afonso360 opened PR #7320 from afonso360:interpret-libcalls
to bytecodealliance:main
:
:wave: Hey,
This is a follow up to one of @alexcrichton 's comments in #7203, where he found some fuzzing issues with RISC-V. I'm still investigating those, but this solves one of them.
Fuzzgen can currently generate a number of libcalls, these also need to be given a handler when building the interpreter. We currently have that in the fuzzgen interpreter, but don't have an equivalent in
test interpret
, so when copying a failing clif case from fuzzgen it's possible to get a second error if the testcase references some libcall.I've been currently just stripping out libcalls from failing clif cases, but probably should have addressed this a while ago.
This adds the same libcalls that we have in the fuzzer, it might be worth having a default interpreter implementation in
cranelift-interpreter
that adds these, not entirely sure.
afonso360 requested wasmtime-compiler-reviewers for a review on PR #7320.
afonso360 requested fitzgen for a review on PR #7320.
afonso360 requested wasmtime-default-reviewers for a review on PR #7320.
alexcrichton submitted PR review.
alexcrichton merged PR #7320.
Last updated: Nov 22 2024 at 16:03 UTC