H5Part development and discussion

[Achim Gsell <achim AT cybercity.ch>]

Hi Mark,

On Jul 1, 2010, at 10:20 PM, Mark Howison wrote:

> I was wondering why there is so much casting in the TRY construct:
> 
> #define TRY( func )                                             \
>        if ((int64_t)(ptrdiff_t)(func) <= (int64_t)H5_ERR)      \
>                return H5_ERR;
> 
> What if we had a convention where every function must return a signed
> int, either h5_ssize_t for any function that returns a number of
> something, or h5_err_t if the function is only returning an
> error/success code. Then we could compare directly to H5_ERR.

Actually I already thought about this several times. But we have functions 
returning a pointer, like h5_open_file() or the memory (re-)allocation 
functions. On error these functions return a pointer value H5_ERR. H5_ERR is 
defined as -1, which isn't a valid pointer value on all systems I know.

Achim