Re: [pkcs11] Proposal: All constants should have UL suffix

[Wan-Teh Chang <wtc@google.com>]

On Tue, May 28, 2013 at 2:12 AM, Stef Walter <stefw@redhat.com> wrote:
> All numeric constants in the specification section "A. Manifest
> Constants" should have the suffix 'UL' on them. This allows the code
> like this to return the correct result:
>
>   sizeof(CKO_DATA)
>
> Among other things, it also allows the constants to be passed correctly
> to C varargs functions.
>
> Since all document edits for this would be tedious. I'll just give a few
> examples:
>
> - #define CKU_SO         0
> + #define CKU_SO         0UL
>
> and
>
> - #define CKA_CLASS      0x00000000
> + #define CKA_CLASS      0x00000000UL
>
> Above applies to every constant in the specification section "A.
> Manifest Constants" as well obviously as the reference headers.

Hi Stef,

I think this proposal is a fine idea. To help us review this proposal,
you need to point out that the manifest constants in question are all
intended for types defined as CK_ULONG. For example, CKO_DATA,
CKO_CERTIFICATE, etc. are for the CK_OBJECT_CLASS type, which is
defined as CK_ULONG.

This fact was not evident to me (although I suspected it is true), so
I had to check the type definitions to convince myself.

A near counter example is the manifest constants CK_TRUE and CK_FALSE,
which I believe are intended for the CK_BBOOL type, but CK_BBOOL is
not defined as CK_ULONG. Since CK_TRUE and CK_FALSE are not listed in
the specification section "A. Manifest Constants", your proposal is
still accurate, but if you can provide such additional information,
it'll make the review easier.

Thanks,
Wan-Teh