fix: building codspeed with clang

[not-matthias]

Things I changed:

• rust-bindgen: Invoke manually with the right args to only generate for our functions -> otherwise assertions for unrelated things will fail
• Fallback to noop implementation when build fails (similar to pytest-codspeed)
• Build with optimizations (removes a warning)
• Replaced gnu17 with c11, otherwise it causes compilation errors when not supported.
• Added cargo-cross tests

[codspeed-hq]

CodSpeed Performance Report

Merging #137 will degrade performances by 70%

_{Comparing cod-1483-error-building-codspeed-rust-with-clang (898484d) with main (27c0199)}

Summary

⚡ 26 improvements
❌ 24 (👁 1) regressions
✅ 313 untouched

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Mode	Benchmark	BASE	HEAD	Change
❌	Instrumentation	Iterative	132.5 ns	161.7 ns	-18.04%
⚡	WallTime	Recursive	50.6 µs	45.6 µs	+10.91%
❌	Instrumentation	Iterative[20]	132.5 ns	161.7 ns	-18.04%
❌	Instrumentation	Iterative[21]	134.2 ns	163.3 ns	-17.86%
⚡	WallTime	Recursive[20]	50.5 µs	45.7 µs	+10.45%
⚡	WallTime	Recursive[21]	82 µs	74.5 µs	+10.07%
❌	WallTime	small_drop	50 ns	60 ns	-16.67%
⚡	WallTime	large_setup	18 ns	14 ns	+28.57%
👁	WallTime	iter_with_setup	36 ns	44 ns	-18.18%
❌	WallTime	iter_batched_large_input	36 ns	44 ns	-18.18%
❌	WallTime	iter_batched_per_iteration	36 ns	48 ns	-25%
❌	WallTime	iter_batched_ref_large_input	36 ns	44 ns	-18.18%
❌	WallTime	iter_batched_ref_per_iteration	36 ns	40 ns	-10%
❌	WallTime	iter_with_setup	36 ns	44 ns	-18.18%
❌	WallTime	from_elem[16384]	891 ns	928 ns	-3.99%
⚡	WallTime	from_elem[4096]	371 ns	351 ns	+5.7%
⚡	WallTime	from_elem[8192]	550 ns	520 ns	+5.77%
⚡	WallTime	from_elem_decimal[2048]	261 ns	241 ns	+8.3%
⚡	WallTime	init_array[1000]	2.1 µs	1.9 µs	+12.51%
❌	Instrumentation	fib_10	830.6 ns	859.7 ns	-3.39%
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

[Copilot]

Pull Request Overview

This PR fixes compilation issues when building the codspeed crate with clang by addressing compiler-specific warnings and improving CI testing for different compilers.

• Adds compiler warning suppressions for C code generated from Zig
• Replaces specific tests with build-only validation for macOS and Windows
• Expands CI matrix to test both GCC and Clang compilers on Linux

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
crates/codspeed/build.rs	Adds warning suppression flags and uses clang-compatible flag handling
.github/workflows/ci.yml	Updates CI matrix to test GCC/Clang separately and changes to build-only for unsupported platforms

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

Pull Request Overview

Copilot reviewed 6 out of 7 changed files in this pull request and generated 1 comment.

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[art049]

Looks good but can you add a check in the CI making sure the bindings generated are up to date?

As well maybe a small section explaining how to generate the bindings in the README/contributing?

[not-matthias]

Looks good but can you add a check in the CI making sure the bindings generated are up to date?

As well maybe a small section explaining how to generate the bindings in the README/contributing?

Added a check in the lint step to regenerate the bindings and fail if they are different. I added the script in the same dir as bindings.rs and added a comment in the Rust code. I feel like it doesn't make that much sense to add it to the README/CONTRIBUTING since it's much harder to discover and not really linked to the code.

[art049]

lgtm!

[art049]

If I'm correct @not-matthias this also fixes #139 right?

[not-matthias]

If I'm correct @not-matthias this also fixes #139 right?

Yes, I've added tests for cargo-cross to ensure it works