================
@@ -0,0 +1,55 @@
+; RUN: %llc_dwarf -O0 -filetype=obj < %s | llvm-dwarfdump - | FileCheck
--implicit-check-not "{{DW_TAG|NULL}}" %s
+
+; Test that retained unused (unreferenced) types emission.
+
+; Compiled from
+; $ clang -cc1 -debug-info-kind=unused-types test.cpp -emit-llvm
+
+; void test_unused() {
+; struct Y {};
+; {
+; struct X {};
+; }
+; }
+
+; CHECK: DW_TAG_compile_unit
+; CHECK: DW_TAG_subprogram
+; CHECK: DW_AT_name ("test_unused")
+; CHECK: DW_TAG_structure_type
+; CHECK: DW_AT_name ("Y")
+
+; FIXME: here should be DW_TAG_lexical_block as a parent of structure 'X'.
+; But it's not possible to reliably emit a lexical block for which a
LexicalScope
+; wasn't created, so we just fallback to the most close parent DIE
+; (see DwarfCompileUnit::getOrCreateLexicalBlockDIE() for details).
----------------
jmorse wrote:
I think it makes sense to test this behaviour to make sure the sane thing
happens; but do we have any test coverage for DW_TAG_lexical_block's that _do_
have LexicalScope's created? i.e., stick in an argument-less function call
inside the inner lexical block?
https://github.com/llvm/llvm-project/pull/165032
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
