diff options
Diffstat (limited to '')
| -rw-r--r-- | src/test/codegen/mem-replace-direct-memcpy.rs | 13 |
1 files changed, 1 insertions, 12 deletions
diff --git a/src/test/codegen/mem-replace-direct-memcpy.rs b/src/test/codegen/mem-replace-direct-memcpy.rs index 8095d309f..4318e926e 100644 --- a/src/test/codegen/mem-replace-direct-memcpy.rs +++ b/src/test/codegen/mem-replace-direct-memcpy.rs @@ -1,21 +1,10 @@ -// ignore-test - -// WHY IS THIS TEST BEING IGNORED: -// -// This test depends on characteristics of how the stdlib was compiled, -// namely that sufficient inlining occurred to ensure that the call to -// `std::mem::replace` boils down to just two calls of `llvm.memcpy`. -// -// But the MIR inlining policy is in flux as of 1.64-beta, and the intermittent -// breakage of this test that results is causing problems for people trying to -// do development. - // This test ensures that `mem::replace::<T>` only ever calls `@llvm.memcpy` // with `size_of::<T>()` as the size, and never goes through any wrapper that // may e.g. multiply `size_of::<T>()` with a variable "count" (which is only // known to be `1` after inlining). // compile-flags: -C no-prepopulate-passes -Zinline-mir=no +// ignore-debug: the debug assertions get in the way #![crate_type = "lib"] |