Frank Steffahn

I’ve made a gist originally, too, but it looks like you can’t see all the files there. (Afaict only 300 of the 401 files show up.) Here’s a repo https://github.com/steffahn/rust_issue_101973...

I don’t really like those relative speedup graphs, since they paint a skewed/biased picture of what’s happening: E.g., if the new implementation is twice as fast, it’ll say +100%, but...

@Voultapher Oh… right, you *did* mention there that it’s symmetric. Maybe it’d be easier to find if it was mentioned inside of the graph, too 😇 E.g. where it currently...

[ @Voultapher by the way, you had tagged the wrong person; careful with using @-sign in e-mail ;-) – also, since you’ve responded by e-mail you might have missed my...

> @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...

> I wonder if the faster sorting could be enabled for a set of trusted types like &str, integers, etc. where we provide the Ord impl so we trust it....

> From my understanding, sorting an array of integers with the builtin Ord implementation is 100% safe. Speeding up just this use case is what I propose. Ah, sorry, I...

> for release builds we document that some implementation defined behaviour happens, for example element duplication, but no UB I think the argument is that *element duplication **is UB***. Even...

The error message behaves funny, too, with the `type must outlive the anonymous lifetime as define here` part. If you change the `impl` to use `add_building>::Output`. If all lifetimes are...

I feel like this whole bullet point might have simply become seriously outdated, because it appears to describe *some* (but not all) of the coercions that now work via `CoerceUnsized`....