linker-diff for RISC-V

[lapla-cogito]

close #776

[marxin]

I would like to continue working on the RISC-V port somewhen in September. @lapla-cogito Are you planning to work on this PR, please? It's definitely something that will make my life easier.

[lapla-cogito]

Sorry for the delay, I'll work on this in this week.

[marxin]

Sorry for the delay, I'll work on this in this week.

Don't hurry, it's great you've been working on it and don't hesitate to ask me about anything if you feel you're stuck.

[lapla-cogito]

I'm not sure why some tests hang and never finish when I run them with WILD_TEST_CROSS=riscv64. The same thing happens both locally and in the CI environment...

[marxin]

I'm not sure why some tests hang and never finish when I run them with WILD_TEST_CROSS=riscv64.

Based on my initial investigation, it's connected to your changes where you decode the instructions with objdump (linker_diff::utils::decode_insn_with_objdump), So I guess you're either doing too many invocations of the tool, or there's some glitch connected to read/write with the sub-process. Please add some verbose logging that will help you.

[lapla-cogito]

This PR didn't implement relaxation itself, which caused some tests to hang due to repeated calls to the relatively heavy add_diff function within report_section_diffs. I addressed this by checking which sections should be ignored before calling add_diff, preventing unnecessary calls where possible.

[davidlattimore]

Looks good to me. I tried running it on my VF2 and the tests passed - well, except one, but it seems it was already failing. Thanks for your work on this!

Perhaps wait a bit in case @marxin has any comments.

[marxin]

Thank you for working on that @lapla-cogito!