DavidSpickett wrote: Yes this is what I was thinking of. Don't go diving into the failures yet, I'll try this on the bot machine first and see if we need other adjustments. https://github.com/llvm/llvm-project/pull/149305