* ValueError: ('invalid number of arguments', mul(Subtensor{0}.0, 18))
This error happened after serializing a model. Scalar ops taking a variable
number of arguments had their ufunc set to that op with -1 input parameters, and
nout output params. Code has been added to check this condition and set ufunc to
None which postpones the building of the ufunc to when we know the actual number of
params (or something like that -see OlivierB.).

* Optimization bug which seemed rather serious.
Fixed bug with Canonizer and DimShuffle
Symptom: ValueError: ('Input dimensions do not match.'

code in a mess, but gemm-optimization works on more systematic test cases including josephs NAACL graph

fixed a bug with random number generation by submodules. See Oliviers email to theano-dev for details.

now you must give the --debug option to be in debug mode. This will be usefull as the file that don't use this main, will have the same default and we should be able to use it everywhere without trouble.

made the import work for "python test_logistic_regression.py" and for "nosetests test_logistic_regression.py"

instead of using Join for scalar values

added a JOSEPHS_BUG_SOLVED global variable to tensor.blas that disables dot22 and gemm optimizations