I think this discussion is really good, it highlights that we don't have a
list of what is missing and what people know needs to be done "todo".

*If a feature has never existed it needs to be a new work card.*
*If the code exists and it fails to work fully or is not well documented I
think it is a bug  that needs a patch, or if it is simple just submit a
patch.*
*All todos should be bugs before the code is accepted, the debug tag/id
should be in the todo*

In this case I think it is important that the actual checked in code has a
significant flaw that a user should know about, it should be a bug to add
the functionality which might take a while to be solved. If LNG pick up
solving it it will become a CARD and eventually it will be a patch either
way.

On 29 August 2014 08:49, Savolainen, Petri (NSN - FI/Espoo) <
[email protected]> wrote:

>  Still book keeping should be done in a single place. It’s confusing if
> TODO/missing feature lists  are spread over doxygen comments, bugzilla bug
> reports, cards, … I’d use doxygen only for documenting the existing code,
>  and keep list missing features somewhere else.
>
>
>
> E.g. in this case, when someone adds the missing function (eventually),
> it’s easy to forget to remove the comment from the reserve function (or N
> other places)
>
In my opinion worries about correctly changing the code in the future is
not a good reason not to fix what does exist now.
Options:

   - In this case we can give the heads up with a simple patch (this one as
   is)
   - Convert it to a todo that lists the bug ID.
   - THE BEST WAY - is in this thread we define the new API and this patch
   can add it with a dummy implementation that returns unsupported.

But I am ok with a todo - is that ok with you Petri?

>
>
> -Petri
>
>
>
>
>
> *From:* ext Bill Fischofer [mailto:[email protected]]
> *Sent:* Friday, August 29, 2014 3:19 PM
> *To:* Mike Holmes
> *Cc:* Savolainen, Petri (NSN - FI/Espoo); [email protected]
> *Subject:* Re: [lng-odp] [PATCH] odp_shared_memory.h: Document
> odp_shm_reserve
>
>
>
> We definitely need the closure APIs to be part of v1.0 for completeness.
>  For now noting their absence as bugs is a good way of reminding us of that
> fact.
>
>
>
> Bill
>
>
>
> On Fri, Aug 29, 2014 at 6:34 AM, Mike Holmes <[email protected]>
> wrote:
>
>
>
>
>
> On 29 August 2014 04:05, Savolainen, Petri (NSN - FI/Espoo) <
> [email protected]> wrote:
>
>
>
> > -----Original Message-----
> > From: [email protected] [mailto:lng-odp-
> > [email protected]] On Behalf Of ext Mike Holmes
> > Sent: Thursday, August 28, 2014 11:54 PM
> > To: [email protected]
> > Subject: [lng-odp] [PATCH] odp_shared_memory.h: Document odp_shm_reserve
> >
> > Signed-off-by: Mike Holmes <[email protected]>
> > ---
> >  platform/linux-generic/include/api/odp_shared_memory.h | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/platform/linux-generic/include/api/odp_shared_memory.h
> > b/platform/linux-generic/include/api/odp_shared_memory.h
> > index 8ac8847..d75b179 100644
> > --- a/platform/linux-generic/include/api/odp_shared_memory.h
> > +++ b/platform/linux-generic/include/api/odp_shared_memory.h
> > @@ -21,7 +21,7 @@ extern "C" {
> >
> >  #include <odp_std_types.h>
> >
> > -/** Maximum shared memory block name lenght in chars */
> > +/** Maximum shared memory block name length in chars */
> >  #define ODP_SHM_NAME_LEN 32
> >
> >
> > @@ -33,6 +33,9 @@ extern "C" {
> >   * @param align  Block alignment in bytes
> >   *
> >   * @return Pointer to the reserved block, or NULL
> > + *
> > + * @note The same block name cannot be reused.
> > + * @note There is no way to unreserve a named block
>
> I think there will be a odp_shm_release call. So I'd not put that last
> note there...otherwise half of the current API should be noted about
> missing features.
>
> The problem I face is that this is the API as it stands,  and we don't
> have any roadmap that says the symmetry will be added by 1.0, we need one.
>
>
>
> It comes into focus right now that the unit tests are running into it.The
> test cases either have to prove an API that does not exist or up date the
> docs to match reality and test against that.
>
>
>
> I feel the best approach rather than add  todo allover is to actually
> document reality, prove it with the unit tests,  and then update the
> code/docs/unit tests as we make progress.
>
>
> -Petri
>
>
> >   */
> >  void *odp_shm_reserve(const char *name, uint64_t size, uint64_t align);
> >
> > --
> > 1.9.1
> >
> >
>
> > _______________________________________________
> > lng-odp mailing list
> > [email protected]
> > http://lists.linaro.org/mailman/listinfo/lng-odp
>
>
>
>
>
> --
>
> *Mike Holmes*
>
> Linaro Technical Manager / Lead
>
> LNG - ODP
>
>
> _______________________________________________
> lng-odp mailing list
> [email protected]
> http://lists.linaro.org/mailman/listinfo/lng-odp
>
>
>



-- 
*Mike Holmes*
Linaro Technical Manager / Lead
LNG - ODP
_______________________________________________
lng-odp mailing list
[email protected]
http://lists.linaro.org/mailman/listinfo/lng-odp

Reply via email to