Repository navigation
has_unique_object_representations returns false for classes whose base has only inherited members #327
Description
Discussed in #290
Originally posted by jakub-labenski October 13, 2026
Issue Summary
type_has_unique_object_representations (folding.c) treats a direct base class as empty when its own field_list is NULL, ignoring the data inherited from its bases. Such a base contributes 0 bytes to the computed layout, so the tail-padding check fails and the trait wrongly evaluates to false.
Observed Behavior
Reproducer:
//remark:type_has_unique_object_representations
//options:--g++ --gnu_version=150100;fp
struct A { char i; };
struct B : A {};
struct C : B {};
static_assert(__has_unique_object_representations(B)); // passes
static_assert(__has_unique_object_representations(C)); // fails, expected to pass
edg-docker-test and then edg-test-run-diff returns:
--- Expected Output (sandbox/test.sft.rt.tar.gz -- edg_x86_64)
+++ Observed Output (sandbox/test.sft.rt.tar.gz -- edg_x86_64)
# fe_only -DTEST_NUMBER=1 --g++ --gnu_version=150100 Test_name.c
@@ -0,0 +1,5 @@
+"Test_name.c", line 7: error: static assertion failed
+ static_assert(__has_unique_object_representations(C)); // fails, expected to pass
+ ^
+
+1 error detected in the compilation of "Test_name.c".
Expected Behavior
Code should compile correctly.
Additional Details
No response
I acknowledge that:
- I have checked open issues and existing discussions before filing this report.
Activity
fix(folding): correct unique object rep checks [GH edgcpp#327]
fix(folding): correct unique object rep checks [GH edgcpp#327]
fix(folding): correct unique object rep checks [GH edgcpp#327]
Fix __has_unique_object_representations for several cases [GH edgcpp#327
Fix __has_unique_object_representations for several cases [GH edgcpp#327
Sync upstream main (ad4cfc4) into multibackend and reconcile build/ho…
mohitmishra786 commentedon Nov 14, 2026
@wchilders-nvidia @nv-cmeerw this looks fixed on main by 98ad882 (#326), and
__has_unique_object_representations(C)from the report is now true. Could this be closed?