http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54876
Bug #: 54876 Summary: [4.8 Regression] LTO bootstrap broken, streaming garbage-collected BLOCK Classification: Unclassified Product: gcc Version: 4.8.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: bootstrap AssignedTo: unassig...@gcc.gnu.org ReportedBy: rgue...@gcc.gnu.org CC: de...@gcc.gnu.org, jamb...@gcc.gnu.org We are streaming a GCed TREE_BLOCK of (gdb) p expr->base.code $5 = MEM_REF (gdb) p expr->base.code $6 = ADDR_EXPR from output_node_opt_summary here: /* At the moment we assume all old trees to be PARM_DECLs, because we have no mechanism to store function local declarations into summaries. */ gcc_assert (parm); streamer_write_uhwi (ob, parm_num); gcc_assert (EXPR_LOCATION (map->new_tree) == UNKNOWN_LOCATION); stream_write_tree (ob, map->new_tree, true); we need to clear all TREE_BLOCKs from this. Not sure if new_tree originates from a tree that eventually comes along prune_expression_for_jf, but static tree prune_expression_for_jf (tree exp) { if (EXPR_P (exp)) { exp = unshare_expr (exp); SET_EXPR_LOCATION (exp, UNKNOWN_LOCATION); } return exp; } is not enough. You need to walk_tree exp, as in this case it is &MEM_REF and all sub-expressions can contain a location.