Re: Warning for unadorned 0 in varargs lists?

[Clifford Wolf <clifford at clifford dot at>]

On Mon, Aug 23, 2004 at 06:25:51PM +0200, Andreas Schwab wrote:
> >>> I used to feel as you do, that since 0 works in practice we should
> >> In practice it does _not_ work.
> > On which system, specifically?
> ia64, for example.  Probably also amd64.

Every system with sizeof(void*) != sizeof(int). This is almost every 64 bit
system which defines int to be 32 bit but a pointer 64 bit.

However, afair ANSI C says that sizeof(void*) == sizeof(long), so it should
be save to use 0L for null pointers (unless the abi defines something like
extending all pointers to 128 bit values in function calls or something
alike - I think some mainframe systems made that with the upper 64 reserved
for the type of the pointer - but I'm not sure if this is ok with ANSI C).

by the end, that pointers can be compared with 0 to check for null
pointers doesn't imply that it would be save to use 0 as a pointer without
explictely casting it to one.

yours,
 - clifford

-- 
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* Prof:    So the American government went to IBM to come up with *
*          a data encryption standard and they came up with ...   *
* Student: EBCDIC!                                                *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

checking whether the linker (ld -static) supports shared libraries... yes