Commit `f78380b936`

LemonBoy <thatlemon@gmail.com>

2020-10-16 12:15:46

stage1: Don't ask LLVM to emit misaligned memcpy

Pay close attention to the RHS type alignment when rendering an assignment op as it may differ from the LHS pointer one. This problem was noticed when debugging a CI failure in #6648: due to sheer luck the misalignment caused a segfault on macos that was also reproduced locally. I tried to write a small test case but it turned out to be a daunting task as I couldn't manage to trigger the problem consistently (and stop the optimizer from simplifying everything). Patches welcome.

master

1 parent 2a62d4b

Changed files (1)

src

stage1

codegen.cpp

@@ -1645,11 +1645,13 @@ static void gen_assign_raw(CodeGen *g, LLVMValueRef ptr, ZigType *ptr_type,
 
         ZigType *usize = g->builtin_types.entry_usize;
         uint64_t size_bytes = LLVMStoreSizeOfType(g->target_data_ref, get_llvm_type(g, child_type));
-        uint64_t align_bytes = get_ptr_align(g, ptr_type);
+        uint64_t src_align_bytes = get_abi_alignment(g, child_type);
+        uint64_t dest_align_bytes = get_ptr_align(g, ptr_type);
         assert(size_bytes > 0);
-        assert(align_bytes > 0);
+        assert(src_align_bytes > 0);
+        assert(dest_align_bytes > 0);
 
-        ZigLLVMBuildMemCpy(g->builder, dest_ptr, align_bytes, src_ptr, align_bytes,
+        ZigLLVMBuildMemCpy(g->builder, dest_ptr, dest_align_bytes, src_ptr, src_align_bytes,
                 LLVMConstInt(usize->llvm_type, size_bytes, false),
                 ptr_type->data.pointer.is_volatile);
         return;

Commit f78380b936

Commit `f78380b936`