Adds Guard Conditions

[christopherjreid]

• Adds the Guard Condition struct encapsulating rcl_guard_condition_t.
• Adds optional callbacks and a trigger method to approximate the rclcpp
  implementation.
• Adds an add_guard_condition method to WaitSet to add the
  GuardCondition to the WaitSet

[nnmm]

I'll check it out this evening!

[nnmm]

Great job as usual. I left a round of comments.

[christopherjreid]

Updated to address all comments!

[nnmm]

BTW, you could add guard conditions (and graph support, which I forgot) to the new rclrs/CHANGELOG.rst file in here too.

[christopherjreid]

All comments addressed, except for the open discussion on the return type of GuardCondition::new.
We may need a new discussion on the impl of PartialEq for GuardCondition. I believe that -only- checking eq on the underlying rcl_guard_condition is sufficient, but I'm open to thoughts on that. @nnmm suggested a static counter that would serve as an ID for each GuardCondition

[nnmm]

(Probably) final comments from my side

[esteve]

@christopherjreid in the future, could you push commit by commit, instead of force-push? It's easier to see the changes that happen while reviewing a PR. We'll squash all the commits in the end when we merge the PR, so there's no need to squash them while the PR is being worked on. Thanks.

[nnmm]

I also took a look at the clippy error – I hadn't thought about that. Maybe it would be nicest to have two methods:

• GuardCondition::new() (does not take a callback)
• GuardCondition::new_with_callback() (takes a callback, i.e. not enclosed in an Option)

That way the callback type could always be deduced. I think it also makes the call sites prettier.

Alternatively, we could just revert to taking boxed callbacks.

[esteve]

@christopherjreid I agree with @nnmm about adding an explicit GuardCondition::new_with_callback(), instead of an Option

[christopherjreid]

Hrm, so we're looking at:

pub fn new(...);

pub fn new_with_callback(...);

pub(crate) fn new_with_rcl_context(...);

pub(crate) fn new_with_rcl_context_with_callback(...);

?

[nnmm]

Hrm, so we're looking at:

pub fn new(...);

pub fn new_with_callback(...);

pub(crate) fn new_with_rcl_context(...);

pub(crate) fn new_with_rcl_context_with_callback(...);

?

No, the rcl_context version can receive a boxed Option callback

[nnmm]

LGTM!

[esteve]

@christopherjreid LGTM! Thank you so much for iterating with us and for your patience 🙂