FreeTDS Development Group

["James K. Lowden" <jklowden AT schemamania.org>]

"Craig A. Berry" <craigberry AT mac.com> wrote:
> James K. Lowden wrote:
> > - if (freopen(options.input_filename, "r", stdin) < 0) {
> > + if ((int)freopen(options.input_filename, "r", stdin) < 0) {
> > 
> > Could I change all these to:
> > 
> > if (freopen(options.input_filename, "r", stdin) < (FILE*)0) {
> > 
> > I think that's more honest.  
> 
> It looks like the compiler will take that, though it surprises me that
> it doesn't complain about comparing zero to an unsigned value. A pointer
> can't be negative can it?

Not only can it not be negative, the code is wrongo.  As you noted,
freopen(3) and friends return NULL, not 0 (where NULL is an
implementation-defined macro representing the "null pointer").  

The code should be:

if (freopen(options.input_filename, "r", stdin) == NULL)

cf. http://www.vmunix.com/~gabor/c/draft.html sections 6.2.2.3 and 7.1.6. 


That's supposed to work, right?  Is you compiler happy with that? 

> HOWEVER, as I belatedly look at the docs, my CRTL documentation and C99
> agree that fopen and freopen both return NULL on failure, not a negative
> number. So apparently what we really should be discussing is changing
> them all to "== (FILE*)NULL".

It's supposed to be unnecessary to cast a void* or NULL to anything.  ISTR
one can use 0 for any null pointer, but perhaps that is wrong or out of
date.  

I think we are approaching the truth, albeit perhaps asymptotically.  ;-)

--jkl