Jeff King <peff@xxxxxxxx> writes： > ...we are ultimately relying on refs_warn_dangling_symrefs(), so we'd > have to update its interface. And we also reuse the list (here, after > your patch, but already in remote.c) to pass to refs_delete_refs(). So > probably not worth it. This patch only adds sorting code to prune_refs(), and as far as I can tell, prune_refs() is only called once during git fetch. So I was just wondering, would it be problematic if we moved the string_list_sort() into refs_warn_dangling_symref() instead? And if it turns out to be safe, could we perhaps even use strset in refs_warn_dangling_symref()?
