> + bool packedp = false; > + > + if (TREE_CODE(to) == COMPONENT_REF > + && (TYPE_PACKED (TREE_TYPE (TREE_OPERAND (to, 0))) > + || (TREE_CODE (TREE_OPERAND (to, 1)) == FIELD_DECL > + && DECL_PACKED (TREE_OPERAND (to, 1))))) > + packedp = true; > > note that this is not a reliable way of determining if a field is packed > (yes, the existing code is broken in the same way). I realize that you > are only using this for diagnostic purposes, still something worth > mentioning.
Indeed, ideally DECL_PACKED shouldn't be used outside stor-layout.c and dwarf2out.c in the middle-end. > I dislike passing around the packedp flag throughout the expander and to > warn inside the expander. First, the flag has a name that can be confusing > (it does not conservatively flag all packed accesses). Second, why don't > you warn for this from inside the frontend when the bitfield access is > generated? This way the diagnostic won't depend on optimization levels > and you avoid uglifying the expander. Seconded. If the -fstrict-volatile-bitfields support is still incomplete, let's take this opportunity to clean it up. Testing DECL_PACKED or issuing a warning from the RTL expander is to be avoided. -- Eric Botcazou
