Re: feature request

[Dirk Herrmann]

Hello everybody,

in CVS guile, SCM is now defined as:
   typedef struct scm_unused_struct * SCM;
Hmmm.  I just realize that it should probably rather be
   typedef struct scm_unused_struct_t * SCM;
I'll fix that soon.

Anyway, SCM_VOIDP_TEST is gone, instead SCM_DEBUG_TYPING_STRICTNESS can
take values 0, 1, 2 where 1 is the default and will use the above
definition for SCM.

The new definition actually showed some errors, which are hopefully also
fixed in CVS.  One compile time warning can't easily be removed without
larger changes to the options implementation, which is currently in some
parts quite, well, let's call it interesting :-)

Another very interesting issue came up in a private discussion with
Michael:  When giving the compile time option -W in addition to -Wall to
gcc, it will display quite some more usefull warning messages.  Among them
are empty if blocks (helped me to fix a bug in ramap.c), comparisons of
unsigned values against >= 0 (which is always true and thus indicates a
potential signedness problem), and comparisons of signed against unsigned
variables.  From the latter there are still quite a lot remaining, but I
guess it doesn't make much sense to check them before we make scm_bits_t
into an unsigned type.

Best regards,
Dirk Herrmann