feat: support risc-v

[lievenhey]

Requires KDAB/perfparser#39 to be merged. Fixes: #705

[lievenhey]

@dendibakh I added some support for risc-v. I can't really test this, but the result looks sane.

[dendibakh]

Thanks @lievenhey! Is there an easy way for me to test this? IIRC, building hotspots from sources is not trivial. But if I can update a few libs and test, then I can test the fix.

[lievenhey]

@dendibakh there is a ci job running which builds an appimage. You can download the current appimage here https://github.com/KDAB/hotspot/actions/runs/14440980603?pr=709

[dendibakh]

Thanks @lievenhey. The appimage is x86, so I can't install it on my P550 board. However, I installed it on a Linux x86 system and opened the profile there. I can see the correct function names, which is good.

I can't see the disassembly for the solution function though -- it shows a blank window.

[lievenhey]

Yeah the disassembly is a little bit funky but I am working on a better solution.

[lievenhey]

I am still not sure if hotspot works with bigger recordings of riscv programs. But I don't have a risc-v board so I can't test it.

[dendibakh]

I see. How hard do you think it would be to add the riscv disasm support to hotspot (assuming you have access to a riscv board)?

Anyway, from my perspective, this patch is already a step forward.

[lievenhey]

We are using objdump for the disassembly. If you have a objdump version which supports riscv you can set that in Settings -> Unwind -> objdump path. Then it should work.

[dendibakh]

I installed riscv binutils with sudo apt install binutils-riscv64-unknown-elf and added /usr/bin/riscv64-unknown-elf-objdump in Settings -> Unwind -> objdump path. Now when I try to show disassembly for a function, the application crashes.

/tmp/.mount_hotspoe5cYDy/AppRun.wrapped: line 6:  8792 Segmentation fault      (core dumped) LD_LIBRARY_PATH="$d/usr/lib":$LD_LIBRARY_PATH "$bin/hotspot" "$@"

[milianw]

Can you get us a backtrace for the crash by any chance? Ideally self-compiled?

[dendibakh]

Can you get us a backtrace for the crash by any chance? Ideally self-compiled?

@milianw , I tried getting a stack trace but it seems the image provided by @lievenhey doesn't have symbols. So, when I run it under gdb it shows no useful data.

[Inferior 1 (process 3227653) exited with code 0213]
Core was generated by `/usr/libexec/ibus-ui-gtk3'.
Program terminated with signal SIGSEGV, Segmentation fault.

I don't know how to build hotspots and I'm afraid it will take me a lot of time to do it.

The crash that I see can be reproduced on an x86 system as I described above.