Re: [Patch] Oracle on Mac OS X with SBCL and OpenMCL

James Bielman wrote:
> I'm not certain if I ran the unit tests correctly; I got 6 failures
> for Oracle on SBCL/x86 before and after applying this patch.  Oracle
> under OpenMCL fails 12/213 tests (including the same 6 as SBCL).  I've
> attached the OpenMCL test output.

Thanks for the report and the patch.

Those unit test failures represent the failure to update a global
variable whose name has changed to *default-string-length*. I fixed it
a few days ago. The next upload of CLSQL will have a fix.


> +;; I've changed the data type of COLNAME to :unsigned-short as per
> +;; the Oracle 9i OCI documentation. -- JJB 20040713

I'll double check this and commit this patch.

> +      ;; Using (SETF DEREF-VP) to initialize this pointer fails in OpenMCL
> +      ;; due to incorrect evaluation of the :POINTER-VOID argument to
> +      ;; UFFI:DEREF-POINTER:
> +      ;;
> +      ;; > Error in process listener(1): Unknown foreign type: :g10
> +      ;; > While executing: ccl::%parse-foreign-type
> +      ;;
> +      ;; The following works around this for OpenMCL as I'm not
> +      ;; certain where the actual problem is. -- JJB 20040713

I wasn't aware the Oracle ran on OSX. I'll install a client for my OSX
system and work on debugging UFFI so that OpenMCL works.

> +      #+openmcl
> +      (setf (uffi:deref-pointer envhp :pointer-void) +null-void-pointer+)
> +      #-openmcl
>        (setf (deref-vp envhp) +null-void-pointer+)

That's an okay work-around. But, in general, I've always been able to
adjust UFFI so that such implementation-specific differences in
UFFI-using applications are not necessary.

Thanks again for the report and patch -- very appreciated.

Kevin