bryancall commented on PR #13541:
URL: https://github.com/apache/trafficserver/pull/13541#issuecomment-5300846775

   Adding the production evidence behind this, since the description referred 
to crash reports without showing one.
   
   Stack as reported, on a 10.0.x-based build:
   
   ```
   EventNotify::signal()
   (anonymous namespace)::AutoStopCont::mainEvent(int, Event*)
   handleEvent
   EThread::process_event(Event*, int, long)
   EThread::execute_regular()
   execute
   EThread::execute()
   spawn_thread_internal
   ```
   
   Two occurrences in a seven day window, on two separate machines.
   
   The attribution is unambiguous: `Log::preproc_notify[i].signal()` is the 
only `EventNotify::signal()` call anywhere in `AutoStopCont::mainEvent`, so a 
fault at that frame from that caller can only be this loop. That is the line 
this PR guards.
   
   Two caveats so this is not read as more than it is. The faulting address was 
not captured for these two events, so I am not claiming a confirmed null 
dereference from this data alone. What the data establishes is that the loop 
faults in production, and inspection establishes that the array can 
legitimately be null there. #13472 described the identical failure mode at the 
two `LogObject.cc` call sites and fixed those.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to