Well playing with it a bit tonight
I tried this patch instead
++ AV *tuples_av =sv_2mortal(newAV());
-- tuples_utf8_av=newAV();
and it worked exactly the same.
I have noticed a little thing though that may be another leak.
Seems If I change the value of 'ora_array_chunk_size' It becomes stable but
then seems to leak again later on.
I will have to investigate a little more. Might just have to juggle things
about a little to clear out memory when dealing with big
''ora_array_chunk_size'
Cheers
John
> Date: Tue, 9 Oct 2012 16:57:05 +0200
> Subject: Re: Patch proposal for leak in DBD::Oracle when calling
> 'execute_array' with UTF-8 NLS...
> From: demer...@gmail.com
> To: byter...@hotmail.com
> CC: martin.ev...@easysoft.com; blan...@worldcom.ch; dbi-dev@perl.org
>
> On 9 October 2012 14:29, John Scoles <byter...@hotmail.com> wrote:
> >
> > I am sort of with Martin on this one as we would like to know why?
> >
> > At first glace decrementing the reference count like this would not seem to
> > be a little overkill as tuples_utf8_av is created with newAV() which should
> > take care of that for us.
>
> Should take care of what?
>
> newAV() doesnt create a mortal.
>
> Yves
>
>
> --
> perl -Mre=debug -e "/just|another|perl|hacker/"
