TimNN added a comment.
Herald added a subscriber: StephenFan.
The ThinLTO related breakage I mentioned above should be fixed as of
https://github.com/llvm/llvm-project/commit/451799bb8261bde52bbfef226d019caf1d82aa42.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
TimNN added a comment.
I'm sorry for the noise. Further investigation has shown that this happens when
Rust is doing (thin) LTO, and I don't think this patch can be considered in any
way "at fault" here, so this is the last you'll hear from me on the topic here.
I don't know whether the fault
TimNN added a comment.
I didn't manage to repro with `opt`, so still no compilable IR. I did some more
debugging, though:
- Inside `removeDeadArgumentsFromCallers`, `CB->getCalledFunction()->dump()`
(after the modification) is `define void
TimNN added a comment.
I've included excerpts from the IR below. It will take me a bit to provide
something compilable. Though you are right, the `noundef` did indeed get
removed from the `call`.
*** IR Dump Before DeadArgumentEliminationPass on [module] ***
; Function Attrs: nonlazybind
TimNN added a comment.
I'm still trying to properly minimize this, but this definitely interacts badly
with other optimizations (which triggers the Rust CI failure I mentioned above):
- We start with two functions, `outer` and `inner`. `outer` calls `inner`.
`outer` has a `noundef` argument
TimNN added a comment.
This also broke Rust when compiled at LLVM head:
https://buildkite.com/llvm-project/rust-llvm-integrate-prototype/builds/13166#0182fb4a-0f2d-4f2e-830f-f62b463b8d48.
(I don't know whether this was some existing UB that got only exposed by this
patch or not, but wanted to
