[postgis-devel] Bogus PROJ4 memory context definition crashes Postgres

[Mark Cave-Ayland]

On Thu, 2007-07-12 at 19:40 -0400, Tom Lane wrote:

> Did you check the pre-8.0 cases?  My recollection is that the stats
> method goes all the way back while is_empty doesn't, so both your
> 7.2 and 7.3 structs look fishy.

Ah - looks like I missed the change in method ordering with the older PostgreSQL 
versions :(  Fix committed to SVN.

> BTW, I recall seeing something yesterday that made me think Postgis was
> setting up more than one kind of specialized memory context.  I didn't
> go back to look more thoroughly, but you should check if there are
> others and if so whether they are up to speed on these points.

I'm sure there is only one type of memory context. When the transform()
function is first called, the PROJ4 context is attached as a child
context of the function context so we can free the library handles as
the transform() context is destroyed.


Kind regards,

Mark.

-- 
ILande - Open Source Consultancy
http://www.ilande.co.uk