Re: [PATCH 04/15] Hurd signals: refactor check_pending_signals

[Jeremie Koenig]

On Sun, Jul 03, 2011 at 09:30:01PM +0200, Jeremie Koenig wrote:
> |   if (act != ignore &&          /* Signals ignored now are forgotten now.  
> */
> |       __sigismember (&ss->blocked, signo) ||
> |       (signo != SIGKILL && _hurd_stopped))
> |     { 
> |       mark_pending ();
> |       act = ignore;
> |     }
> (...)
> 
> So the patch probably changes behaviour only in corner cases, if at all.

Hmm, actually the following modified example does terminate:

        signal(SIGINT, SIG_IGN);
        sighold(SIGINT);
        raise(SIGSTOP);

        $ kill -INT %1
        $ fg

        signal(SIGINT, SIG_DFL);
        sigrelse(SIGINT);
        /* interruped here */

        printf("foo\n");

In my interpretation this behaviour is still compiliant, but maybe I
should reintroduce the check to make it more consistent, or shuffle the
parenthesis in the code snippet above:

|   if (act != ignore &&          /* Signals ignored now are forgotten now.  */
|       (__sigismember (&ss->blocked, signo) ||
|        signo != SIGKILL && _hurd_stopped))

-- 
Jeremie Koenig <jk@jk.fr.eu.org>
http://jk.fr.eu.org