Hi,
We have an issue where MLIR compiled with clang, exposed to a user building their code with gcc fails: [MLIR] Project compiled using `gcc` links incorrectly to `mlir` compiled with `clang` causing invalid memory access. · Issue #62918 · llvm/llvm-project · GitHub
The issue seems to be around ABI incompatibility on lambda captures. Namely clang and gcc seem to capture with a different order.
What’s clang aim in terms of compatibility with gcc? Is this a bug or expected? If expected, is this case (and others?) documented? Do we have “sanitizers” to defend against this (clang-tidy checks maybe)?