You'd like to delay the reception of a signal, possibly to prevent unpredictable behavior from signals that can interrupt your program at any point.
Use the POSIX module's interface to the
sigprocmask
system call. This is only available if your system is POSIX conformant.
To block a signal around an operation:
use POSIX qw(:signal_h); $sigset = POSIX::SigSet->new(SIGINT); # define the signals to block $old_sigset = POSIX::SigSet->new; # where the old sigmask will be kept unless (defined sigprocmask(SIG_BLOCK, $sigset, $old_sigset)) { die "Could not block SIGINT\n"; }
To unblock:
unless (defined sigprocmask(SIG_UNBLOCK, $old_sigset)) { die "Could not unblock SIGINT\n"; }
The POSIX standard introduced
sigaction
and
sigprocmask
to give you better control over how signals are delivered. The
sigprocmask
function controls delayed delivery of signals and
sigaction
installs handlers. If available, Perl uses
sigaction
when you change
%SIG
.
To use
sigprocmask
, first build a signal set using
POSIX::SigSet->new
. This takes a list of signal numbers. The POSIX module exports functions named after the signals, which return their signal numbers.
use POSIX qw(:signal_h); $sigset = POSIX::SigSet->new( SIGINT, SIGKILL );
Pass the POSIX::SigSet object to
sigprocmask
with the SIG_BLOCK flag to delay signal delivery, SIG_UNBLOCK to restore delivery of the signals, or SIG_SETMASK to block only signals in the POSIX::SigSet. The most paranoid of programmers block signals for a
fork
to prevent a signal handler in the child process being called before Perl can update the child's
$$
variable, its process id. If the signal handler were called immediately and reported
$$
in that handler, it could possibly report its parent's
$$
, not its own. This issue does not arise often.
Your system's sigprocmask (2) manpage (if you have one); the documentation for the standard POSIX module in Chapter 7 of Programming Perl