https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108880
--- Comment #14 from Marek Polacek <mpolacek at gcc dot gnu.org> --- (In reply to Jakub Jelinek from comment #13) > (In reply to Marek Polacek from comment #12) > > Sure, it worked for the testcase because the STATEMENT_LIST only had two > > stmts. I'm testing: > > > > --- a/gcc/c-family/c-gimplify.cc > > +++ b/gcc/c-family/c-gimplify.cc > > @@ -516,7 +516,8 @@ c_genericize_control_stmt (tree *stmt_p, int > > *walk_subtrees, void *data, > > tree t = tsi_stmt (i); > > if (TREE_CODE (t) != DEBUG_BEGIN_STMT && nondebug_stmts < 2) > > nondebug_stmts++; > > - walk_tree_1 (tsi_stmt_ptr (i), func, data, NULL, lh); > > + walk_tree_1 (tsi_stmt_ptr (i), func, data, > > + static_cast<hash_set<tree> *>(data), lh); > > I'd limit this change to !c_dialect_cxx () only, I'm afraid if it is done > for C++ too, then cp_genericize_r won't then walk into those trees later on > and could avoid replacing there something important. While for C, there is > walk_tree solely for the purposes of c_genericize_control* and nothing else. > To avoid testing it in every iteration you could have a hash_set<tree> *pset > temporary initialized based on c_dialect_cxx () to NULL or data and then > just use it in the loop. Yeah, in C++ I get a crash in check_complete_insertion anyway. I don't really see why but best to limit it to C.