Question about rmw_wait.cpp

[yingflower]

rmw_fastrtps/rmw_fastrtps_shared_cpp/src/rmw_wait.cpp

Line 287 in d98cace

if (changed_statuses.is_active(

I have some questions about this code. If has_triggered_condition in line 149 return true, and changed_statuses.is_active in line 289 return false, this thread will in running state forever. Is this a bug?

[fujitatomoya]

@yingflower i do not understand what you are trying to say as possible bug here? can you share more details of sequence?

[yingflower]

I encountered a bug where, in certain scenarios, a thread's CPU usage would spike to nearly 100%. After investigation, I traced it to the file rmw_fastrtps/rmw_fastrtps_shared_cpp/src/rmw_wait.cpp. The issue arises because skip_wait on line 149 is set to true, but on line 287, changed_statuses.is_active() returns false, causing the status_condition state not to be cleared. As a result, every subsequent loop iteration follows the same processing flow, preventing the current thread from ever entering sleep.

[fujitatomoya]

thanks for sharing that. could you share reproducible self-contained sample to make that happen?

[yingflower]

Sorry, I discovered this issue in a complex system and currently don't know how to simplify the reproduction method into a simple example. But I can use GDB to dump the values of some variables. The internal values of changed_statuses are as follows:

‘’‘
(gdb) p changed_statuses
$7 = {<std::bitset<16>> = std::bitset = {[10] = 1, [14] = 1}, }
’‘’

If you need to see the values of other variables, I can dump them for you.