FUZZER: eval_proc_main crash 0330

[allight]

Describe the bug Fuzzer crash found on 9-23 hash 0330

crashers/crasher_2024-09-23_0330.x

Seems to timeout, possibly in LLVM.

[allight]

This seems to be an IR that llvm just takes an inordinately long time to process...

Setting llvm opt level to 0 results in finishing in <1 second.

[ericastor]

Profiling indicates that LLVM might be getting stuck in the DAGCombiner; it's spending 99.97% of the first two minutes inside DAGCombiner::Run. I might try to dump the LLVM IR & see if I can get the issue reported upstream, especially if I can reduce it at all.