:farmer: Flaky `TestGuardCondition.construction_and_destruction` throwing SEH Exception in Windows repeated

[Crola1702]

Bug report

Required Info:

• Operating System:
  • Windows
• Installation type:
  • Source
• Version or commit hash:
  • Rollingo
• Client library (if applicable):
  • rclcpp

Steps to reproduce issue

1. Run a build in nightly_win_rep
2. See TestGuardCondition.construction_and_destruction test failure

Expected behavior

Test should pass

Actual behavior

Test is failing

Additional information

Reference build: https://ci.ros2.org/view/nightly/job/nightly_win_rep/3273/#showFailuresLink

Test regressions:

• projectroot.test.rclcpp.test_guard_condition
• rclcpp.TestGuardCondition.construction_and_destruction

Log output:

[ RUN      ] TestGuardCondition.construction_and_destruction
unknown file: error: SEH exception with code 0xc0000005 thrown in the test body.
Stack trace:


[  FAILED  ] TestGuardCondition.construction_and_destruction (0 ms)

[Crola1702]

Flaky ratio in the last 30 days (Apr 11th):

job_name	last_fail	first_fail	build_count	failure_count	failure_percentage
nightly_win_rep	2024-04-09	2024-03-13	29	11	37.93

[Crola1702]

Flaky ratio in the last 30 days (Jun 18th):

job_name	last_fail	first_fail	build_count	failure_count	failure_percentage
nightly_win_rep	2024-06-16	2024-05-21	25	8	32.0

[ahcorde]

The line that is failing is https://github.com/ros2/rclcpp/blob/rolling/rclcpp/test/rclcpp/test_guard_condition.cpp#L61-L62

    auto mock = mocking_utils::inject_on_return(
      "lib:rclcpp", rcl_guard_condition_fini, RCL_RET_ERROR);

[ahcorde]

Segfault is happening in this line inside the inject_on_return

mock_ = mmk_mock(target_c_str(), mock_type);

[mjcarroll]

Thanks for isolating the area where the segfault is coming from!

is target_c_str() or mock_type null at this point?

[ahcorde]

target_c_str() has some value (lib:rclcpp) but I'm still trying to understand what's inside the mock_type

[ahcorde]

more details:

Code is failing here

mmk_vfprintf_ = (void*) plt_get_real_fn(ctx, vfprintf);

In particular plt_get_real_fn is calling CreateToolhelp32Snapshot() which is returning an INVALID_HANLDE_VALUE and the code abort.

[ahcorde]

The error in particular is The program issued a command but the command lenght is incorrect

[ahcorde]

I made this patch in plt-pe.c

HANDLE snap;
do {
  snap = CreateToolhelp32Snapshot(TH32CS_NSAPMODULE, GetCurrentProcessId());
} while (snap == INVALID_HANDLE_VALUE);

[ahcorde]

Related fix https://github.com/ros2/Mimick/pull/38

[clalancette]

Given what we merged yesterday, I'm going to consider that this one is "fixed", and close this out. If it turns out not to fix the issue, please feel free to reopen.