Yoann Vandoorselaere <[EMAIL PROTECTED]> wrote:
> On Mon, 2005-05-16 at 15:20 +0200, Jim Meyering wrote:
>> Yoann Vandoorselaere <[EMAIL PROTECTED]> wrote:
>> > This patch fix constness warning in the GnuLib hash module.
>>
>> I'm all for avoiding warnings, but not when it detracts from what I
>> think of as the correctness of an interface, as it would in this case.
>>
>> Those `const void *entry' parameters constitute a promise
>> by the hash module that it will not try to write through
>> those pointers. Removing `const' here might make users of
>> these functions wonder whether the functions write through
>> the void pointers.
>
> Since the interface let the user provide a function for deleting data
> (which thus does not use const), you might then as well workaround this
> issue using:
>
> union {
> const void *val_ro;
> void *val_rw;
> } user_data;
>
> And use the later in order to provide the deletion callback with a not
> const pointer.
> Proof of concept patch attached.
I'll try to make it clearer this time. The promise is that the
named functions and their callees will not dereference through
the const pointers. Those `const' attributes say nothing about
what other functions (like the user-supplied deletion callback)
will do to the underlying data.
This is similar to using a const pointer to iterate through
non-const data because you're performing a search or doing some
other read-only job. The added (some might say `unnecessary')
const makes the code a little more readable and might also give
the compiler clues to let it generate better code.
Thanks for the patch, but IMHO such a change is not necessary
and would make hash.c (and its interfaces) less readable.
Can't you just turn off that particular warning when compiling hash.c?
_______________________________________________
bug-gnulib mailing list
bug-gnulib@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-gnulib
