On 18 December 2014 at 19:54, Bill Fischofer <[email protected]> wrote:
> There is no architectural requirement that they be distinct and, as Raras
> noted, most implementations will just make them synonyms.  That's why the
> patch is classified as a documentation change--there's no actual change in
> observable behavior.
If the semantics is the same, why then not have only one symbolic
constant? How is the architecture better by having two symbolic
constants that mean the same?

>
> Bill
>
> On Thursday, December 18, 2014, Ola Liljedahl <[email protected]>
> wrote:
>>
>> On 18 December 2014 at 16:47, Taras Kondratiuk
>> <[email protected]> wrote:
>> > On 12/18/2014 05:41 PM, Savolainen, Petri (NSN - FI/Espoo) wrote:
>> >>
>> >>
>> >>
>> >>> -----Original Message-----
>> >>> From: [email protected] [mailto:lng-odp-
>> >>> [email protected]] On Behalf Of ext Taras Kondratiuk
>> >>> Sent: Thursday, December 18, 2014 4:36 PM
>> >>> To: Bill Fischofer; [email protected]
>> >>> Subject: Re: [lng-odp] [PATCH] Documentation: Change
>> >>> odp_buffer_pool_info_t output
>> >>>
>> >>> On 12/16/2014 01:43 PM, Bill Fischofer wrote:
>> >>>>
>> >>>> Change odp_buffer_pool_info() output to use ODP_SHM_INVALID instead
>> >>>> of
>> >>>> ODP_SHM_NULL.
>> >>>>
>> >>>> Signed-off-by: Bill Fischofer <[email protected]>
>> >>>
>> >>>
>> >>> To follow our naming convention it should be named:
>> >>> api: buffer: Change odp_buffer_pool_info_t output
>> >>>
>> >>> Otherwise
>> >>> Reviewed-by: Taras Kondratiuk <[email protected]>
>> >>>
>> >>> Should ODP_SHM_NULL definition be removed completely now from API?
>> >>
>> >>
>> >> No. The idea is that user input param is XXX_NULL if the param is
>> >> optional
>> >> and the user don't specify a handle. Output from ODP would be always
>> >> XXX_INVALID or valid handle.
>> >>
>> >>
>> >> shm = odp_shm_reserve(...) // returns INVALID
>> >>
>> >> odp_buffer_pool_create(..., shm, ...); // I care but didn't check the
>> >> shm
>> >> output
>> >>
>> >> vs.
>> >>
>> >> odp_buffer_pool_create(..., ODP_SHM_NULL, ...); // I don't care
>> >
>> >
>> > Sorry if it was already discussed before, but what is the advantage of
>> > having both NULL and INVALID? Why not to leave only one of them?
>> > I assume on most of platforms they will anyway map to the same value.
>> If Petri's example above should work (as I expect it to work), they
>> cannot have the same value.
>>
>> I think it is confusing to provide both a _NULL and an _INVALID
>> symbol. Perhaps the _NULL symbol should be renamed to _DEFAULT or
>> something else that is semantically far away from _INVALID.
>>
>> The odp_buffer_pool_create() function could also specify that
>> ODP_SHM_INVALID has the meaning of do default behavior. If the user is
>> too lazy to check the return value of odp_shm_reserve() and
>> unexpectedly passes ODP_SHM_INVALID, it could be viewed as their
>> problem.
>> >
>> >
>> > _______________________________________________
>> > lng-odp mailing list
>> > [email protected]
>> > http://lists.linaro.org/mailman/listinfo/lng-odp

_______________________________________________
lng-odp mailing list
[email protected]
http://lists.linaro.org/mailman/listinfo/lng-odp

Reply via email to