Lukas Bergdoll

As a starting point this is how I run tests with miri in github actions https://github.com/Voultapher/self_cell/blob/main/.github/workflows/rust.yml

@EdorianDark I'm running some benchmarks tonight, preliminary results suggest it is slightly faster than the existing sort, but significantly slower than new_stable for copy types. From what I can tell,...

Plus I just discovered UB by running my test suite with miri and the proposed `wpwoodjr_stable_sort`.

Regarding the failing test, I'm not sure what the expected behaviour should be. ```rust // Sort using a completely random comparison function. // This will reorder the elements *somehow*, but...

@Frank I fully agree about 'directional' speedup graphs. This one is deliberately symmetric. Which means +100% means the new one is twice as fast as the old one. -100% means...

Indeed I currently have no access to my computer and will take a closer look on Monday. Good point about tagging, I didn't realize it turns into an actual tag,...

@Mark-Simulacrum I just noticed you changed the label from waiting on review to waiting on author. I'm waiting on review, regarding the failing CI, I commented about it here https://github.com/rust-lang/rust/pull/100856#issuecomment-1225999974....

@est31 > Good point, it seems that items are being duplicated and others are silently leaked. With types that impl Drop this might lead to a double free. Please read...

@Mark-Simulacrum thanks for the feedback. > As an aside, it looks like this PR's diff currently is such that it replaces the main body of the implementation too, even if...

@steffahn good points I hand't thought about pointers and in general unsafe code. Arguably the integer overflow comparison holds to some degree, given that offset calculation overflow could cause UB...