Update bindgen

[hannesweisbach]

Updating for a solution for issue rust-lang/rust-bindgen#2312 of rust-bindgen.

I'm in the process to generate bindings for softdevice 332, when I encountered the error

+ for s in s332
+ ./nrf-softdevice-gen/target/release/nrf-softdevice-gen ./softdevice/s332/headers ./nrf-softdevice-s332/src/bindings.rs
thread 'main' panicked at /home/hannesweisbach/.cargo/registry/src/index.crates.io-6f17d22bba15001f/bindgen-0.55.1/src/ir/context.rs:838:9:
"ble_gap_evt_timeout_t_union_(unnamed_at__/tmp/ble_gap_h_1196_3)" is not a valid Ident
stack backtrace:

This SO post https://stackoverflow.com/questions/76443280/rust-bindgen-causes-a-is-not-a-valid-ident-error-on-build pointed me to rust-lang/rust-bindgen#2312.

The version bump solves the error above.

[hannesweisbach]

Edit: changed from 0.71.1 down to 0.70.0, because 0.71.1 would generate stray "unsafe" keywords. Not sure what's up. 0.70.0 seems to work though.

[alexmoon]

It looks like bindings generated with the new version of bindgen are not backwards-compatible. If you want to update bindgen, please run gen.sh so that the repository remains consistent and reproducible and make sure that all tests pass (by running ci.sh). Ideally you can get it to work by just adjusting the bindgen configuration.

[hannesweisbach]

Thanks for your feedback!

So not quite as straight-forward as I hoped. I want to mention that I get these types of warnings when I run gen.sh:

+ ./nrf-softdevice-gen/target/release/nrf-softdevice-gen ./softdevice/s140/headers ./nrf-softdevice-s140/src/bindings.rs
Warning: can't set `imports_granularity = Module`, unstable features are only available in nightly channel.
Warning: can't set `group_imports = StdExternalCrate`, unstable features are only available in nightly channel.
Warning: can't set `imports_granularity = Module`, unstable features are only available in nightly channel.
Warning: can't set `group_imports = StdExternalCrate`, unstable features are only available in nightly channel.
+ rustfmt ./nrf-softdevice-s140/src/bindings.rs
Warning: can't set `imports_granularity = Module`, unstable features are only available in nightly channel.
Warning: can't set `group_imports = StdExternalCrate`, unstable features are only available in nightly channel.
+ cd nrf-softdevice-s140
+ cargo build --target thumbv7em-none-eabihf

Do you know if they are expected/benign?

The new generated bindings do not have as nice formatted docs as the old ones. Instead of multiple lines there is just one #[doc]macro with embedded newlines. There is also an issue open rust-lang/rust-bindgen#3073 regarding this issue. As a stop-gap there is a set of rustfmt options to re-format the comments in multiline but with /// instead of #[doc]. Would that be acceptable?

The new bindgen version introduced _bitfield_align_1-members in structs which required to be initialized manually, so I added initializers in nrf-softdevice/ and examples/. ./ci.sh executes now without errors. I'm still investigating whether there is a setting for bindgen to control generation of these members.

[alexmoon]

Do you know if they are expected/benign?

Yes, those warnings are fine. We use some rustfmt options that are only on nightly.

The new generated bindings do not have as nice formatted docs as the old ones. Instead of multiple lines there is just one #[doc]macro with embedded newlines. There is also an issue open rust-lang/rust-bindgen#3073 regarding this issue. As a stop-gap there is a set of rustfmt options to re-format the comments in multiline but with /// instead of #[doc]. Would that be acceptable?

#[doc] macros with embedded newlines are exactly identical to multiple #[doc] macros as far as rustdoc is concerned, so that's fine.

The new bindgen version introduced _bitfield_align_1-members in structs which required to be initialized manually, so I added initializers in nrf-softdevice/ and examples/. ./ci.sh executes now without errors. I'm still investigating whether there is a setting for bindgen to control generation of these members.

This is more of a problem, because it breaks existing code. In most cases it's also useless because the bitfield is byte aligned anyway. Let me know what you find.