Jonathan Schwender

> It would be very difficult to detect the okay assume_init call such as MaybeUninit -> [MaybeUninit;N] on the library side. It may be difficult if we wanted to detect...

> This is not always correct. The type could be Copy, in which case it remains initialized (and its not possible to check whether its Copy in generic code). I...

Probably benchmarking `production` is also interesting, since that is where the long build-time of script is really noticable, due to `codegen-units = 1`.

You could run `export RUST_LOG=debug` before running servo to get some additional debug messages printed, which might give some hint.

> @jschwe, I noticed you've merged https://github.com/rust-lang/rust-bindgen/pull/3103, and I'm wondering about the usefulness of process_comment in its current form. Shouldn't we invoke all callbacks so they can be layered? Just...

Shouldn't we just bump `MACOSX_DEPLOYMENT_TARGET` in servo? The oldest available version in github CI is macOS-12, so I don't think we can even test at all if servo can actually...

I think this should be resolved after https://github.com/servo/servo/pull/34484, which bumped the target macos version in servo.

If you run cbindgen on the static libs, will cbindgen not generate the bindings if you re-export the public API of your rlib crate? It doesn't make much sense to...

> Adding this to _generator_add_package_targets before the "bin" case fixes the issue for me: I don't think corrosion will support importing `rlib`s before the `rlib` format is stabilized. > cbindgen...

> Not sure how it relates to my case, removing unused/unreachable binary code is totally fine. > Something like libB is not being produced, only binaries. Thats your specific use-case,...