> Btw, would be really nice to have libbacktrace support for ada ...
OK, I'll keep that in mind.
> While the patch looks technically ok I think you'll run into the same issue
> with a non-zero offset MEM_REF as that will get you a POINTER_PLUS_EXPR
> from build_fold_addr_expr. We might be lucky not to ICE in
> recompute_tree_invariant_for_addr_expr because we can access operand
> zero of that of course. I think recompute_tree_invariant_for_addr_expr
> misses an assert that it receives an ADDR_EXPR and the gimplify.c
> caller would need to handle POINTER_PLUS_EXPR specially.
>
> Or change your patch to also handle non-zero offset MEM_REFs by
> simply gimplifying to POINTER_PLUS_EXPR op0, op1.
I couldn't cover the new case though, because you need a record with variable
size and an array of those yields a non-constant offset so no MEM_REF and a
record with fixed offset doesn't yield a MEM_REF either for some reason...
But I can add the assert in recompute_tree_invariant_for_addr_expr:
Index: tree.c
===================================================================
--- tree.c (revision 228794)
+++ tree.c (working copy)
@@ -4248,6 +4248,8 @@ recompute_tree_invariant_for_addr_expr (
tree node;
bool tc = true, se = false;
+ gcc_assert (TREE_CODE (t) == ADDR_EXPR);
+
/* We started out assuming this address is both invariant and constant, but
does not have side effects. Now go down any handled components and see if
any of them involve offsets that are either non-constant or non-invariant.
--
Eric Botcazou
