You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Yeah, that is a magic number that should be cleaned up. It might make sense to actually pick the sleep period in relation to the timeout in general, something like 100th of the timeout. The rate of polling needs to be traded off against performance. At the low end it will end up saturating at the minimum sleep period. And at the high end it would cause ~100 calls per timeout consistently.
… (ros#330)
* Clear callback from transformable_callbacks_ when callback is done or not needed
Co-authored-by: marting87 <martinganeff@gmail.com>
Co-authored-by: Matthijs den Toom <mdentoom@lely.com>
Co-authored-by: Chris Lalancette <clalancette@gmail.com>
seanyen
pushed a commit
to ms-iot/geometry2
that referenced
this issue
Mar 22, 2021
* Make BufferCore callback handle maintenance internal.
That is, there is really no reason for external users to know
about TranformableCallbackHandle at all. Instead, hide all of
the implementation detail about the handles within BufferCore
itself. This gets simplifies the API, and allows us to ensure
that cleanup happens properly.
Signed-off-by: Chris Lalancette <clalancette@openrobotics.org>
Co-authored-by: Chris Lalancette <clalancette@openrobotics.org>
Currently,
Buffer::canTransform
's timeout is implemented by polling with a cycle of 0.01ms. Therefore shorter timeouts than 0.01ms are not possible.I would like to suggest that, if
time_ou
t < 0.01ms, the polling-cycle should be adapted.At least, if not changed, this behavior should be documented anywhere...
The text was updated successfully, but these errors were encountered: