github-actions[bot] commented on code in PR #67053:
URL: https://github.com/apache/doris/pull/67053#discussion_r3841440866


##########
be/src/exec/pipeline/pipeline_fragment_context.cpp:
##########
@@ -2604,18 +2619,23 @@ void 
PipelineFragmentContext::_coordinator_callback(const ReportStatusRequest& r
         LOG_INFO("Going to cancel query {} since report exec status got rpc 
failed: {}",
                  print_id(req.query_id), rpc_status.to_string());

Review Comment:
   The current generation/sequence change does not yet fix this interleaving. 
Both builders capture completion state and counters before calling 
`next_query_statistics_sequence()`: periodic P can capture unfinished S0 and 
pause; final F can capture S1, obtain N, send/ack, and unregister; then P 
obtains N+1 and FE accepts S0 before reaching the terminal guard. The sequence 
must be reserved before reading either completion or counters so it is the 
snapshot's linearization point, with a barrier-controlled test that resumes P 
last.



##########
be/src/runtime/query_context.cpp:
##########
@@ -248,6 +248,13 @@ QueryContext::~QueryContext() {
                 
PrettyPrinter::print_bytes(query_mem_tracker()->peak_consumption()));
     }
     _resource_ctx->task_controller()->finish();
+#ifndef BE_TEST
+    if (_resource_ctx->task_controller()->query_type() == TQueryType::LOAD) {

Review Comment:
   The current call order does not establish the claimed recursive-CTE close 
edge. `FINAL_CLOSE` first calls `listen_wait_close(..., true)`, which 
immediately evaluates `_finish_load_fragment()` and sends without retaining or 
waiting on the guard; only afterward does `rerun_fragment()` call 
`notify_close()`. If the target tasks close later, normal 
`_close_fragment_instance()` calls `_finish_load_fragment()` again for the same 
PFC, so the exact-equality counter can publish early or overshoot permanently. 
Please advance the logical PFC count exactly once after target close and cover 
FINAL_CLOSE racing the last target task.



-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to