On 17-Apr-08, at 8:11 AM, Christos Zoulas wrote:
It is not an issue of avoiding the warnings. For example isspace() is defined in:http://www.opengroup.org/onlinepubs/009695399/functions/ isspace.htmlas: The c argument is an int, the value of which the application shall ensure is a character representable as an unsigned char or equal to the value of the macro EOF. If the argument has any other value, the behavior is undefined.So by casting a signed char to an int, you can get a large negative numberand you end up with "undefined behavior" which might mean a core dump, so don't do it!
Argh! Sorry, I should have been able to figure that out myself, given the usual implementations. My brain doesn't seem so good at seeing the possibility of negative array indexing -- too many years of sloppy assumptions.
However what you and Alan said about simply casting to (unsigned char) isn't sufficient, at least on some platforms (IRIX-6, IIRC), at least not to keep the compiler happy.
Sadly I do get the least warnings from every platform when the cast is only to (int). I wonder if the compiler can be taught to detect uses where a parameter may have a value out of range even though the data type it is declared as is wide enough. The only solution the compiler can really believe though, I think, would be a double cast where the value is narrowed sufficiently before being widened again for the call.
Just as sadly Joerg's advice to read about the value in the CAVEATS section doesn't help, at least not on NetBSD-4.0 or earlier. Unfortunately I also missed the section you quoted from the SUS standard, so I'm not sure peeking at -current would have helped me. Neither the SUSv2 docs online, nor any variant of the NetBSD manual pages, contain any good examples.
--
Greg A. Woods; Planix, Inc.
<woods%planix.ca@localhost>