Some cleanup/documentation issues

[David Ronis <ronis at ronispc dot chem dot mcgill dot ca>]

I'm using gsl-1.1.1 on an i686-linux-gnu box.  Two small problems and
one question.

1.  I've written some code that has #define GSL_RANGE_CHECK_OFF
#define HAVE_INLINE defined; when compiling with -Wall I get errors
like:

In file included from /usr/local/include/gsl/gsl_matrix.h:4,
                 from /usr/local/include/gsl/gsl_multiroots.h:26,
                 from carl.c:5:
/usr/local/include/gsl/gsl_matrix_complex_long_double.h: In function `gsl_matrix_complex_long_double_get':
/usr/local/include/gsl/gsl_matrix_complex_long_double.h:230: warning: unused variable `zero'
In file included from /usr/local/include/gsl/gsl_matrix.h:5,
                 from /usr/local/include/gsl/gsl_multiroots.h:26,
                 from carl.c:5:
/usr/local/include/gsl/gsl_matrix_complex_double.h: In function `gsl_matrix_complex_get':
/usr/local/include/gsl/gsl_matrix_complex_double.h:230: warning: unused variable `zero'
In file included from /usr/local/include/gsl/gsl_matrix.h:6,
                 from /usr/local/include/gsl/gsl_multiroots.h:26,
                 from carl.c:5:
/usr/local/include/gsl/gsl_matrix_complex_float.h: In function `gsl_matrix_complex_float_get':
/usr/local/include/gsl/gsl_matrix_complex_float.h:230: warning: unused variable `zero'

The definitions of the const zero should be moved inside the blocks
protected by #ifndef GSL_RANGE_CHECK_OFF

2.  I'm using gsl_multiroot_test_delta() to check whether a root
    finding iteration has converged.  There doesn't seem to be a
    function defined (or documented) that gives me the dx's, analogous
    to gsl_multiroot_fsolver_root().  Am I correct in assuming that if
    S gsl_multiroot_fsolver struct, then what I want is S.dx?  If it's
    that simple, then shouldn't gsl_multiroot_fsolver_root simply
    implement itself as a macro:
    
    #define gsl_multiroot_fsolver_root(s) ((s)->x)

    In any event, the documentation needs to mention how to get dx.

3.  Finally, I have to do a lot of vector manipulations in the
function I'm trying to zero, and I'd like to eliminate all the
overhead associated with calls to gsl_vector_get/set.  Can I assume
that stride is 1 for all vectors?  If so, I can access the data member
directly, and faster.


David