rv64_atomic_widths_orders.arm32.skip (483B)
1 This rv64-targeted case exercises 64-bit atomics through `long`, relying on long being 64-bit (LP64) so i64_loc can hold values like 0x100000000. On arm32 ILP32 long is 32-bit: the store of the long long literal 0x100000000L truncates to 0, so the != comparison against the untruncated 64-bit literal fails and the case returns 9 (kit and clang agree). arm32's 32-bit atomic codegen (LDREX/STREX, fetch_add, CAS) is covered by builtin_06..25, which pass on arm32. Same skip as rv32.