On 07/04/2011 03:46 PM, Jim Meyering wrote:
Finally, to avoid the risk of error with negative "index", I suggest to change the semantics by incrementing that argument by 1 at each macro use and changing the name to "n_items".
At this point, and if we stick with unsigned types for sizes, the assertion on 0 < (n_required) becomes pointless (it should become 0 <= (n_required) and that is always true for unsigned types). It even gets in the way when applying REALLOC_IF_NECESSARY to the copy function, because it is sometimes called with a zero-element source. What about removing the assertion altogether?
Paolo
