Re: Octave/Win32 update

[David Bateman]

John W. Eaton wrote:
> On 29-Jan-2007, David Bateman wrote:
> 
> | Ouch... That's a pretty silly dependency.. I can see why it is there, as
> | "hid_t" must be defined for the save_hdf5 and load_hdf5 functions.. I'd
> | be inclined to remove the include of hdf5.h and replace it with
> | 
> | typedef int hid_t;
> | 
> | Can you try editing c:\programs\octave\include\octave-2.9.9+\octave\ov.h
> | and do that? John what change do you want to make to support this, or do
> | you want to require that HDF5 development files are installed to build
> | oct-files?
> 
> Eliminating this dependency would be fine, but I don't want to
> duplicate the typedef in an Octave header.  So we need to find some
> other way to hide hid_t, maybe wrapping it in a class?  Then the ov.h
> header could just include a
> 
>   class octave_hdf5_hid_type;
> 
> forward declaration and the argument list of the load/save functions
> could be changed to use
> 
>   const octave_hdf5_hid_type& loc_id
> 
> instead of using
> 
>   hid_t loc_id
> 
> directly.  Would that work?
> 
> jwe
> 

Would this require that all hid_t declarations in the save_hdf5 and
load_hdf5 function would need to be modified to use this class? If so
that might be a bit of a pain as there is some octave-forge code that
implements these functions.. Anything that hides the hid_t arg at the
level seen by mkoctfile would be ok..

D.