Re: [Mingw-cross-env-list] Merging MXE Octave - first steps

[Tony Theodore]

On 27 Feb 2014, at 05:14, John W. Eaton <address@hidden> wrote:

> On 02/26/2014 12:55 PM, Tony Theodore wrote:
> 
>> Hdf5 is now done[1] and I’ll keep going with the existing deps.
> 
> OK.  I see that you have a patch to alter the configure script for
> HDF5 that removes AC_TRY_RUN tests.  It's also possible to just set
> those cache values on the configure command line:
> 
>  configure ... hdf5_cv_gettimeofday_tz=no ...
> 
> Either way will work, but if you set the values on the command line
> instead of modifying the configure.ac file, then you shouldn't have to
> regenerate the configure script and you can eliminate a patch file.

There are a couple of AC_TRY_RUN tests that can’t be set on the command line, 
but I agree that the ones that can, should be removed from the patch. I also 
think you have wine installed, the hdf5 build requires host generated sources 
which we have to patch in and step through the build as if those were created 
during the build. These are CPU specific (99% sure) and don’t really need to be 
generated for each target - though the textual info in H5lib_settings.c will be 
misleading.

Do you specify wine as a requirement?

>> One question: does ENABLE_64 mean “build x86_64-w64-mingw32” or “use
>> 64-bit integers for array dimensions and indexing”?
> 
> It currently means both.  We really need to have a third option, which
> is to build with x86_64-w64 (for 64-bit pointers) but to use the usual
> 32-bit integers for Fortran indexing.  I intend to do that by
> introducing another option, one to control the compiler and one to
> control the 64-bit integer indexing option.

OK, we use targets to control the compiler, and the yet-to-be-determined 
customisation hooks for things like the indexing options. Now I’m thinking we 
should add the customisation hooks as we go.

Cheers,

Tony