Re: Solaris 2.5, Netscape Server 2.0, and Oracle 7.3.2

elf_at_coho.halcyon.com (Elf Sternberg) wrote:
> I'm running magnus 2.0, the core Netscape Server product
>they're currently hawking over at Netscape, and I'm trying to write an
>interface between the Oracle API and the Netscape Server API. The 2.0
>Server has full multithreading capacity as given by Solaris's confor-
>mance with POSIX threads. If I put the Oracle log-in, exec, and log-
>out all in the same routine, it works fine. If I put the Oracle log-
>in in an intialization routine, and then have sub-threads making the
>EXEC SQL calls, the first call gets
>
> 03113, "end-of-file on communication channel"
>
>and each subsequent call creates
>
> 03114, "not connected to ORACLE"
>
> I suspect that, because of the multithreading capacities of
>magnus, the database is getting a weird LDA (Login Data Area) from some
>subthread instantiated after initialization. Anyone have any ideas?
>I'm not doing the Oracle from scratch, if that's any difference--
>I'm running the program through ProC first, then compiling with GCC.
>Anyone got any hints?

I must be behind the times. Are all the Oracle libraries MT-safe all of a sudden? For that matter, is your gcc linking against MT-safe system libraries?

On the face of it, it sounds like synchronization primitives aren't being used to control access to ORACLE per-process resources, e.g. file descriptors.

Is this using SQL*NET v2, BTW?

--Karl

-- 
        == Karl Wiebe == karl_at_dnai.com ==         
"Order is a form of repetition compulsion" --Freud
"Order is a form of repetition compulsion" --Freud
"Order is a form of repetition compulsion" --Freud