Messages in this thread |  | | From | Zefram <> | Subject | Re: On SIGCHLD signal semantics | Date | Tue, 18 Jun 1996 04:16:39 +0100 (BST) |
| |
You missed a case. What if a SysV program expects to inherit a SIG_IGN handler? Or if it checks which handler it has, and then expects SysV semantics? More generally, what about programs that assume that SIG_IGN is SIG_IGN, regardless of who set it? There are plenty of library routines out there that will save and restore a signal handler, and calling programs will then often assume that that handler's behaviour remains unchanged.
I think the idea of distinguishing between inherited and explicitly set signal handlers is broken. You can't tell which you have, you can't get back to the inherited state (except by re-execing yourself), and it violates the principle of least surprise.
-zefram
|  |