bug: Need to regenerate InstCombine with elab.async false
Since Lean added asynchronous elaboration, running our evaluation scripts on InstCombine spawns a huge number of lean processes, which saturates our machines and skews our results. Therefore, we turn off async elaboration with set_option elab.async false (https://github.com/opencompl/lean-mlir/pull/1084). This was added ~24 hours pre OOPSLA 2025, and was thus accomplished by find-and-replace.
We elected not to perform the principled solution of regenerating the InstCombine files using our python scripts, as we do not have precisely pinned version of python, xdsl, and llvm-project to accomplish this.
We record that the files need to be regenerated, and we suggest that this regeneration runs as a part of (perhaps nightly) CI, to make sure that these scripts are not stale.
Also, when we regenerate the InstCombine files, consider saving the problems as one-test-case-per-file.
AFAIK this is fixed now. I will close the issue.