westonpace commented on a change in pull request #9095:
URL: https://github.com/apache/arrow/pull/9095#discussion_r564862274
##########
File path: cpp/src/arrow/util/task_group.cc
##########
@@ -135,6 +149,18 @@ class ThreadedTaskGroup : public TaskGroup {
// before cv.notify_one() has returned
std::unique_lock<std::mutex> lock(mutex_);
cv_.notify_one();
+ if (completion_future_.has_value()) {
+ // MarkFinished could be slow. We don't want to call it while we are
holding
+ // the lock.
+ // TODO: If optional is thread safe then we can skip this locking
entirely
+ auto future = *completion_future_;
+ auto finished = completion_future_->is_finished();
+ auto status = status_;
Review comment:
Technically we don't set `finished_` when calling `FinishAsync` and so
conceivably a task could be scheduled after we call unlock that fails quickly
and modifies `status_` as we are using it...but we should probably mark
`finished_` when we call `FinishAsync`. I will do that.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
