make writer_guid uint8_t[] instead of int8_t[] for consistency with rmw_gid_t

[ihasdapie]

This PR address another inconsistency I noticed with with gids in rmw, namely that rmw_gid_t is an uint8_t array and rmw_request_id_t.writer_guid is a int8_t array. Changing this would silence some rcl compilation warnings introduced by the service introspection feature (ros2/ros2#1285).

Reference:

rmw/rmw/include/rmw/types.h

Lines 356 to 363 in 2259c3f

	typedef struct RMW_PUBLIC_TYPE rmw_request_id_s
	{
	/// The guid of the writer associated with this request
	int8_t writer_guid[16];
	/// Sequence number of this service
	int64_t sequence_number;
	} rmw_request_id_t;

rmw/rmw/include/rmw/types.h

Lines 620 to 627 in 2259c3f

	typedef struct RMW_PUBLIC_TYPE rmw_gid_s
	{
	/// Name of the rmw implementation
	const char * implementation_identifier;
	/// Bype data Gid value
	uint8_t data[RMW_GID_STORAGE_SIZE];
	} rmw_gid_t;

I'm going to keep this PR as draft until I review the rmw implementations to check if any notable assumptions are made about the signedness of writer_guid, but I don't expect it to be an issue.

Signed-off-by: Brian Chen brian.chen@openrobotics.org

[fujitatomoya]

AFAIK, Fast-DDS takes it just byte code, either int8_t or uint8_t during copy.

https://github.com/ros2/rmw_fastrtps/blob/1dcfe806874956158e1d93a77034c740e8790e19/rmw_fastrtps_shared_cpp/include/rmw_fastrtps_shared_cpp/guid_utils.hpp#L34-L36

[ihasdapie]

AFAIK, Fast-DDS takes it just byte code, either int8_t or uint8_t during copy.

I've dug through a number of rmw implementations and all of them take it bytecode as well. I doubt making this change will break anything for "unofficial" rmws and we can run CI for the official impls

See: #328 (comment)

[ihasdapie]

• Linux
• Linux-aarch64
• Windows

[clalancette]

CI:

• Linux
• Linux-aarch64
• Windows
• Windows Debug