On Sun, Sep 13, 2015 at 02:49:21PM +0200, Simon Richter wrote: > The C++ specification disallows this to permit implementations that use > the address-of operator and sizeof to calculate the address of the first > argument inside the ellipsis; sizeof(foo &) gives the size of the "foo" > class rather than the space taken up by the reference.
Pretty obscure rule, never heard of that :D I know that the 'thing' in the vararg are pretty limited (IIRC only ints, doubles and pointers) but never heard of a limitation in the address before. Well, it's C++ only, C doesn't suffer from that problem:P I suppose that is talking about the implementation of va_start: address of last+sizeof of last gives the starting point of the va_list (in the 'correct' stack direction, obviously); am I correct? Otherwise there should be 'compiler magic' involved in the va_* implementation (it knows that a reference is passed by address). gcc probably seems to be using this magic: https://gcc.gnu.org/onlinedocs/gccint/Varargs.html ... let me guess, a clang warning? -- Lorenzo Marcantonio Logos Srl _______________________________________________ Mailing list: https://launchpad.net/~kicad-developers Post to : [email protected] Unsubscribe : https://launchpad.net/~kicad-developers More help : https://help.launchpad.net/ListHelp
