Hi! <cut>
Thanks for the patch and for finding the bug!
However, Serg already removed the thread count handling from do_handle_bootstrap. I will today check if the problem still exists and if yes, fix it by always sending the broadcast. (A quick look suggest it still exist).
Just noticed that you had pushed your patch into the main 10.2 tree. I checked it and think that's the right way to fix this. Thanks also for the great comment in the patch that explains it! I also agree that's a bit strange that read_init_file is using the bootstrap() to read the init file. This is probably perfectly safe, but it would have been good if the comment for the bootstrap() function would have said so! Will add this next time I touch the code in this area. Regards, Monty