Watch out for the Networking: Implementing RPC chapter! There are myriad errors in the code examples. Why doesn't someone review the code, and perhaps actually try to run it?
For example:
The peer demonstration (between two RPC servers) just won't work. Besides the fact that the code has a typo (new_rpc_server instead of new_server, mentioned in the errata), they each create a listening socket, supposedly to each opposite machine. That *won't* work. Even if they created a listening socket on their respective hosts, they wouldn't know who to contact for the TCP connection. The lower-level Msg library doesn't facilitate a server sending messages to some arbitrary client (as the RPC example implies).
Someone should *really* sit and go through the code.
Want another more explicit example:
BEGIN EXCERPT
sub rpc {
my $conn = shift;
my $subname = shift;
$subname = (caller() . '::' . $subname) unless $subname =~ /:/;
my $gimme = wantarray ? 'a' : 's'; # Array or scalar
my $msg_id = ++$g_msg_id;
my $serialized_msg = freeze ('>', $msg_id, $gimme, @_);
END EXCERPT
Notice that he shifts subname off the param list? Then proceeds to build the $serialized_msg with only the @_ params (NOT INCLUDING THE SUBNAME).
This is quite disasterous when the recevier gets here:
BEGIN EXCERPT
my $instant = thaw($msg);
my ($dir, $id, @args) = @$instant;
my ($result, @results);
if ($dir eq '>') {
# New request message
my $gimme = shift @args;
my $sub_name = shift @args;
try {
no strict 'refs'; # Because we call the subroutine using
# a symbolic reference
if ($gimme eq 'a') { # Want an array back
@results = &{$sub_name} (@args);
} else {
$result = &{$sub_name} (@args);
}
END EXCERPT
Notice again that he tries to shift the subname off the thaw'd @args array... Won't work, will it?!
I knew this already but for your knowledge - Don't trust ANYONE else's code. You'll pay for it with a significant loss of time.
|
