OG> On Thu, Jan 20, 2000 at 03:17:43AM -0800, Martin Buchholz wrote: >> >> OG> On Wed, Jan 19, 2000 at 04:35:23AM -0500, Olivier Galibert wrote: OG> And here it is. There are some LISP_TO_VOID and back, but that's OG> where I thought it was the clearer thing to do. Martin? OG> Note that these "errors" are only because you're compiling with a C++ OG> compiler. The code is perfectly valid ISO C code. And frankly, I OG> tend to consider all the added casts unnecessary noise. OG> OTOH, being able to get the better warnings a C++ compiler can give OG> you is a good thing, so I won't veto any patch on this basis. I should document more clearly that XEmacs is written in "Clean C", that is the intersection of ANSI C++ and ANSI C, for reasons of cleanliness and having more compilation tools available. OG> The code does: OG> if (the lisp object is actually a pointer) then OG> add <delta> to the pointer OG> endif I'm looking at this code: if (POINTER_TYPE_P (XTYPE (dobj))) { char *objp = (char *) LISP_TO_VOID (dobj); if (objp) objp += delta; VOID_TO_LISP (*pobj, objp); } Should we ever be storing a NULL pointer? Shouldn't we be doing assert (objp != NULL); /* Here is code that doesn't use the fact that there is only one Lisp_Type that is a pointer. It would need to be adjusted if *pobj could be Qzero_bits. */ enum Lisp_Type type = XTYPE (*pobj); if (POINTER_TYPE_P (type)) { XSETOBJ (*pobj, type, (char *) XPNTR (*pobj) + delta); }