On 9/19/22 09:52, Patrick Palka wrote:
It looks like some xtreme-header-* tests are failing after the libstdc++
change r13-2158-g02f6b405f0e9dc ultimately because we're neglecting
to stream PACK_EXPANSION_EXTRA_ARGS, which leads to false equivalences
of different partial instantiations of _TupleConstraints::__constructible.

Tested on x86_64-pc-linux-gnu, does this look OK for trunk?

        PR c++/106761

gcc/cp/ChangeLog:

        * module.cc (trees_out::type_node) <case TYPE_PACK_EXPANSION>:
        Stream PACK_EXPANSION_EXTRA_ARGS.
        (trees_in::tree_node) <case TYPE_PACK_EXPANSION>: Likewise.


Looks good, I wonder why I missed that. (I guess extracting a testcase out of the headers was too tricky?)

nathan
---
  gcc/cp/module.cc | 3 +++
  1 file changed, 3 insertions(+)

diff --git a/gcc/cp/module.cc b/gcc/cp/module.cc
index 1a1ff5be574..9a9ef4e3332 100644
--- a/gcc/cp/module.cc
+++ b/gcc/cp/module.cc
@@ -8922,6 +8922,7 @@ trees_out::type_node (tree type)
        if (streaming_p ())
        u (PACK_EXPANSION_LOCAL_P (type));
        tree_node (PACK_EXPANSION_PARAMETER_PACKS (type));
+      tree_node (PACK_EXPANSION_EXTRA_ARGS (type));
        break;
case TYPENAME_TYPE:
@@ -9455,12 +9456,14 @@ trees_in::tree_node (bool is_use)
            {
              bool local = u ();
              tree param_packs = tree_node ();
+             tree extra_args = tree_node ();
              if (!get_overrun ())
                {
                  tree expn = cxx_make_type (TYPE_PACK_EXPANSION);
                  SET_TYPE_STRUCTURAL_EQUALITY (expn);
                  PACK_EXPANSION_PATTERN (expn) = res;
                  PACK_EXPANSION_PARAMETER_PACKS (expn) = param_packs;
+                 PACK_EXPANSION_EXTRA_ARGS (expn) = extra_args;
                  PACK_EXPANSION_LOCAL_P (expn) = local;
                  res = expn;
                }

--
Nathan Sidwell

Reply via email to