SingleAccretion opened PR #6931 from SingleAccretion:DI-Fixes
to bytecodealliance:main
:
Two changes:
1) Fix a small bug where a constant DWARF "block" wasn't handled in transformation.
2) Fix up instruction offsets after code emission to account for branch removal.See also the commit messages for a bit more detail.
With these two changes, I am able to step through the code generated by the compiler we are working on.
Fixes #3999.
SingleAccretion submitted PR review.
SingleAccretion created PR review comment:
It is interesting to note that
optimize_branches
is invoked after this code one last time and can nominally modify the current buffer pointer, makefunc_body_len
stale, etc. At the same time, for it to remove a branch, that branch needs to point past the end of the function, which doesn't seem possible?
Last updated: Nov 22 2024 at 16:03 UTC