Missing stdbool.h import is breaking downstream C clients

[banerjs-overland]

Generated by Generative AI

Diagnosed by Claude Opus 4.6

Operating System:

ubuntu 22.04

ROS version or commit hash:

jazzy

RMW implementation (if applicable):

No response

RMW Configuration (if applicable):

No response

Client library (if applicable):

No response

'ros2 doctor --report' output

No response

Steps to reproduce issue

  /* C translation unit that includes rmw/time.h */
  #include "rmw/time.h"

  void test_func(void) {
      rmw_time_t a = {0, 0};
      rmw_time_t b = {0, 0};
      bool result = rmw_time_equal(a, b);
      (void)result;
  }

• Create a minimal colcon workspace with rcutils:jazzy and rmw:jazzy (unfixed) and the above C file in a test C package
• colcon build in the workspace reproduced the above error
• Changing the rmw package to include the fix in this MR allowed the build to succeed.

Expected behavior

Able to build rmw.

Actual behavior

Unable to build rmw

Additional information

I put up a PR with #423 but I'm guessing it's not getting visibility because of a missing bug report? Hoping that this is the missing link. I'm hoping the patch makes it in before a new apt package is released to downstream consumers.

[impala454]

Seeing this in humble as well, notably when building with testing off. I think stdbool.h is getting "accidentally" included by some test.

[impala454]

I did a little sleuthing if anyone cares- it looks like this the chain of commits that broke a transitive include:

In rcutils last week:
ros2/rcutils@d812bf2 - removed rcutils/types.h which in turn included rcutils/types/uint8_array.h, which in turn included rcutiles/allocator.h which includes <stdbool.h>.

[kymikoloco]

@banerjs-overland - Thank you for adding the fix to humble with #428 ! I'm generally ignorant of the particulars of the ros2 ecosystem in terms of building, but is there value in running IWYU to help prevent these types of issues in the future?

Barring that, we pin the repos to humble which can cause a cascade effect, is there a "safer" way to pin/do this, or is this such an infrequent issue that it's an outlier?

[banerjs-overland]

Good questions. I'm going to have to defer an answer to the maintainers of this repo such as @fujitatomoya or @mjcarroll , et al. though. From what I've seen within ROS, this is not unknown but it is relatively rare; that said, I cannot speak to any additional considerations that might be at play when narrowing of headers (such as what happened here) occurs.