Implement Cgc<T> - a concurrent threadsafe garbage collector #6
Conversation
This update to the Trace trait should make it easier to implement the concurrent garbage collector in parallel to the current threadlocal garbage collector. Unfortunately, it also causes object safety problems due to the type parameter in the _trace method. Further investigation should be done to see if this can be fixed.
|
Is this merge-ready? 😄 GC in Rust seems to be waking up again. |
|
@Manishearth IIRC the biggest problem with it was that it wasn't very well tested... Not sure about any other problems... |
|
We should probably ask the community to poke this crate a bit more |
|
No kidding ^.^ - we should also test it manually a bunch more too. I'd honestly probably feel the most comfortable with letting the concurrent collector bitrot a bit, and focusing on the threadlocal one. It seems the least likely to have any serious problems. (Also I think I remember there was some problem with CGCs inside of Gcs or something like that) |
|
Wouldn't it be more consistent to call it |
|
@torkleyy : this Gc isn't atomic, it's concurrent, so Agc would be misleading. |
|
Is it staled or still a work in progress? |
|
This isn't going to happen any time soon. Closing |
Don't merge this yet - this is a work in progress PR for us to talk about the implementation and potential problems with it.