fix abort in some cases with DDS #13627
Conversation
|
@maloel can you review? the condition for joinable and the actual join actions are not atomic. Currently this issue was also raised by validation as a critical BUG. |
| @@ -72,8 +72,14 @@ class network_adapter_watcher_singleton | |||
| ~network_adapter_watcher_singleton() | |||
| { | |||
| _adapter_watcher.reset(); // signal the thread to finish | |||
| if( _th.joinable() ) | |||
| _th.join(); | |||
| try { | |||
There was a problem hiding this comment.
moved the opening bracket to a new line
| } | ||
| catch (std::system_error&) | ||
| { | ||
| std::cerr << "join failed, thread might no longer be joinable"; |
There was a problem hiding this comment.
And, if there's an error, there's an error message too, no?
| } | ||
| catch (std::system_error& e) | ||
| { | ||
| LOG_ERROR("join failed, thread might no longer be joinable, Error:" << e.what()); |
There was a problem hiding this comment.
Not sure this will work.
Can you try to see if it works and you see the e.what()?
You might need to use rsutils::from() <<
There was a problem hiding this comment.
LOG_ERROR has <<, if that was the concern
There was a problem hiding this comment.
I would ask that we make error messages a little nicer, though: capitalization, e.g.:
LOG_ERROR( "Network adapter watcher termination failed: " << e.what() );
@Nir-Az up to you whether this is an error the user should see - we might want to leave it debug?
There was a problem hiding this comment.
Sure why not, @AviaAv can you handle please?
Fixed an issue where we randomly see an error where abort() is called after we finish using DDS (in the d'tor), added a try catch block due to the race condition there
Tracked on: [LRS-1212]