SCM_NEWCELL{2}?

[Dirk Herrmann <dirk at ida dot ing dot tu-bs dot de>]

Hi!

In order to get incrementally rid of the type inconsitencies in guile I'd
like to change the SCM_NEWCELL macros as the following example
shows:  (the lines that change are marked ***)

The current definition:

#define SCM_NEWCELL(_into) \
        do { \
          if (SCM_IMP (scm_freelist)) \
             _into = scm_gc_for_newcell (&scm_master_freelist, \
                                         &scm_freelist); \
          else \
            { \
               _into = scm_freelist; \
***            scm_freelist = SCM_CDR (scm_freelist);\
***            SCM_SETCAR (_into, scm_tc16_allocated); \
            } \
        } while(0)


becomes:


#define SCM_NEWCELL(_into) \
        do { \
          if (SCM_IMP (scm_freelist)) \
             _into = scm_gc_for_newcell (&scm_master_freelist, \
                                         &scm_freelist); \
          else \
            { \ 
               _into = scm_freelist; \
***            scm_freelist = SCM_CELL_OBJECT_1 (scm_freelist);\
***            SCM_SET_CELL_TYPE (_into, scm_tc16_allocated); \
            } \
        } while(0)


The reason is, that neither the cells on the freelist from valid scheme
pairs (i. e. SCM_CONSP is wrong), nor is the newly allocated cell a valid
scheme cell (same reason).  Thus, accessing cells from the freelist with
SCM_CDR is not appropriate, nor is SCM_SETCAR appropriate to access the
allocated cell.  The other SCM_NEWCELL definitions would have to be
changed accordingly.

Please note that the resulting code will not be different (I hope
:-).  It's just the type strictness that (I think) will be improved.

Are there objections to my interpretations of scm_freelist and
scm_tc16_allocated and the consequences for the SCM_NEWCELL macros?  If
not, I will apply the changes.

Best regards
Dirk Herrmann