Re: Segfault on VARRAY fetch

On Thu, Sep 06, 2001 at 03:13:31PM -0500, John Gorkos wrote:
> Help-
>   I am trying to read spatial data from a standard SDO structure, as shown in the readgeom.c example in 
> $ORACLE_HOME/md/demo/examples (8.1.7)
> Using Oracle::OCI, I am able to connect and prepare the query, after describing the objects that 
> I want to pull back.  However, when I execute the query, I get segfaults under both Solaris (with coredump) and Linux.  
> Here are some code snippets that are relevant (I'm assuming you're looking at readgeom.c 
> for the missing pieces).  Commentary follows the code, before the DBI->trace dump.
> 
> ...
> 
> According to readgeom.c, I need ARRAY_SIZE (which is 32) as my 'iters' parameter in the OCIStmtExecute.  
> When I set this to 0 (zero), the code executes flawlessly, but OCI_ATTR_ROW_COUNT also returns zero
> (but the column count always returns 2, which is correct for the query).
> I suspect strongly that the error is in the OCIDefineObject statement.  According to the OCI reference book, I need to be
> allocating memory space for the statement to return the values into, but I just don't know how to do it.  The keys are the 4th-7th
> params of OCIDefineObject.  I'm NOT a C programmer, meaning many of the arcane uses of pointers to pointers and the concept
> of pointers pointing to arrays of pointers REALLY fogs my brain, and I don't know how to translate this to Perl.  I suspect that if
> I managed to get the right structures, it would all work.
> Any thoughts?  Please?

Wow! I admire your bravery! You've charged in where angels might fear to tread.

When I get a moment I'll take a look at readgeom.c. Chase me if you've not heard anything
within a couple of days.


Tim.