(Razvan's branch will introduce a test that catches this bug)

Fix bug in infer_shape of Flatten

Check the graphviz executable are there.

Even if the pydot Python package is installed, it is not certain that
the graphviz executable are (that is the case with EPD for instance).
In that case, we do not want to use it.

Infershape would incorrectly assign a shape of 1 on the last dimension,
if all dimension were preserved.
In particular, calling .flatten() on some vector would infer a shape of
(1,), which is incorrect.

C code for GpuIncsub, and a new CUDA kernel for one special case

Fix an optimization crash of local_gpu_elemwise_0.

Call to theano.compile.rebuild_collect_shared will clone variable that have owner.

Change dtype of shared variable to floatX

Add erfinv and erfcinv ops

Fix test for when floatX=float32

Borrow=True is dangerous if one output destroys another

Re-add part of the dtype constraint on out grads

Batched dot22

This is a temporary fix until we fix the behaviour of borrow=True
for the general case.

Re-add (and update) bug fix description.

Err msg unalign

Doc python mem

C version of view op

Fixes after code review

INTERFACE CHANGE: Disable using unaligned tensor in all case. We supported it for python code before.