The problem with these types of C tricks is that they only work right if the platform they're used on has the property of sizeof(void*) == sizeof(int). Unfortunately this is not always the case and not mandated by the C standard. That's the reason for these warnings. Its also a reason not to turn off the warnings.
Mike > -----Ursprüngliche Nachricht----- > Von: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] Im Auftrag von Dan > Gesendet: Samstag, 25. Oktober 2008 19:22 > An: General Discussion of SQLite Database > Betreff: Re: [sqlite] rtree cast warnings on 64bit OS - > strange parameter use > > > On Oct 25, 2008, at 11:40 PM, William Kyngesburye wrote: > > > I added rtree to my sqlite compilation for the first time and got > > these warnings for OSX 64bit: > > > > /Users/Shared/src/sqlite/sqlite-3.6.4/sqlite3.c: In function > > rtreeCreate: > > /Users/Shared/src/sqlite/sqlite-3.6.4/sqlite3.c:94784: > warning: cast > > from pointer to integer of different size > > /Users/Shared/src/sqlite/sqlite-3.6.4/sqlite3.c: In function > > rtreeConnect: > > /Users/Shared/src/sqlite/sqlite-3.6.4/sqlite3.c:94797: > warning: cast > > from pointer to integer of different size > > > > These are in the rtreeCreate() and rtreeConnect() functions, calling > > rtreeInit(): > > > > return rtreeInit(db, pAux, argc, argv, ppVtab, pzErr, 1, (int)pAux); > > > > I thought the two pAux parameters were odd - one bare and > one cast to > > (int), so I looked up rtreeInit(). > > Good point. I removed the first of the two "pAux" parameters > from rtreeInit(). It was not being used. > > http://www.sqlite.org/cvstrac/chngview?cn=5842 > > > static int rtreeInit(sqlite3 *db, void *pAux, int argc, const char > > *const*argv, sqlite3_vtab **ppVtab, char **pzErr, int isCreate, int > > eCoordType) > > > > The first pAux is a pointer, so this one looks correct. But the > > second is an int (eCoordType), and the only two values I > found defined > > are 0 & 1: > > > > #define RTREE_COORD_REAL32 0 > > #define RTREE_COORD_INT32 1 > > > > Forget the cast warnings now - why is pAux used to set the > eCoordType? > > My C skills are pretty basic, so maybe there is some pointer/cast > > magic happening? Or maybe it's simply screwed up? > > The two functions that call rtreeInit() are registered as > callbacks with SQLite. When you register the callback > function you also specify a void* pointer that is passed to > the callback whenever it is invoked. > This is not an uncommon pattern in C code. > > So, since the interface allows us to pass a void* as context > to the callback function, but in this instance we really just > want an integer, the value has to be cast to a void* when the > callback function is registered, and back to an integer when > the callback is invoked. > > A lot of compilers throw a warning when they encounter this. > In my opinion (having never had anything to do with compiler design or > implementation) they shouldn't. > > Dan. > > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
