alexcrichton requested saulecabrera for a review on PR #8243.
alexcrichton requested elliottt for a review on PR #8243.
alexcrichton opened PR #8243 from alexcrichton:migrate-winch-tests
to bytecodealliance:main
:
This commit builds on https://github.com/bytecodealliance/wasmtime/pull/8237 to migrate all Winch filetests to the top-level
tests/disas/
test suite. All files are copied to atests/disas/winch
folder.This PR has a number of commit splitting out the various stages of this migration. The main changes are:
test = "winch"
is now supported indisas
tests- Support for the Winch ABI in Cranelift is ungated (it no longer panics) to get
wasmtime compile
working to get the tests working- Winch tests now use ATT syntax instead of Intel syntax (the Cranelift default)
The next follow-up I'd like to do is to remove the filetest infrastructure for Winch, which will probably also remove the Winch executable itself. I'll do that as a follow-up though.
alexcrichton requested wasmtime-compiler-reviewers for a review on PR #8243.
alexcrichton requested fitzgen for a review on PR #8243.
alexcrichton requested wasmtime-core-reviewers for a review on PR #8243.
elliottt commented on PR #8243:
One change that we made to the winch disasm tests that I liked was to only print offsets after branch instructions (and always after a return). This made updates to filetests easier to read, as the diff was usually more precise. Would that be easy to add back in before this migration, or should we do that as a separate update?
alexcrichton commented on PR #8243:
I saw that yeah from Winch but I ended up leaving it out in favor of printing all offsets. I was a bit confused as to why the address after a jump instruction was printed because that didn't account for fall-through-style jumps (e.g. it didn't reliably catch all jump targets). For after-ret I was also not sure that would handle the case of multiple
ret
instructions in a function.I can add it back in though if desired!
saulecabrera submitted PR review:
Looks good to me, thanks! I'd be in favor of considering adding back the file test formatting; even though not all cases are covered (as you've pointed out) the original motivation that Trevor and I discussed was to make any diffs easier to read.
saulecabrera submitted PR review:
Looks good to me, thanks! I'd be in favor of considering adding back the file test formatting; even though not all cases are covered (as you've pointed out) the original motivation that Trevor and I discussed was to make any diffs easier to read.
saulecabrera created PR review comment:
"The Winch calling convention is only implemented for x86_64 and aarch64"
alexcrichton updated PR #8243.
alexcrichton updated PR #8243.
alexcrichton updated PR #8243.
alexcrichton commented on PR #8243:
Ah ok that makes sense. I've updated with that added back in plus some comments
elliottt submitted PR review.
elliottt commented on PR #8243:
Thanks for changing the output back!
alexcrichton updated PR #8243.
alexcrichton has enabled auto merge for PR #8243.
alexcrichton merged PR #8243.
Last updated: Dec 23 2024 at 12:05 UTC