[patch] Fix for issue #125

[Jules Bergmann]

While working on Ext_data for the reference implementation, I ended up 
fixing issue #125 "Fft cannot take an expression as an argument" by 
allowing Ext_data objects to be created for expression blocks (the 
ref-impl needs to use Ext_data on expression blocks to implement some of 
the matrix vector operations, such as prodh()).

The root problems are Ext_data (and Rt_ext_data)

(a) cannot be instantiated for const blocks, because it results in 
ambiguous constructors, and

(b) cannot be instantiated for blocks without a put() method, because 
the sync value is a run-time parameter.

The fixes are:

- explicitly disambiguate the constructors for non-const and const block 
types.

- make Block_copy from a pointer to a const block a runtime error.  To 
reduce duplication of code for the copy to a pointer (the other 
direction), Block_copy was split into separate classes Block_copy_to_ptr 
and Block_copy_from_ptr.


The patch also includes a regression test for the issue, and some 
additional argument checking for Fft to help debug the regression test :).

I'm currently running the regression suite to make sure the Block_copy 
change did not affect other code.  I will apply after that looks good.

				-- Jules

-- 
Jules Bergmann
CodeSourcery
jules at codesourcery.com
(650) 331-3385 x705
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 125.diff
URL: <http://sourcerytools.com/pipermail/vsipl++/attachments/20061205/d937e0eb/attachment.ksh>