parallel: family_pids tested on PolarHome platforms.

src/parallel

@@ -2439,7 +2439,7 @@ sub reaper {
                 if($opt::halt_on_error) {
                     if($opt::halt_on_error == 1
 			or
-		       ($opt::halt_on_error < 1 and $Global::total_failed > 1 
+		       ($opt::halt_on_error < 1 and $Global::total_failed > 3 
 			and
 			$Global::total_failed / $Global::total_started > $opt::halt_on_error)) {
                         # If halt on error == 1 or --halt 10%
@@ -4771,19 +4771,37 @@ sub kill {
     }
 }
 
-sub family_pids {
+{
+    my %pid_parentpid_cmd;
+
+    sub family_pids {
 	# Find the pids with this->pid as (grand)*parent
-    # TODO test this on different OS as 'ps' is known to be different
 	my $self = shift;
 	my $pid = $self->pid();
+
+       	%pid_parentpid_cmd or %pid_parentpid_cmd = 
+	    (
+	     'aix' => q( ps -ef | awk '{print $2" "$3}' ),
+	     'cygwin' => q( ps -ef | awk '{print $2" "$3}' ),
+	     'dec_osf' => q( ps -ef | awk '{print $2" "$3}' ),
+	     'darwin' => q( ps -o pid,ppid -ax ),
+	     'dragonfly' => q( ps -o pid,ppid -ax ),
+	     'freebsd' => q( ps -o pid,ppid -ax ),
+	     'gnu' => q( ps -ef | awk '{print $2" "$3}' ),
+	     'hpux' => q( ps -ef | awk '{print $2" "$3}' ),
+	     'linux' => q( ps -ef | awk '{print $2" "$3}' ),
+	     'mirbsd' => q( ps -o pid,ppid -ax ),
+	     'netbsd' => q( ps -o pid,ppid -ax ),
+	     'nto' => q( ps -ef | awk '{print $2" "$3}' ),
+	     'openbsd' => q( ps -o pid,ppid -ax ),
+	     'solaris' => q( ps -ef | awk '{print $2" "$3}' ),
+	     'svr5' => q( ps -ef | awk '{print $2" "$3}' ),
+	    );
+	$pid_parentpid_cmd{$^O} or ::die_bug("pid_parentpid_cmd for $^O missing");
+
 	my (@pidtable,%children_of,@pids);
-    if ($^O eq 'linux' or $^O eq 'solaris' or $^O eq 'cygwin') {
-	# Table with pid parentpid (SysV + GNU)
-	@pidtable = `ps -ef | awk '{print \$2" "\$3}'`;
-    } else {
-	# Table with pid parentpid (BSD)
-	@pidtable = `ps -o pid,ppid -ax`;
-    }
+	# Table with pid parentpid
+	@pidtable = `$pid_parentpid_cmd{$^O}`;
 	for (@pidtable) {
 	    /(\S+)\s+(\S+)/ or ::die_bug("pidtable format");
 	    push(@{$children_of{$2}},$1);
@@ -4801,6 +4819,7 @@ sub family_pids {
 	}
 
 	return (@pids);
+    }
 }
 
 sub failed {

src/parallel.pod

@@ -639,11 +639,13 @@ See also: B<--line-buffer> B<--ungroup>
 Print a summary of the options to GNU B<parallel> and exit.
 
 
-=item B<--halt-on-error> <0|1|2>
+=item B<--halt-on-error> I<val> (alpha testing)
 
-=item B<--halt> <0|1|2>
+=item B<--halt> I<val> (alpha testing)
 
-=over 3
+How should GNU B<parallel> terminate if one of more jobs fail?
+
+=over 7
 
 =item Z<>0
 
@@ -661,6 +663,12 @@ last failing job.
 Kill off all jobs immediately and exit without cleanup. The exit
 status will be the exit status from the failing job.
 
+=item Z<>1-99%
+
+If I<val>% of the jobs fail and minimum 3: Do not start new jobs, but
+complete the running jobs including cleanup. The exit status will be
+the exit status from the last failing job.
+
 =back
 
 
@@ -845,9 +853,9 @@ the same syntax as B<--jobs>, so I<100%> for one per CPU is a valid
 setting. Only difference is 0 which is interpreted as 0.01.
 
 
-=item B<--controlmaster> (experimental)
+=item B<--controlmaster>
 
-=item B<-M> (experimental)
+=item B<-M>
 
 Use ssh's ControlMaster to make ssh connections faster. Useful if jobs
 run remote and are very fast to run. This is disabled for sshlogins

src/parallel.texi

@@ -737,11 +737,13 @@ See also: @strong{--line-buffer} @strong{--ungroup}
 
 Print a summary of the options to GNU @strong{parallel} and exit.
 
-@item @strong{--halt-on-error} <0|1|2>
-@anchor{@strong{--halt-on-error} <0|1|2>}
+@item @strong{--halt-on-error} @emph{val} (alpha testing)
+@anchor{@strong{--halt-on-error} @emph{val} (alpha testing)}
 
-@item @strong{--halt} <0|1|2>
-@anchor{@strong{--halt} <0|1|2>}
+@item @strong{--halt} @emph{val} (alpha testing)
+@anchor{@strong{--halt} @emph{val} (alpha testing)}
+
+How should GNU @strong{parallel} terminate if one of more jobs fail?
 
 @table @asis
 @item 0
@@ -763,6 +765,13 @@ last failing job.
 Kill off all jobs immediately and exit without cleanup. The exit
 status will be the exit status from the failing job.
 
+@item 1-99%
+@anchor{1-99%}
+
+If @emph{val}% of the jobs fail: Do not start new jobs, but complete the
+running jobs including cleanup. The exit status will be the exit
+status from the last failing job.
+
 @end table
 
 @item @strong{--header} @emph{regexp}

src/parallel_tutorial.pod

@@ -1170,6 +1170,25 @@ Output:
   parallel: This job failed:
   echo 1; exit 1
 
+If --halt is given a percentage this percentage of the jobs must fail
+(though minimum 3) before GNU Parallel stops spawning more jobs:
+
+  parallel -j2 --halt 20% echo {}\; exit {} ::: 0 0 1 2 3 4 5 6 7
+
+Output:
+
+  0
+  0
+  1
+  2
+  3
+  4
+  parallel: Starting no more jobs. Waiting for 2 jobs to finish. This job failed:
+      echo 4; exit 4
+  5
+  parallel: Starting no more jobs. Waiting for 1 jobs to finish. This job failed:
+      echo 5; exit 5
+
 GNU Parallel can retry the command with --retries. This is useful if a
 command fails for unkown reasons now and then.
 

testsuite/tests-to-run/parallel-polarhome.sh

@@ -1,6 +1,6 @@
 #!/bin/bash
 
-P_ALL="vax freebsd solaris openbsd netbsd debian alpha aix redhat hpux ultrix qnx irix tru64 openindiana suse solaris-x86 mandriva ubuntu scosysv unixware dragonfly centos miros hurd raspberrypi"
+P_ALL="vax freebsd solaris openbsd netbsd debian alpha aix redhat hpux ultrix minix qnx irix tru64 openindiana suse solaris-x86 mandriva ubuntu scosysv unixware dragonfly centos miros hurd raspberrypi"
 P_NOTWORKING="vax alpha openstep"
 P_NOTWORKING_YET="ultrix irix"