Re: patch: workaround for ooracle segfaults in unicode sbcl

James Bielman wrote:
> I think I've come up with a reasonable workaround for the segfaults
> that occur when using the Oracle backend under SBCL.
> [...]
> I would love to figure out if this is an OCI bug or an SBCL issue, but
> so far I haven't been able to narrow it down to a test case without
> Oracle...

Hi James,

You've certainly been working hard on this difficult bug hunt. I think
your workaround overall is a reasonable one. But, it does have one
significant philosophical issue.

In writing CLSQL, I've been diligent to avoid any lisp implementation
FFI differences. In fact, CLSQL is the main reason I wrote UFFI -- so
CLSQL wouldn't have to know anything about the underlying FFI. I've
been quite strict about that policy, resulting in a concomminant
improvement in UFFI.

So, you patch brings up interesting questions. Long-term, I don't want
to keep such a violation of CLSQL philosophy in the code base. Short
term, I could tolerate it. But, I wonder what the plan would be to fix
the underlying problem and remove the SBCL-specific code from CLSQL.

If it's really a SBCL bug, then perhaps the right place to move the
work-around is UFFI so that other applications wouldn't be affected by
the bug. If it is a bug in Oracle (which seems less likely), then
opening a support ticket with Oracle would make sense.

Your thoughts?

Kevin