test: add tests for overflowing arith

[mooori]

Addresses the overflowing part of #912.

Even though it doesn't fully resolve #912, I'd like to put this PR up for review to get feedback on the approach of using a generic function instead of macros.

Generated files in expected/

Should they be generated and committed for these tests? For now I did that to be consistent with existing tests.

For passing tests I did not review these files, assuming matching rust_out and vm_out is sufficient. Though the generated files can be very helpful to get an idea of generated code and inspect it in case of failures. So another option could be writing these files only if specifically requested or when an error occurs.

[greenhat]

I bet the reason is rodata being different due to the file path difference that is in it - https://github.com/0xMiden/compiler/actions/runs/22096960736/job/63856846720?pr=959#step:7:816. We fix it with trim-path cargo option (see example at

compiler/tests/integration/src/compiler_test.rs

Line 580 in 0570d59

trim-paths = ["diagnostics", "object"]

) but in this case you're using CompilerTest::rust_fn_body that calls rustc directly. So you can either find trim-path equivalent in rustc options or switch to CompilerTest::rust_fn_body_with_* that uses cargo.

[greenhat]

The globals are placed after the 'rodata'.

[mooori]

Thanks for the hints! It was due to the path difference in rodata indeed.

With rustc's --remap-path-prefix path prefix I haven't been able to modify the path in the embedded panic message. Using rust_fn_body_with_stdlib_sys sets panic="abort", which solves the problem by removing the panic message. No panics are expected for these tests, so we can benefit from more compact wasm.

[greenhat]

Looking good! Nice job!