There appears to be one additional failure that is causing a lot of
regressions. it has to do with
processing va_lists. The code in
stabilize_va_list (tree valist, int needs_lvalue)
{
tree vatype = targetm.canonical_va_list_type (TREE_TYPE (valist));
gcc_assert (vatype != NULL_TREE);
seems to assert in a number of tests. I assume that this is related
to no quite getting everything in the va type correct. This needs to
be looked into further.
Aside from that there were only the normal merging and drifting issues.
