Re: Renaming libcruft

[Rik]

On 08/24/2012 01:50 PM, John W. Eaton wrote:
> | >
> | > | The fact that it is
> | > | Fortran really segregates it in my mind from the rest of the C++ 
> sources. 
> | > | I would still change the directory name away from libcruft to 
> libfortran.
> | >
> | > Not all the code in libcruft is Fortran.  Look at the misc directory.
> | > That's mostly for interfacing C with Fortran, but it also has the
> | > lowest-level error handling functions and variables.
> | Does this rule out calling the new directory libfortran then, or can we
> | overlook the ~2.7% of the code in misc which is non-Fortran? 
> | Alternatively, we could somehow reference netlib if that's where this code
> | came from.
>
> How about extracting the misc directory out of libfortran and putting
> it with other miscellaneous functions in some other subdirectory of
> liboctave?
Yeah that's good.
>
> Also, I don't see why we need to subdirectory "lib"fortran.  The
> library built there will be there for convenience only.  We don't use
> that naming for other directories that hold the sources that are used
> to build other convenience libraries like corefcn.  Maybe just
> numerical, misc-numerical, or something similar.  Naming the directory
> "fortran" does let me know that the files there are probably written
> in Fortran, but it doesn't tell me what purpose the files there have.
I agree that once libcruft is no longer a top-level directory it should
drop the lib prefix and just be whatever is descriptive of the contents. 
If "fortran" isn't good, what about netlib since that is the source for
this code?  Also, if this directory is parallel to the 'math' directory
then maybe it should be "fortran-math".  I notice that netlib refers to
pieces of code as routines so maybe we could work that in.  Are
"netlib-rtns", "numerical-fortrtns" any better?

--Rik