Changeset 17840 for gsdl/trunk

Timestamp:

2008-11-12T16:18:25+13:00 (15 years ago)

Author:

ak19

Message:

Terminating Wget: Removed last commit's temporary fix. Now this code has gone back to using open3 since that works on both Linux and Windows, avoids the use of 2>&1 (since that launches a subshell on Linux which becomes the perl script's subprocess rather than the wget we're monitoring in order to terminate it). Wget's output is to STDERR which is what we need to monitor and we do that with open3--although the wget output when using open3 comes into its stdout handle instead. But since we are only using one of the two outputstreams of wget, we use the same handle for both and this means the need for extra stream handle variables and the need to close them is now gone.

File:

• gsdl/trunk/perllib/downloaders/WgetDownload.pm (modified) (6 diffs)

gsdl/trunk/perllib/downloaders/WgetDownload.pm

@@ -35 +35 @@
 use strict;
 use Cwd;
-use IPC::Open2;
+use IPC::Open3;
 use IO::Select;
 use IO::Socket;
@@ -199 +199 @@
 }
 
-sub get_childpid {
-    my ($self, $pid) = @_;
-    my $os = $^O;
-    # for windows, we do nothing special
-    if ($os =~ m/mswin/i) {
-    return $pid;
-    }
-    
-    # else $os is macos or linux
-    # This means the wget child process may have been spawned from a subshell
-    # (the real child) that was launched by this perl script. However, we want
-    # the pid of the wget process, since that is what we want to terminate.
-
-    sleep(2); # give it some time to start up, else we'd have finished searching beforehand
-
-    # Look through any processes spawned immediately/soon after the subshell.
-    # Look for any pid greater than the subshell's, but within a certain limit (10)
-    my $child = $pid;
-    for(my $i = 1; $i <= 10; $i++) {
-    $child = $child+$i;
-    if(kill(0, $child)) { # this process exists
-        return $child;
-    }
-    }
-    
-    return $pid; # could not find any process within limit, so return the original process id?
-}
 
 sub useWget
@@ -266 +239 @@
 
     my $wget_file_path = &util::filename_cat($ENV{'GSDLHOME'}, "bin", $ENV{'GSDLOS'}, "wget");
-    $command = "\"$wget_file_path\" $cmdWget 2>&1";
-    # print STDERR "Command is: $command\n";
-    $childpid = open2($chld_out, $chld_in, $command) || die "wget request failed: $!\n";
-    
-    # for linux the wget process starts off as achild of a subshell launched as the child of 
-    # this perl script. We want the pid of the wget process, not that of the subshell. 
-    $childpid = $self->get_childpid($childpid);
+    $command = "\"$wget_file_path\" $cmdWget";
+    #print STDOUT "Command is: $command\n";
+
+    # Wget's output needs to be monitored to find out when it has naturally termianted.
+    # Wget's output is sent to its STDERR so we can't use open2 without doing 2>&1.
+    # On linux, 2>&1 launches a subshell which then launches wget, meaning that killing
+    # the childpid does not kill wget on Linux but the subshell that launched it instead. 
+    # Therefore, we use open3. Though the child process wget sends output only to its stdout, 
+    # using open3 says chld_err is undefined and the output of wget only comes in chld_out(!)
+    # However that may be, it works with open3. But to avoid the confusion of managing and
+    # closing an extra unused handle, a single handle is used instead for both the child's
+    # stderr and stdout.
+
+    # Both open2 and open3 don't return on failure, but raise an exception. The handling
+    # of the exception is described on p.568 of the Perl Cookbook
+    eval { 
+    $childpid = open3($chld_in, $chld_out, $chld_out, $command);
+    };
+    if ($@) {
+    if($@ =~ m/^open3/) {
+        die "open3 failed in $0: $!\n$@\n";     
+    }
+    die "Tried to launch open3 in $0, got unexpected exception: $@";
+    }
 
     my $loop = 1;
@@ -341 +331 @@
                 close($rh);             #close($serverSocket);
                 print $client "Perl terminated wget and is about to exit\n";
+                last;                           # out of inner for loop
             }
             else { # the process may just be starting up, wait
                 sleep(1);
             }
-            last;                           # out of inner for loop
             }
             last;                               # out of foreach loop
@@ -398 +388 @@
 
     my $wget_file_path = &util::filename_cat($ENV{'GSDLHOME'}, "bin", $ENV{'GSDLOS'}, "wget");
-    #my $command = "\"$wget_file_path\" $cmdWget 2>&1 |"; #open(*WIN,$command) || die "wget request failed: $!\n";
-    my $command = "\"$wget_file_path\" $cmdWget 2>&1";
-    # print STDERR "Command is: $command\n";
-    $childpid = open2($chld_out, $chld_in, $command) || die "wget request failed: $!\n";
-
-    # for linux the wget process starts off as achild of a subshell launched as the child of 
-    # this perl script. We want the pid of the wget process, not that of the subshell. 
-    $childpid = $self->get_childpid($childpid);
+    my $command = "\"$wget_file_path\" $cmdWget";
+    #print STDOUT "Command is: $command\n";
+
+    eval {     # see p.568 of Perl Cookbook
+    $childpid = open3($chld_in, $chld_out, $chld_out, $command);
+    };
+    if ($@) {
+    if($@ =~ m/^open3/) {
+        die "open3 failed in $0: $!\n$@\n";     
+    }
+    die "Tried to launch open3 in $0, got unexpected exception: $@";
+    }
 
     my $full_text = "";
@@ -495 +489 @@
                 close($rh);             #close($serverSocket);
                 print $client "Perl terminated wget and is about to exit\n";
+                last;                           # out of inner for loop
             }
             else { # the process may just be starting up, wait
                 sleep(1);
             }
-            last;                           # out of inner for loop
             }
             last;                               # out of foreach loop