On 04/06/15 10:22, Stuart Haslam wrote:
On Wed, Jun 03, 2015 at 09:37:54PM +0200, Ola Liljedahl wrote:
On 3 June 2015 at 20:42, Zoltan Kiss <[email protected]> wrote:
On that note it turned out this will be a lot bigger than I thought ...
And I already have two questions:
- what should we do if an assert over odp_queue_enq fails? E.g.
"CU_ASSERT(odp_queue_enq(queue, ev) == 0)" Should we free the events, or
just leave it?
If we want the validation program to terminate cleanly, I assume the caller
has to free any buffers that weren't enqueued or they will be leaked and
then things will go bad when the program is terminating.
- there are numerous places where you don't know what kind of event you
are actually queue, so should we make an odp_event_free() call?, e.g.
I have also encountered this situation. I even started to write a post to
the list about it. But in the end I hacked around it, probably by switching
on the event type and calling different free functions. Having an
odp_event_free() would be simpler and clearer so I second your suggestion
here.
-- Ola
Me too. Actually it looks like I did;
odp_buffer_free(odp_buffer_from_event(ev));
Is this not sufficient?..
No, it releases the buffer as it would be a raw buffer. A related issue
is that odp_buffer_* functions seem to handle the input as a raw buffer,
while an unsuspecting developer might think they are generic buffer
functions which can handle any kind of buffers.
anyway I agree odp_event_free() would be
clearer.
I'll send in a patch for that too
_______________________________________________
lng-odp mailing list
[email protected]
https://lists.linaro.org/mailman/listinfo/lng-odp
