https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82737
--- Comment #22 from Martin Liška <marxin at gcc dot gnu.org> --- (In reply to Nathan Sidwell from comment #21) > yeah, I messed up. I had presumed that once set, DECL_ASSEMBLER_NAME does > not change. That is untrue. > In this case it's even worse, in that DECL_ASSEMBLER_NAME is set on a new > decl that duplicate_decl then deletes. This is a latent bug that never > triggered on the old code, but would have left the global namespace pointing > at dead storage. Good, I'm happy you're able to reproduce and to fix that ;)