[postgis-devel] BUGFIX in transform()

[strk]

Yes, another bug found in transform().
Thanks to Mark I've discovered the --enable-cassert configure
flag of postgresql, which enables many more checks about memory
management.
I suggest everyone willing to help compile postgresql with
that flag enabled.

Looking at the deserializer functions I've also incurred
in other places where unsigned char is casted to BOX2DFLOAT4,
which is a "NO NO!" when it comes to memory alignment constraints.

I wonder why nobody is getting weird results on constrained platforms.

--strk;