Conversation

Notices

1. Embed this notice
   fasterthanlime 🌌 (fasterthanlime@hachyderm.io)'s status on Sunday, 30-Jun-2024 23:56:51 JST fasterthanlime 🌌

   Found another snag re: dynamic linking Rust code — N copies of `tokio` are present, so that's N distinct thread-locals storing the "current tokio context"

   ...and the binary is calling the library from an executor, but the library doesn't think so.

   https://github.com/tokio-rs/tokio/blob/65d0e08d39f06eba88661d3d6de665b14588f30d/tokio/src/runtime/context.rs#L77-L122

   • Embed this notice
     esmevane, sorry (ironchamber@mastodon.esmevane.com)'s status on Sunday, 30-Jun-2024 23:56:49 JST esmevane, sorry

     in reply to

     • Christian Kruse

     @cjk @fasterthanlime haha, I understand this is a thing Not To Be Done, but I'm super curious about how you did it & what happened. did you write anything up or post code anywhere, maybe?

   • Embed this notice
     Christian Kruse (cjk@chaos.social)'s status on Sunday, 30-Jun-2024 23:56:50 JST Christian Kruse

     in reply to

     @fasterthanlime I once wrote a task runner library with different queues and priorities and all the bells and whistles.

     I spawned a tokio runtime for every queue, with a configurable amount of workers.

     Oh boy, did I regret that. 🫣

   • Embed this notice
     fasterthanlime 🌌 (fasterthanlime@hachyderm.io)'s status on Sunday, 30-Jun-2024 23:56:51 JST fasterthanlime 🌌

     in reply to

     Well, it’s nothing nested tokio runtimes cannot fix 🤡

   • Embed this notice
     esmevane, sorry (ironchamber@mastodon.esmevane.com)'s status on Monday, 01-Jul-2024 02:50:58 JST esmevane, sorry

     in reply to

     • Christian Kruse

     @cjk @fasterthanlime Gotcha! That makes sense, most of my work is just the same. If you do wind up writing anything up one day, I'd love to read it though!

   • Embed this notice
     Christian Kruse (cjk@chaos.social)'s status on Monday, 01-Jul-2024 02:51:00 JST Christian Kruse

     in reply to

     • esmevane, sorry

     @ironchamber @fasterthanlime no, sorry. This was an internal project, and I couldn’t fix the problems, so I just threw it away and wrote a much simpler approach with Redis as a message broker and distinct worker processes. I did not write about it.

     But maybe I should 🤔 the code is still available in the git history.