José Augusto Bolina
José Augusto Bolina
@tishun, yes. I'll push what I have in a second commit.
I added a second commit to the PR. It re-works some of the StringMatchResultOutput a little and registers into the OutputRegistry. This change removes the boolean in the constructor and...
Yeah, right on point :dart: This approach in steps makes more sense. We can close this PR and handle the issues in increments. If any of the changes make sense...
+1 on @wburns comment. Can you provide more information on what the `keys` set contains? And also, is this all running locally on the same host? And also, confirm you...
@gleven0822, not at all, we very much appreciate when users report possible issues! Thanks for taking the time to check on your side as well. I'll close this one, but...
The biggest caveat is the time for probing. Even if it identifies the deadlock and rollback one of the transactions, the operation already took a long time (15s, for example)....
I believe the win with the algorithm instead of random timeouts would be to avoid aborting transactions unnecessarily. However, it is much more intricate to implement than random timeouts in...
@pruivo, I've updated the proposal. The idea is to have the deadlock detection running concurrently with the transactions. The transaction does not depend on the deadlock to finish, so the...
Just so I don't forget. I need to add details about partition handling and topology changes. :writing_hand:
CI seems to have related failures.