https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120929
qinzhao at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |qinzhao at gcc dot gnu.org
Assignee|unassigned at gcc dot gnu.org |qinzhao at gcc dot
gnu.org
--- Comment #2 from qinzhao at gcc dot gnu.org ---
could you please reduce the testing case to help me debug?
I suspect that this relate to the following change in gcc/tree-object-size.cc
of r10-1905-g7165ca43caf470:
/* Handle the following stmt #2 to propagate the size from the
1860 stmt #1 to #3:
1861 1 _1 = .ACCESS_WITH_SIZE (_3, _4, 1, 0, -1, 0B);
1862 2 _5 = *_1;
1863 3 _6 = __builtin_dynamic_object_size (_5, 1);
1864 */
1865 else if (TREE_CODE (rhs) == MEM_REF
1866 && POINTER_TYPE_P (TREE_TYPE (rhs))
1867 && TREE_CODE (TREE_OPERAND (rhs, 0)) == SSA_NAME
1868 && integer_zerop (TREE_OPERAND (rhs, 1)))
1869 reexamine = merge_object_sizes (osi, var, TREE_OPERAND
(rhs, 0));
This change enabled more opportunities to propagate the size information, as a
result, the __builtin_dynamic_object_size might return a valid value now.
Are you sure that the new buffer overflow is a false positive?
