Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Tue, 21 Sep 2010 11:23:56 -0500

On Tuesday 21 September 2010 05:50:20 Dmitry Teytelman wrote:
>
> > It would be great if you opened a bug report against the Launchpad bug
> > tracker for EPICS Base [1] to report this issue. (And maybe even file a
> > patch if you were able to fix it.) Thanks a lot!
>
> Done (with a patch).

Thanks, I will include an equivalent change in R3.14.12, or maybe something a 
little more radical depending on any answers to the following question:

Can anyone (other than Dmitry!) remember/explain why to ignore the SIGHUP and 
SIGPIPE signals, we install our own null handlers (which then have to call any 
already-installed handler, but that's broken if they specify SA_SIGINFO) 
rather than just setting SIG_IGN if the current handler is SIG_DFL?  I'm 
thinking something like this would be much simpler and seems to work:

epicsShareFunc void epicsShareAPI epicsSignalInstallSigHupIgnore (void)
{
    struct sigaction curAction;
    int status = sigaction(SIGHUP, NULL, &curAction);

    if (status >=0 &&
        curAction.sa_handler == SIG_DFL) {
        curAction.sa_handler = SIG_IGN;
        status = sigaction(SIGHUP, &curAction, NULL);
    }
    if (status < 0) {
        fprintf(stderr, "%s: SIGHUP action failed, %s\n",
            __FILE__, strerror(errno));
    }
}

Note that our existing code is not thread-safe either, this routine is called 
once in iocInit().

- Andrew
-- 
The best FOSS code is written to be read by other humans -- Harald Welte

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System