Fix failing unit tests under ARM

[volsa]

This PR does does two things to fix the failing ARM test:

1. It replaces compile_and_run_no_params with compile_and_run for String-Function tests, which would otherwise fail due to invalid pointer access (we previously did something similar)
2. (My understanding of this and it might not be 100% correct) It ignores some #[should_panic] stdlib tests under ARM because of undefined behaviour. More specifically some of our FFI functions call panic!, the current rustc implementation however assumes that any extern "C" function cannot unwind and doing so would trigger undefined behaviour. Consequentially the #[should_panic] attribute isn't able to catch panics / unwind the stack. I did some experiments if we could bypass this somehow (see code below) but while these tests worked under x86_64, they did not under ARM. One potential fix to this would be to not panic inside FFI functions and instead return error codes which would be a somewhat big-ish refactor. Luckily there's a RFC being stabilized, so maybe in a few Rust versions we can just change the function signature of FFIs from pub extern "C" to pub extern "C-unwind", potentially fixing this issue.

TL;DR: Wait for some future Rust version where C-unwind is introduced, in the meantime #[ignore] some tests under ARM.

Some more references I found

• https://internals.rust-lang.org/t/does-extern-c-actually-support-unwind/18002
• https://doc.rust-lang.org/nomicon/ffi.html#rust-panic-with-c-unwind
• https://www.github.com/PyO3/pyo3/issues/766
• https://www.github.com/PyO3/pyo3/issues/492
• https://www.github.com/rust-lang/rust/issues/88622

...and the bypass attempts

#[test]
fn right_string_substring_too_long() {
    let src = r#"
	    FUNCTION main : STRING
	    VAR_TEMP
	        in : STRING;
	    END_VAR
	        in := 'sample text';
	        main := RIGHT(in, 12);
	    END_FUNCTION
        "#;

    // These attempts worked under x86_64, but not under aarch64.
    // 1. Attempt: Run code inside catch_unwind closure and check if the result is an error (= paniced).
    let catch = std::panic::catch_unwind(|| {
        let sources = add_std!(src, "string_functions.st");
        let _: [u8; 81] = compile_and_run_no_params(sources);
    });
    assert!(catch.is_err());

    // 2. Attempt: Spawn a thread, join it and check if the thread result is an error (= paniced).
    let thread = std::thread::spawn(move || {
        let sources = add_std!(src, "string_functions.st");
        let _: [u8; 81] = compile_and_run_no_params(sources);
    });
    assert!(thread.join().is_err());
}

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: +0.02% 🎉

Comparison is base (fe3a9e2) 95.42% compared to head (76c9ef2) 95.45%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #960      +/-   ##
==========================================
+ Coverage   95.42%   95.45%   +0.02%     
==========================================
  Files         152      152              
  Lines       40774    40818      +44     
==========================================
+ Hits        38909    38962      +53     
+ Misses       1865     1856       -9     

Files Changed	Coverage Δ
.../stdlib/tests/date_time_numeric_functions_tests.rs	100.00% <ø> (ø)
libs/stdlib/src/string_functions.rs	98.18% <100.00%> (ø)
libs/stdlib/tests/string_function_tests.rs	97.12% <100.00%> (+0.08%)	⬆️

... and 4 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.