Sander Saarend

Rebased onto main 👍 Note, my branch also included the changes from #3077 which was already merged but is currently missing in main

These changes on the original 0.17 branch have been verified live in https://lemm.ee - deadlocks are no longer killing the scheduled tasks thread (but deadlocks still occur as expected)

CI test failures seem unrelated to my changes. I think I can't re-run CI myself, is that correct?

Just FYI, you can run pict-rs as a standalone process on any IP and port you want, and configure Lemmy to connect through the `pictrs` block in `lemmy.hjson`.

After further investigation, I am unfortunately starting to think that this issue is not caused by multiple concurrent scheduled task executions (at least not exclusively). I have been running the...

I just received confirmation that I am not the only one experiencing this, and it's also happening on an instance with a single lemmy_server process. So far, I have only...

My current theory is that this deadlock is caused by `Comment::create` here: https://github.com/LemmyNet/lemmy/blob/87aa0acaf167a105b16acbd7d0551b0733502e0d/crates/db_schema/src/impls/comment.rs#L54 This function also updates the `comment_aggregates` table, and it is being invoked constantly even by federated comments

I'm already creating exactly that PR, will submit for review soon

The deadlocks crashing the scheduled tasks thread was fixed in #3090, which is included in 0.18 The deadlocks themselves should be significantly reduced (in practice, on lemm.ee, completely resolved) in...

Trigger changes removed and hot rank calculations added to the respective `receive` functions. Is this what you had in mind, @dessalines ?