
Table size is between 100 and 20.000.
Funny thing it happened with the small table's (124 rows)

regards,



Johann Zuschlag
zuschlag2@online.de



Re: Re.:ODBC changes about parameter handling

From
Hiroshi Inoue
Date:
Johann Zuschlag wrote:
>
> On Sun, 1 Apr 2001 23:45:28 +0900, Hiroshi Inoue wrote:
>
> >
> >I couldn't reproduce it here.
> >What's your environment ?
> >
> I knew that one was going to be a difficult one.
> I use:
>
> Win 98 SE
> MS driver manager 3.510.3711.0
>
> Maybe my application is the reason. But it's
> just sending simple SELECT's.
> Table size is between 100 and 20.000.
> Funny thing it happened with the small table's (124 rows)
>

Could you increase the stack size of your application ?

regards,
Hiroshi Inoue

Re: Re.:ODBC changes about parameter handling

From
"Johann Zuschlag"
Date:
On Mon, 02 Apr 2001 11:12:51 +0900, Hiroshi Inoue wrote:

>> Maybe my application is the reason. But it's
>> just sending simple SELECT's.
>> Table size is between 100 and 20.000.
>> Funny thing it happened with the small table's (124 rows)
>>
>
>Could you increase the stack size of your application ?
>

I'm sorry,  but I don't have the sources.
I will try something else. I just try different
cursor cache sizes. Maybe there is some kind
of relationship. Maybe I get the chance
to trace something.

Did you have any idea about my first point:
Page fault when leaving.
But not if I switch on CommLog?

regards




Johann Zuschlag
zuschlag2@online.de