[clang] [clang] Match -isysroot behaviour with system compiler on Darwin (PR #80524)

[Louis Dionne via cfe-commits]

ldionne wrote:

If we do that, we’ll just create churn. It’s a moving target.

You will « fix » upstream Clang to match « the system compiler » temporarily, but by doing so you’re causing the downstream Clang to ingest that change too via auto-merging and that means you’ll flip-flop the state of downstream. That is, unless we manually undo the change downstream but then upstream Clang is the one that will eventually need to play catch up with downstream.

IMO that’s just churn. It’s a lot easier and better to let upstream be the canonical version and let downstream play catch up like it normally does. It just sucks that it took so long for this change to make it into downstream clang but that’s a separate story.

https://github.com/llvm/llvm-project/pull/80524