paleolimbot commented on PR #14328: URL: https://github.com/apache/arrow/pull/14328#issuecomment-1292123473
Closing this because none of the experiments fixed anything, although some may be worth persuing separately. Notably, this PR tried: - A "trash can" singleton approach for tracking exec plans that haven't finished yet but that we've called `StopProducing()` on - A `thread_safe_sexp` class which made sure that no deleters called R code (if, for example, the last remaining instance of a shared pointer that held a reference to an `sexp` was deleted on another thread). This never seemed to occur, or at least the warning was never printed in any logs. - Disabling the version of `RunWithCapturedR()` that launches another thread. I remember trying this several months ago too and it also didn't fix anything then. It's worth noting that most of the valgrind jobs timed out, which may be a failure mode (deadlock somewhere) or may just be a result of really slow runners. In any case, the strategies on this PR should be tried one at a time rather than in serial (as they were here). -- 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]
