Tim Hutt

I think we also shouldn't be checking in binary ELFs anymore either. We finally have some infrastructure for first party tests but I guess if we want something from newer...

I also think we should update the riscv-tests to build from source and then we can update to the latest version. If it requires a recent Clang, so be it....

I'd quite like to get #1262 and then #792 merged first. #1262 could do with a review. It does remove some enums that will be needed in this PR but...

Actually it automatically selected Yes even without `-y` (I guess it uses isatty or something?) and successfully installed from source. The CI action didn't fail - the failure is that...

Yeah I noticed this too. The OCaml emulator doesn't actually support float currently anyway so it's probably not a big deal. This will change when the float implementation is moved...

Ditto VLUXSEGTYPE, VLOXSEGTYPE, maybe other.

What's the motivation here? Adding explicit signal handlers is usually something that should be avoided if at all possible.

Yeah we should avoid touching signal handlers at all costs. They're *extremely* error prone (see Alex's comment for example) and this doesn't seem like a big enough benefit.

So I was thinking 1. Move the struct definitions to C structs: ``` bitfield RVFI_DII_Execution_PacketV2 : bits(512) = { magic : 63 .. 0, // must be set to 'trace-v2'...

Right but we currently cant use `read_CSR()` twice because it isn't idempotent. In fact it's not really well defined what the value that we report via the callback should be....