Error associated with graphrhc.cpp

[dodsonmg]

I'm cross-compiling ROS2 dashing for FreeBSD.

When compiling rmw_cyclonedds_cpp, I get the following errors:

Starting >>> rmw_cyclonedds_cpp                                                                                                                                                                             
--- stderr: rmw_cyclonedds_cpp                                                                                                                                                                              
/home/broomstick/cheri/ros2/src/ros2/rmw_cyclonedds/rmw_cyclonedds_cpp/src/graphrhc.cpp:169:3: error: cannot initialize a member subobject of type 'dds_rhc_read_take_cdr_t' (aka 'int (*)(dds_rhc *, bool, 
ddsi_serdata **, dds_sample_info *, unsigned int, unsigned int, unsigned int, unsigned int, unsigned long)') with an lvalue of type 'bool (struct dds_rhc *, struct dds_readcond *)': different number of p$
rameters (9 vs 2)                                                                                                                                                                                           
  graphrhc_add_readcondition,                                                                                                                                                                               
  ^~~~~~~~~~~~~~~~~~~~~~~~~~                                                                                                                                                                                
/home/broomstick/cheri/ros2/src/ros2/rmw_cyclonedds/rmw_cyclonedds_cpp/src/graphrhc.cpp:170:3: error: cannot initialize a member subobject of type 'dds_rhc_add_readcondition_t' (aka 'bool (*)(struct dds_$
hc *, struct dds_readcond *)') with an lvalue of type 'void (struct dds_rhc *, struct dds_readcond *)': different return type ('bool' vs 'void')                                                            
  graphrhc_remove_readcondition,
  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/broomstick/cheri/ros2/src/ros2/rmw_cyclonedds/rmw_cyclonedds_cpp/src/graphrhc.cpp:171:3: error: cannot initialize a member subobject of type 'dds_rhc_remove_readcondition_t' (aka 'void (*)(struct dd
s_rhc *, struct dds_readcond *)') with an lvalue of type 'uint32_t (struct dds_rhc *)' (aka 'unsigned int (dds_rhc *)'): different number of parameters (2 vs 1)
  graphrhc_lock_samples,
  ^~~~~~~~~~~~~~~~~~~~~
/home/broomstick/cheri/ros2/src/ros2/rmw_cyclonedds/rmw_cyclonedds_cpp/src/graphrhc.cpp:172:3: error: cannot initialize a member subobject of type 'dds_rhc_lock_samples_t' (aka 'unsigned int (*)(dds_rhc *
)') with an lvalue of type 'dds_return_t (struct dds_rhc *, struct dds_reader *, const struct ddsi_sertopic *, struct ddsi_tkmap *)' (aka 'int (dds_rhc *, dds_reader *, const ddsi_sertopic *, ddsi_tkmap $
)'): different number of parameters (1 vs 4)
  graphrhc_associate

I was previously building cyclonedds from commit 0b50928. I think commits 9aef055 and 0006e09 made some changes associated with dds_rhc (that I don't really understand... sorry) that might be related to my problems above.

Should I be able to continue building ROS2 dashing using the the cyclonedds master branch?

It looks like the repos file for ROS2 dashing pulls all the way back from releases/0.5.x (269f18e), but I wanted to check.

Thanks!

[eboasson]

Hi @dodsonmg, your analysis is entirely correct. The thing is that the Cyclone DDS RMW layer depends on some "evolving" interfaces — a nice word for "haven't had a chance to do it properly yet" — to be precise:

1. It uses a custom implementation of samples so it can do (de)serialization straight from ROS2 to CDR and back;
2. In Dashing and Eloquent, it implements a custom "reader history cache" (hence the "rhc") for getting notifications of changes to the ROS2 graph without incurring the memory usage you would otherwise have (as I was aware of some small device where memory use really mattered). In Foxy, a lot of major changes occurred that made this one superfluous.

The branching is to prevent regressions on released, but it doesn't mean I have completely given up on the ability to build Dashing with the current version of Cyclone (that'd be stupid). What you have run into is more like the forgetting of merging a PR on the Cyclone RMW implementation (ros2/rmw_cyclonedds#169) in the flurry of activity with Foxy and Cyclone releases. As you can see from the PR, it is trivial. It simply extends the range of Cyclone versions it works with, it doesn't break anything for older versions.

Please give it a try. I did try it before doing the PR, but if you could confirm it works, that'd a much nicer endorsement for merging it.

[dodsonmg]

Thanks @eboasson. I appreciate both the quick confirmation and the additional information.

I can successfully build cyclonedds from master (currently 84abf48) and rmw_cyclonedds from dashing-eloquent with your patch from ros2/rmw_cyclonedds#169.

[eboasson]

Thanks!