[ViewVC] Diff of: cvs/AnyEvent/lib/AnyEvent/Handle.pm

Comparing AnyEvent/lib/AnyEvent/Handle.pm (file contents):
Revision 1.182 by root, Thu Sep 3 12:35:01 2009 UTC vs.
Revision 1.239 by root, Tue Dec 10 20:39:12 2013 UTC

 =head1 NAME
-AnyEvent::Handle - non-blocking I/O on file handles via AnyEvent
+AnyEvent::Handle - non-blocking I/O on streaming handles via AnyEvent
 =head1 SYNOPSIS
    use AnyEvent;
    use AnyEvent::Handle;
    my $hdl; $hdl = new AnyEvent::Handle
       fh => \*STDIN,
       on_error => sub {
          my ($hdl, $fatal, $msg) = @_;
-         warn "got error $msg\n";
+         AE::log error => $msg;
          $hdl->destroy;
          $cv->send;
-      );
+      };
    # send some request line
    $hdl->push_write ("getinfo\015\012");
    # read the response line
    $hdl->push_read (line => sub {
       my ($hdl, $line) = @_;
-      warn "got line <$line>\n";
+      say "got line <$line>";
       $cv->send;
    });
    $cv->recv;
 =head1 DESCRIPTION
-This module is a helper module to make it easier to do event-based I/O on
+This is a helper module to make it easier to do event-based I/O on
-filehandles.
+stream-based filehandles (sockets, pipes, and other stream things).
 The L<AnyEvent::Intro> tutorial contains some well-documented
 AnyEvent::Handle examples.
-In the following, when the documentation refers to of "bytes" then this
+In the following, where the documentation refers to "bytes", it means
-means characters. As sysread and syswrite are used for all I/O, their
+characters. As sysread and syswrite are used for all I/O, their
 treatment of characters applies to this module as well.
 At the very minimum, you should specify C<fh> or C<connect>, and the
 C<on_error> callback.
 use AnyEvent (); BEGIN { AnyEvent::common_sense }
 use AnyEvent::Util qw(WSAEWOULDBLOCK);
 our $VERSION = $AnyEvent::VERSION;
+sub _load_func($) {
+   my $func = $_[0];
+   unless (defined &$func) {
+      my $pkg = $func;
+      do {
+         $pkg =~ s/::[^:]+$//
+            or return;
+         eval "require $pkg";
+      } until defined &$func;
+   }
+   \&$func
+}
+sub MAX_READ_SIZE() { 131072 }
 =head1 METHODS
 =over 4
-=item $handle = B<new> AnyEvent::TLS fh => $filehandle, key => value...
+=item $handle = B<new> AnyEvent::Handle fh => $filehandle, key => value...
 The constructor supports these arguments (all as C<< key => value >> pairs).
 =over 4
 =over 4
 =item on_prepare => $cb->($handle)
 This (rarely used) callback is called before a new connection is
-attempted, but after the file handle has been created. It could be used to
+attempted, but after the file handle has been created (you can access that
+file handle via C<< $handle->{fh} >>). It could be used to prepare the
-prepare the file handle with parameters required for the actual connect
+file handle with parameters required for the actual connect (as opposed to
-(as opposed to settings that can be changed when the connection is already
+settings that can be changed when the connection is already established).
-established).
 The return value of this callback should be the connect timeout value in
-seconds (or C<0>, or C<undef>, or the empty list, to indicate the default
+seconds (or C<0>, or C<undef>, or the empty list, to indicate that the
-timeout is to be used).
+default timeout is to be used).
 =item on_connect => $cb->($handle, $host, $port, $retry->())
 This callback is called when a connection has been successfully established.
-The actual numeric host and port (the socket peername) are passed as
+The peer's numeric host and port (the socket peername) are passed as
-parameters, together with a retry callback.
+parameters, together with a retry callback. At the time it is called the
+read and write queues, EOF status, TLS status and similar properties of
+the handle will have been reset.
+It is not allowed to use the read or write queues while the handle object
+is connecting.
-When, for some reason, the handle is not acceptable, then calling
+If, for some reason, the handle is not acceptable, calling C<$retry> will
-C<$retry> will continue with the next conenction target (in case of
+continue with the next connection target (in case of multi-homed hosts or
-multi-homed hosts or SRV records there can be multiple connection
+SRV records there can be multiple connection endpoints). The C<$retry>
-endpoints). When it is called then the read and write queues, eof status,
+callback can be invoked after the connect callback returns, i.e. one can
-tls status and similar properties of the handle are being reset.
+start a handshake and then decide to retry with the next host if the
+handshake fails.
-In most cases, ignoring the C<$retry> parameter is the way to go.
+In most cases, you should ignore the C<$retry> parameter.
 =item on_connect_error => $cb->($handle, $message)
-This callback is called when the conenction could not be
+This callback is called when the connection could not be
 established. C<$!> will contain the relevant error code, and C<$message> a
 message describing it (usually the same as C<"$!">).
 If this callback isn't specified, then C<on_error> will be called with a
 fatal error instead.
 =item on_error => $cb->($handle, $fatal, $message)
 This is the error callback, which is called when, well, some error
 occured, such as not being able to resolve the hostname, failure to
-connect or a read error.
+connect, or a read error.
 Some errors are fatal (which is indicated by C<$fatal> being true). On
 fatal errors the handle object will be destroyed (by a call to C<< ->
 destroy >>) after invoking the error callback (which means you are free to
 examine the handle object). Examples of fatal errors are an EOF condition
-with active (but unsatisifable) read watchers (C<EPIPE>) or I/O errors. In
+with active (but unsatisfiable) read watchers (C<EPIPE>) or I/O errors. In
-cases where the other side can close the connection at their will it is
+cases where the other side can close the connection at will, it is
 often easiest to not report C<EPIPE> errors in this callback.
 AnyEvent::Handle tries to find an appropriate error code for you to check
-against, but in some cases (TLS errors), this does not work well. It is
+against, but in some cases (TLS errors), this does not work well.
-recommended to always output the C<$message> argument in human-readable
-error messages (it's usually the same as C<"$!">).
+If you report the error to the user, it is recommended to always output
+the C<$message> argument in human-readable error messages (you don't need
+to report C<"$!"> if you report C<$message>).
+If you want to react programmatically to the error, then looking at C<$!>
+and comparing it against some of the documented C<Errno> values is usually
+better than looking at the C<$message>.
-Non-fatal errors can be retried by simply returning, but it is recommended
+Non-fatal errors can be retried by returning, but it is recommended
 to simply ignore this parameter and instead abondon the handle object
 when this callback is invoked. Examples of non-fatal errors are timeouts
 C<ETIMEDOUT>) or badly-formatted data (C<EBADMSG>).
-On callback entrance, the value of C<$!> contains the operating system
+On entry to the callback, the value of C<$!> contains the operating
-error code (or C<ENOSPC>, C<EPIPE>, C<ETIMEDOUT>, C<EBADMSG> or
+system error code (or C<ENOSPC>, C<EPIPE>, C<ETIMEDOUT>, C<EBADMSG> or
 C<EPROTO>).
 While not mandatory, it is I<highly> recommended to set this callback, as
-you will not be notified of errors otherwise. The default simply calls
+you will not be notified of errors otherwise. The default just calls
 C<croak>.
 =item on_read => $cb->($handle)
 This sets the default read callback, which is called when data arrives
 To access (and remove data from) the read buffer, use the C<< ->rbuf >>
 method or access the C<< $handle->{rbuf} >> member directly. Note that you
 must not enlarge or modify the read buffer, you can only remove data at
 the beginning from it.
+You can also call C<< ->push_read (...) >> or any other function that
+modifies the read queue. Or do both. Or ...
-When an EOF condition is detected then AnyEvent::Handle will first try to
+When an EOF condition is detected, AnyEvent::Handle will first try to
 feed all the remaining data to the queued callbacks and C<on_read> before
 calling the C<on_eof> callback. If no progress can be made, then a fatal
 error will be raised (with C<$!> set to C<EPIPE>).
 Note that, unlike requests in the read queue, an C<on_read> callback
 If an EOF condition has been detected but no C<on_eof> callback has been
 set, then a fatal error will be raised with C<$!> set to <0>.
 =item on_drain => $cb->($handle)
-This sets the callback that is called when the write buffer becomes empty
+This sets the callback that is called once when the write buffer becomes
-(or when the callback is set and the buffer is empty already).
+empty (and immediately when the handle object is created).
 To append to the write buffer, use the C<< ->push_write >> method.
 This callback is useful when you don't want to put all of your write data
 into the queue at once, for example, when you want to write the contents
 many seconds pass without a successful read or write on the underlying
 file handle (or a call to C<timeout_reset>), the C<on_timeout> callback
 will be invoked (and if that one is missing, a non-fatal C<ETIMEDOUT>
 error will be raised).
-There are three variants of the timeouts that work fully independent
+There are three variants of the timeouts that work independently of each
-of each other, for both read and write, just read, and just write:
+other, for both read and write (triggered when nothing was read I<OR>
+written), just read (triggered when nothing was read), and just write:
 C<timeout>, C<rtimeout> and C<wtimeout>, with corresponding callbacks
 C<on_timeout>, C<on_rtimeout> and C<on_wtimeout>, and reset functions
 C<timeout_reset>, C<rtimeout_reset>, and C<wtimeout_reset>.
-Note that timeout processing is also active when you currently do not have
+Note that timeout processing is active even when you do not have any
-any outstanding read or write requests: If you plan to keep the connection
+outstanding read or write requests: If you plan to keep the connection
-idle then you should disable the timout temporarily or ignore the timeout
+idle then you should disable the timeout temporarily or ignore the
-in the C<on_timeout> callback, in which case AnyEvent::Handle will simply
+timeout in the corresponding C<on_timeout> callback, in which case
-restart the timeout.
+AnyEvent::Handle will simply restart the timeout.
-Zero (the default) disables this timeout.
+Zero (the default) disables the corresponding timeout.
 =item on_timeout => $cb->($handle)
+=item on_rtimeout => $cb->($handle)
+=item on_wtimeout => $cb->($handle)
 Called whenever the inactivity timeout passes. If you return from this
 callback, then the timeout will be reset as if some activity had happened,
 so this condition is not fatal in any way.
 be configured to accept only so-and-so much data that it cannot act on
 (for example, when expecting a line, an attacker could send an unlimited
 amount of data without a callback ever being called as long as the line
 isn't finished).
+=item wbuf_max => <bytes>
+If defined, then a fatal error will be raised (with C<$!> set to C<ENOSPC>)
+when the write buffer ever (strictly) exceeds this size. This is useful to
+avoid some forms of denial-of-service attacks.
+Although the units of this parameter is bytes, this is the I<raw> number
+of bytes not yet accepted by the kernel. This can make a difference when
+you e.g. use TLS, as TLS typically makes your write data larger (but it
+can also make it smaller due to compression).
+As an example of when this limit is useful, take a chat server that sends
+chat messages to a client. If the client does not read those in a timely
+manner then the send buffer in the server would grow unbounded.
 =item autocork => <boolean>
-When disabled (the default), then C<push_write> will try to immediately
+When disabled (the default), C<push_write> will try to immediately
-write the data to the handle, if possible. This avoids having to register
+write the data to the handle if possible. This avoids having to register
 a write watcher and wait for the next event loop iteration, but can
 be inefficient if you write multiple small chunks (on the wire, this
 disadvantage is usually avoided by your kernel's nagle algorithm, see
 C<no_delay>, but this option can save costly syscalls).
-When enabled, then writes will always be queued till the next event loop
+When enabled, writes will always be queued till the next event loop
 iteration. This is efficient when you do many small writes per iteration,
 but less efficient when you do a single write only per iteration (or when
 the write buffer often is full). It also increases write latency.
 =item no_delay => <boolean>
 the Nagle algorithm, and usually it is beneficial.
 In some situations you want as low a delay as possible, which can be
 accomplishd by setting this option to a true value.
-The default is your opertaing system's default behaviour (most likely
+The default is your operating system's default behaviour (most likely
-enabled), this option explicitly enables or disables it, if possible.
+enabled). This option explicitly enables or disables it, if possible.
 =item keepalive => <boolean>
 Enables (default disable) the SO_KEEPALIVE option on the stream socket:
 normally, TCP connections have no time-out once established, so TCP
-conenctions, once established, can stay alive forever even when the other
+connections, once established, can stay alive forever even when the other
 side has long gone. TCP keepalives are a cheap way to take down long-lived
-TCP connections whent he other side becomes unreachable. While the default
+TCP connections when the other side becomes unreachable. While the default
 is OS-dependent, TCP keepalives usually kick in after around two hours,
 and, if the other side doesn't reply, take down the TCP connection some 10
 to 15 minutes later.
 It is harmless to specify this option for file handles that do not support
 BSD majorly fucked up the implementation of TCP urgent data. The result
 is that almost no OS implements TCP according to the specs, and every OS
 implements it slightly differently.
-If you want to handle TCP urgent data, then setting this flag gives you
+If you want to handle TCP urgent data, then setting this flag (the default
-the most portable way of getting urgent data, by putting it into the
+is enabled) gives you the most portable way of getting urgent data, by
-stream.
+putting it into the stream.
+Since BSD emulation of OOB data on top of TCP's urgent data can have
+security implications, AnyEvent::Handle sets this flag automatically
+unless explicitly specified. Note that setting this flag after
+establishing a connection I<may> be a bit too late (data loss could
+already have occured on BSD systems), but at least it will protect you
+from most attacks.
 =item read_size => <bytes>
-The default read block size (the amount of bytes this module will
+The initial read block size, the number of bytes this module will try
-try to read during each loop iteration, which affects memory
+to read during each loop iteration. Each handle object will consume
-requirements). Default: C<8192>.
+at least this amount of memory for the read buffer as well, so when
+handling many connections watch out for memory requirements). See also
+C<max_read_size>. Default: C<2048>.
+=item max_read_size => <bytes>
+The maximum read buffer size used by the dynamic adjustment
+algorithm: Each time AnyEvent::Handle can read C<read_size> bytes in
+one go it will double C<read_size> up to the maximum given by this
+option. Default: C<131072> or C<read_size>, whichever is higher.
 =item low_water_mark => <bytes>
-Sets the amount of bytes (default: C<0>) that make up an "empty" write
+Sets the number of bytes (default: C<0>) that make up an "empty" write
-buffer: If the write reaches this size or gets even samller it is
+buffer: If the buffer reaches this size or gets even samller it is
 considered empty.
 Sometimes it can be beneficial (for performance reasons) to add data to
 the write buffer before it is fully drained, but this is a rare case, as
 the operating system kernel usually buffers data as well, so the default
 is good in almost all cases.
 =item linger => <seconds>
-If non-zero (default: C<3600>), then the destructor of the
+If this is non-zero (default: C<3600>), the destructor of the
 AnyEvent::Handle object will check whether there is still outstanding
 write data and will install a watcher that will write this data to the
 socket. No errors will be reported (this mostly matches how the operating
 system treats outstanding data at socket close time).
 A string used to identify the remote site - usually the DNS hostname
 (I<not> IDN!) used to create the connection, rarely the IP address.
 Apart from being useful in error messages, this string is also used in TLS
 peername verification (see C<verify_peername> in L<AnyEvent::TLS>). This
-verification will be skipped when C<peername> is not specified or
+verification will be skipped when C<peername> is not specified or is
 C<undef>.
 =item tls => "accept" | "connect" | Net::SSLeay::SSL object
 When this parameter is given, it enables TLS (SSL) mode, that means
-AnyEvent will start a TLS handshake as soon as the conenction has been
+AnyEvent will start a TLS handshake as soon as the connection has been
 established and will transparently encrypt/decrypt data afterwards.
 All TLS protocol errors will be signalled as C<EPROTO>, with an
 appropriate error message.
 TLS mode requires Net::SSLeay to be installed (it will be loaded
 automatically when you try to create a TLS handle): this module doesn't
 have a dependency on that module, so if your module requires it, you have
-to add the dependency yourself.
+to add the dependency yourself. If Net::SSLeay cannot be loaded or is too
+old, you get an C<EPROTO> error.
 Unlike TCP, TLS has a server and client side: for the TLS server side, use
 C<accept>, and for the TLS client side of a connection, use C<connect>
 mode.
 B<IMPORTANT:> since Net::SSLeay "objects" are really only integers,
 passing in the wrong integer will lead to certain crash. This most often
 happens when one uses a stylish C<< tls => 1 >> and is surprised about the
 segmentation fault.
-See the C<< ->starttls >> method for when need to start TLS negotiation later.
+Use the C<< ->starttls >> method if you need to start TLS negotiation later.
 =item tls_ctx => $anyevent_tls
 Use the given C<AnyEvent::TLS> object to create the new TLS connection
-(unless a connection object was specified directly). If this parameter is
+(unless a connection object was specified directly). If this
-missing, then AnyEvent::Handle will use C<AnyEvent::Handle::TLS_CTX>.
+parameter is missing (or C<undef>), then AnyEvent::Handle will use
+C<AnyEvent::Handle::TLS_CTX>.
 Instead of an object, you can also specify a hash reference with C<< key
 => value >> pairs. Those will be passed to L<AnyEvent::TLS> to create a
 new TLS context object.
 TLS handshake failures will not cause C<on_error> to be invoked when this
 callback is in effect, instead, the error message will be passed to C<on_starttls>.
 Without this callback, handshake failures lead to C<on_error> being
-called, as normal.
+called as usual.
-Note that you cannot call C<starttls> right again in this callback. If you
+Note that you cannot just call C<starttls> again in this callback. If you
 need to do that, start an zero-second timer instead whose callback can
 then call C<< ->starttls >> again.
 =item on_stoptls => $cb->($handle)
 callback.
 This callback will only be called on TLS shutdowns, not when the
 underlying handle signals EOF.
-=item json => JSON or JSON::XS object
+=item json => L<JSON> or L<JSON::XS> object
 This is the json coder object used by the C<json> read and write types.
 If you don't supply it, then AnyEvent::Handle will create and use a
 suitable one (on demand), which will write and expect UTF-8 encoded JSON
 texts.
-Note that you are responsible to depend on the JSON module if you want to
+Note that you are responsible to depend on the L<JSON> module if you want
-use this functionality, as AnyEvent does not have a dependency itself.
+to use this functionality, as AnyEvent does not have a dependency on it
+itself.
+=item cbor => L<CBOR::XS> object
+This is the cbor coder object used by the C<cbor> read and write types.
+If you don't supply it, then AnyEvent::Handle will create and use a
+suitable one (on demand), which will write CBOR without using extensions,
+if possible. texts.
+Note that you are responsible to depend on the L<CBOR::XS> module if you
+want to use this functionality, as AnyEvent does not have a dependency on
+it itself.
 =back
 =cut
                $self->{connect}[0],
                $self->{connect}[1],
                sub {
                   my ($fh, $host, $port, $retry) = @_;
+                  delete $self->{_connect}; # no longer needed
                   if ($fh) {
                      $self->{fh} = $fh;
                      delete $self->{_skip_drain_rbuf};
                      $self->_start;
                             });
                   } else {
                      if ($self->{on_connect_error}) {
                         $self->{on_connect_error}($self, "$!");
-                        $self->destroy;
+                        $self->destroy if $self;
                      } else {
                         $self->_error ($!, 1);
                      }
                   }
                },
                sub {
                   local $self->{fh} = $_[0];
                   $self->{on_prepare}
-                     ?  $self->{on_prepare}->($self)
+                     ? $self->{on_prepare}->($self)
                      : ()
                }
             );
       }
 }
 sub _start {
    my ($self) = @_;
+   # too many clueless people try to use udp and similar sockets
+   # with AnyEvent::Handle, do them a favour.
+   my $type = getsockopt $self->{fh}, Socket::SOL_SOCKET (), Socket::SO_TYPE ();
+   Carp::croak "AnyEvent::Handle: only stream sockets supported, anything else will NOT work!"
+      if Socket::SOCK_STREAM () != (unpack "I", $type) && defined $type;
    AnyEvent::Util::fh_nonblocking $self->{fh}, 1;
    $self->{_activity}  =
    $self->{_ractivity} =
    $self->{_wactivity} = AE::now;
+   $self->{read_size} ||= 2048;
+   $self->{max_read_size} = $self->{read_size}
+      if $self->{read_size} > ($self->{max_read_size} || MAX_READ_SIZE);
    $self->timeout   (delete $self->{timeout}  ) if $self->{timeout};
    $self->rtimeout  (delete $self->{rtimeout} ) if $self->{rtimeout};
    $self->wtimeout  (delete $self->{wtimeout} ) if $self->{wtimeout};
-   $self->no_delay  (delete $self->{no_delay} ) if exists $self->{no_delay};
+   $self->no_delay  (delete $self->{no_delay} ) if exists $self->{no_delay}  && $self->{no_delay};
-   $self->keepalive (delete $self->{keepalive}) if exists $self->{keepalive};
+   $self->keepalive (delete $self->{keepalive}) if exists $self->{keepalive} && $self->{keepalive};
-   $self->oobinline (delete $self->{oobinline}) if exists $self->{oobinline};
+   $self->oobinline (exists $self->{oobinline} ? delete $self->{oobinline} : 1);
    $self->starttls  (delete $self->{tls}, delete $self->{tls_ctx})
       if $self->{tls};
-   $self->on_drain  (delete $self->{on_drain}) if $self->{on_drain};
+   $self->on_drain  (delete $self->{on_drain} ) if $self->{on_drain};
    $self->start_read
       if $self->{on_read} || @{ $self->{_queue} };
    $self->_drain_wbuf;
    $message ||= "$!";
    if ($self->{on_error}) {
       $self->{on_error}($self, $fatal, $message);
       $self->destroy if $fatal;
-   } elsif ($self->{fh}) {
+   } elsif ($self->{fh} || $self->{connect}) {
       $self->destroy;
       Carp::croak "AnyEvent::Handle uncaught error: $message";
    }
 }
 =cut
 sub no_delay {
    $_[0]{no_delay} = $_[1];
-   eval {
-      local $SIG{__DIE__};
-      setsockopt $_[0]{fh}, Socket::IPPROTO_TCP (), Socket::TCP_NODELAY (), int $_[1]
+   setsockopt $_[0]{fh}, Socket::IPPROTO_TCP (), Socket::TCP_NODELAY (), int $_[1]
-         if $_[0]{fh};
+      if $_[0]{fh};
-   };
 }
 =item $handle->keepalive ($boolean)
 Enables or disables the C<keepalive> setting (see constructor argument of
 Replace the current C<on_stoptls> callback (see the C<on_stoptls> constructor argument).
 =cut
-sub on_starttls {
+sub on_stoptls {
    $_[0]{on_stoptls} = $_[1];
 }
 =item $handle->rbuf_max ($max_octets)
 Configures the C<rbuf_max> setting (C<undef> disables it).
+=item $handle->wbuf_max ($max_octets)
+Configures the C<wbuf_max> setting (C<undef> disables it).
 =cut
 sub rbuf_max {
    $_[0]{rbuf_max} = $_[1];
 }
+sub wbuf_max {
+   $_[0]{wbuf_max} = $_[1];
+}
 #############################################################################
 =item $handle->timeout ($seconds)
 =item $handle->rtimeout ($seconds)
 =item $handle->wtimeout ($seconds)
 Configures (or disables) the inactivity timeout.
+The timeout will be checked instantly, so this method might destroy the
+handle before it returns.
 =item $handle->timeout_reset
 =item $handle->rtimeout_reset
       $_[0]{$on_timeout} = $_[1];
    };
    *$timeout = sub {
       my ($self, $new_value) = @_;
+      $new_value >= 0
+         or Carp::croak "AnyEvent::Handle->$timeout called with negative timeout ($new_value), caught";
       $self->{$timeout} = $new_value;
       delete $self->{$tw}; &$cb;
    };
 The write queue is very simple: you can add data to its end, and
 AnyEvent::Handle will automatically try to get rid of it for you.
 When data could be written and the write buffer is shorter then the low
-water mark, the C<on_drain> callback will be invoked.
+water mark, the C<on_drain> callback will be invoked once.
 =over 4
 =item $handle->on_drain ($cb)
 Sets the C<on_drain> callback or clears it (see the description of
 C<on_drain> in the constructor).
+This method may invoke callbacks (and therefore the handle might be
+destroyed after it returns).
 =cut
 sub on_drain {
    my ($self, $cb) = @_;
       if $cb && $self->{low_water_mark} >= (length $self->{wbuf}) + (length $self->{_tls_wbuf});
 }
 =item $handle->push_write ($data)
-Queues the given scalar to be written. You can push as much data as you
+Queues the given scalar to be written. You can push as much data as
-want (only limited by the available memory), as C<AnyEvent::Handle>
+you want (only limited by the available memory and C<wbuf_max>), as
-buffers it independently of the kernel.
+C<AnyEvent::Handle> buffers it independently of the kernel.
+This method may invoke callbacks (and therefore the handle might be
+destroyed after it returns).
 =cut
 sub _drain_wbuf {
    my ($self) = @_;
       $cb->() unless $self->{autocork};
       # if still data left in wbuf, we need to poll
       $self->{_ww} = AE::io $self->{fh}, 1, $cb
          if length $self->{wbuf};
+      if (
+         defined $self->{wbuf_max}
+         && $self->{wbuf_max} < length $self->{wbuf}
+      ) {
+         $self->_error (Errno::ENOSPC, 1), return;
+      }
    };
 }
 our %WH;
+# deprecated
 sub register_write_type($$) {
    $WH{$_[0]} = $_[1];
 }
 sub push_write {
    my $self = shift;
    if (@_ > 1) {
       my $type = shift;
+      @_ = ($WH{$type} ||= _load_func "$type\::anyevent_write_type"
-      @_ = ($WH{$type} or Carp::croak "unsupported type passed to AnyEvent::Handle::push_write")
+            or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::push_write")
            ->($self, @_);
    }
+   # we downgrade here to avoid hard-to-track-down bugs,
+   # and diagnose the problem earlier and better.
    if ($self->{tls}) {
-      $self->{_tls_wbuf} .= $_[0];
+      utf8::downgrade $self->{_tls_wbuf} .= $_[0];
       &_dotls ($self)    if $self->{fh};
    } else {
-      $self->{wbuf}      .= $_[0];
+      utf8::downgrade $self->{wbuf}      .= $_[0];
       $self->_drain_wbuf if $self->{fh};
    }
 }
 =item $handle->push_write (type => @args)
-Instead of formatting your data yourself, you can also let this module do
+Instead of formatting your data yourself, you can also let this module
-the job by specifying a type and type-specific arguments.
+do the job by specifying a type and type-specific arguments. You
+can also specify the (fully qualified) name of a package, in which
+case AnyEvent tries to load the package and then expects to find the
+C<anyevent_write_type> function inside (see "custom write types", below).
 Predefined types are (if you have ideas for additional types, feel free to
 drop by and tell us):
 =over 4
 The generated JSON text is guaranteed not to contain any newlines: While
 this module doesn't need delimiters after or between JSON texts to be
 able to read them, many other languages depend on that.
-A simple RPC protocol that interoperates easily with others is to send
+A simple RPC protocol that interoperates easily with other languages is
-JSON arrays (or objects, although arrays are usually the better choice as
+to send JSON arrays (or objects, although arrays are usually the better
-they mimic how function argument passing works) and a newline after each
+choice as they mimic how function argument passing works) and a newline
-JSON text:
+after each JSON text:
    $handle->push_write (json => ["method", "arg1", "arg2"]); # whatever
    $handle->push_write ("\012");
 An AnyEvent::Handle receiver would simply use the C<json> read type and
    $handle->push_read (json => sub { my $array = $_[1]; ... });
 Other languages could read single lines terminated by a newline and pass
 this line into their JSON decoder of choice.
+=item cbor => $perl_scalar
+Encodes the given scalar into a CBOR value. Unless you provide your own
+L<CBOR::XS> object, this means it will be encoded to a CBOR string not
+using any extensions, if possible.
+CBOR values are self-delimiting, so you can write CBOR at one end of
+a handle and read them at the other end without using any additional
+framing.
+A simple nd very very fast RPC protocol that interoperates with
+other languages is to send CBOR and receive CBOR values (arrays are
+recommended):
+   $handle->push_write (cbor => ["method", "arg1", "arg2"]); # whatever
+An AnyEvent::Handle receiver would simply use the C<cbor> read type:
+   $handle->push_read (cbor => sub { my $array = $_[1]; ... });
 =cut
 sub json_coder() {
    eval { require JSON::XS; JSON::XS->new->utf8 }
       || do { require JSON; JSON->new->utf8 }
 }
 register_write_type json => sub {
    my ($self, $ref) = @_;
-   my $json = $self->{json} ||= json_coder;
+   ($self->{json} ||= json_coder)
-   $json->encode ($ref)
+      ->encode ($ref)
+};
+sub cbor_coder() {
+   require CBOR::XS;
+   CBOR::XS->new
+}
+register_write_type cbor => sub {
+   my ($self, $scalar) = @_;
+   ($self->{cbor} ||= cbor_coder)
+      ->encode ($scalar)
 };
 =item storable => $reference
 Freezes the given reference using L<Storable> and writes it to the
 =cut
 register_write_type storable => sub {
    my ($self, $ref) = @_;
-   require Storable;
+   require Storable unless $Storable::VERSION;
    pack "w/a*", Storable::nfreeze ($ref)
 };
 =back
 before it was actually written. One way to do that is to replace your
 C<on_drain> handler by a callback that shuts down the socket (and set
 C<low_water_mark> to C<0>). This method is a shorthand for just that, and
 replaces the C<on_drain> callback with:
-   sub { shutdown $_[0]{fh}, 1 }    # for push_shutdown
+   sub { shutdown $_[0]{fh}, 1 }
 This simply shuts down the write side and signals an EOF condition to the
 the peer.
 You can rely on the normal read queue and C<on_eof> handling
 afterwards. This is the cleanest way to close a connection.
+This method may invoke callbacks (and therefore the handle might be
+destroyed after it returns).
 =cut
 sub push_shutdown {
    my ($self) = @_;
    delete $self->{low_water_mark};
    $self->on_drain (sub { shutdown $_[0]{fh}, 1 });
 }
-=item AnyEvent::Handle::register_write_type type => $coderef->($handle, @args)
+=item custom write types - Package::anyevent_write_type $handle, @args
-This function (not method) lets you add your own types to C<push_write>.
+Instead of one of the predefined types, you can also specify the name of
+a package. AnyEvent will try to load the package and then expects to find
+a function named C<anyevent_write_type> inside. If it isn't found, it
+progressively tries to load the parent package until it either finds the
+function (good) or runs out of packages (bad).
-Whenever the given C<type> is used, C<push_write> will invoke the code
+Whenever the given C<type> is used, C<push_write> will the function with
-reference with the handle object and the remaining arguments.
+the handle object and the remaining arguments.
-The code reference is supposed to return a single octet string that will
+The function is supposed to return a single octet string that will be
-be appended to the write buffer.
+appended to the write buffer, so you can mentally treat this function as a
+"arguments to on-the-wire-format" converter.
-Note that this is a function, and all types registered this way will be
+Example: implement a custom write type C<join> that joins the remaining
-global, so try to use unique names.
+arguments using the first one.
+   $handle->push_write (My::Type => " ", 1,2,3);
+   # uses the following package, which can be defined in the "My::Type" or in
+   # the "My" modules to be auto-loaded, or just about anywhere when the
+   # My::Type::anyevent_write_type is defined before invoking it.
+   package My::Type;
+   sub anyevent_write_type {
+      my ($handle, $delim, @args) = @_;
+      join $delim, @args
+   }
 =cut
 #############################################################################
 ways, the "simple" way, using only C<on_read> and the "complex" way, using
 a queue.
 In the simple case, you just install an C<on_read> callback and whenever
 new data arrives, it will be called. You can then remove some data (if
-enough is there) from the read buffer (C<< $handle->rbuf >>). Or you cna
+enough is there) from the read buffer (C<< $handle->rbuf >>). Or you can
 leave the data there if you want to accumulate more (e.g. when only a
-partial message has been received so far).
+partial message has been received so far), or change the read queue with
+e.g. C<push_read>.
 In the more complex case, you want to queue multiple callbacks. In this
 case, AnyEvent::Handle will call the first queued callback each time new
-data arrives (also the first time it is queued) and removes it when it has
+data arrives (also the first time it is queued) and remove it when it has
 done its job (see C<push_read>, below).
 This way you can, for example, push three line-reads, followed by reading
 a chunk of data, and AnyEvent::Handle will execute them in order.
 This replaces the currently set C<on_read> callback, or clears it (when
 the new callback is C<undef>). See the description of C<on_read> in the
 constructor.
+This method may invoke callbacks (and therefore the handle might be
+destroyed after it returns).
 =cut
 sub on_read {
    my ($self, $cb) = @_;
    $self->_drain_rbuf if $cb;
 }
 =item $handle->rbuf
-Returns the read buffer (as a modifiable lvalue).
+Returns the read buffer (as a modifiable lvalue). You can also access the
+read buffer directly as the C<< ->{rbuf} >> member, if you want (this is
+much faster, and no less clean).
-You can access the read buffer directly as the C<< ->{rbuf} >>
+The only operation allowed on the read buffer (apart from looking at it)
-member, if you want. However, the only operation allowed on the
+is removing data from its beginning. Otherwise modifying or appending to
-read buffer (apart from looking at it) is removing data from its
+it is not allowed and will lead to hard-to-track-down bugs.
-beginning. Otherwise modifying or appending to it is not allowed and will
-lead to hard-to-track-down bugs.
-NOTE: The read buffer should only be used or modified if the C<on_read>,
+NOTE: The read buffer should only be used or modified in the C<on_read>
-C<push_read> or C<unshift_read> methods are used. The other read methods
+callback or when C<push_read> or C<unshift_read> are used with a single
-automatically manage the read buffer.
+callback (i.e. untyped). Typed C<push_read> and C<unshift_read> methods
+will manage the read buffer on their own.
 =cut
 sub rbuf : lvalue {
    $_[0]{rbuf}
 If enough data was available, then the callback must remove all data it is
 interested in (which can be none at all) and return a true value. After returning
 true, it will be removed from the queue.
+These methods may invoke callbacks (and therefore the handle might be
+destroyed after it returns).
 =cut
 our %RH;
 sub register_read_type($$) {
    my $cb = pop;
    if (@_) {
       my $type = shift;
+      $cb = ($RH{$type} ||= _load_func "$type\::anyevent_read_type"
-      $cb = ($RH{$type} or Carp::croak "unsupported type passed to AnyEvent::Handle::push_read")
+             or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::push_read")
             ->($self, $cb, @_);
    }
    push @{ $self->{_queue} }, $cb;
    $self->_drain_rbuf;
    my $cb = pop;
    if (@_) {
       my $type = shift;
+      $cb = ($RH{$type} ||= _load_func "$type\::anyevent_read_type"
-      $cb = ($RH{$type} or Carp::croak "unsupported type passed to AnyEvent::Handle::unshift_read")
+             or Carp::croak "unsupported/unloadable type '$type' passed to AnyEvent::Handle::unshift_read")
             ->($self, $cb, @_);
    }
    unshift @{ $self->{_queue} }, $cb;
    $self->_drain_rbuf;
 =item $handle->unshift_read (type => @args, $cb)
 Instead of providing a callback that parses the data itself you can chose
 between a number of predefined parsing formats, for chunks of data, lines
-etc.
+etc. You can also specify the (fully qualified) name of a package, in
+which case AnyEvent tries to load the package and then expects to find the
+C<anyevent_read_type> function inside (see "custom read types", below).
 Predefined types are (if you have ideas for additional types, feel free to
 drop by and tell us):
 =over 4
 data.
 Example: read 2 bytes.
    $handle->push_read (chunk => 2, sub {
-      warn "yay ", unpack "H*", $_[1];
+      say "yay " . unpack "H*", $_[1];
    });
 =cut
 register_read_type chunk => sub {
 register_read_type line => sub {
    my ($self, $cb, $eol) = @_;
    if (@_ < 3) {
-      # this is more than twice as fast as the generic code below
+      # this is faster then the generic code below
       sub {
-         $_[0]{rbuf} =~ s/^([^\015\012]*)(\015?\012)// or return;
+         (my $pos = index $_[0]{rbuf}, "\012") >= 0
+            or return;
+         (my $str = substr $_[0]{rbuf}, 0, $pos + 1, "") =~ s/(\015?\012)\Z// or die;
-         $cb->($_[0], $1, $2);
+         $cb->($_[0], $str, "$1");
          1
       }
    } else {
       $eol = quotemeta $eol unless ref $eol;
       $eol = qr|^(.*?)($eol)|s;
       sub {
          $_[0]{rbuf} =~ s/$eol// or return;
-         $cb->($_[0], $1, $2);
+         $cb->($_[0], "$1", "$2");
          1
       }
    }
 };
 the receive buffer when neither C<$accept> nor C<$reject> match,
 and everything preceding and including the match will be accepted
 unconditionally. This is useful to skip large amounts of data that you
 know cannot be matched, so that the C<$accept> or C<$reject> regex do not
 have to start matching from the beginning. This is purely an optimisation
-and is usually worth only when you expect more than a few kilobytes.
+and is usually worth it only when you expect more than a few kilobytes.
 Example: expect a http header, which ends at C<\015\012\015\012>. Since we
-expect the header to be very large (it isn't in practise, but...), we use
+expect the header to be very large (it isn't in practice, but...), we use
 a skip regex to skip initial portions. The skip regex is tricky in that
 it only accepts something not ending in either \015 or \012, as these are
 required for the accept regex.
    $handle->push_read (regex =>
    sub {
       # accept
       if ($$rbuf =~ $accept) {
          $data .= substr $$rbuf, 0, $+[0], "";
-         $cb->($self, $data);
+         $cb->($_[0], $data);
          return 1;
       }
       # reject
       if ($reject && $$rbuf =~ $reject) {
-         $self->_error (Errno::EBADMSG);
+         $_[0]->_error (Errno::EBADMSG);
       }
       # skip
       if ($skip && $$rbuf =~ $skip) {
          $data .= substr $$rbuf, 0, $+[0], "";
    my ($self, $cb) = @_;
    sub {
       unless ($_[0]{rbuf} =~ s/^(0|[1-9][0-9]*)://) {
          if ($_[0]{rbuf} =~ /[^0-9]/) {
-            $self->_error (Errno::EBADMSG);
+            $_[0]->_error (Errno::EBADMSG);
          }
          return;
       }
       my $len = $1;
-      $self->unshift_read (chunk => $len, sub {
+      $_[0]->unshift_read (chunk => $len, sub {
          my $string = $_[1];
          $_[0]->unshift_read (chunk => 1, sub {
             if ($_[1] eq ",") {
                $cb->($_[0], $string);
             } else {
-               $self->_error (Errno::EBADMSG);
+               $_[0]->_error (Errno::EBADMSG);
             }
          });
       });
       1
    my ($self, $cb) = @_;
    my $json = $self->{json} ||= json_coder;
    my $data;
-   my $rbuf = \$self->{rbuf};
    sub {
-      my $ref = eval { $json->incr_parse ($self->{rbuf}) };
+      my $ref = eval { $json->incr_parse ($_[0]{rbuf}) };
       if ($ref) {
-         $self->{rbuf} = $json->incr_text;
+         $_[0]{rbuf} = $json->incr_text;
          $json->incr_text = "";
-         $cb->($self, $ref);
+         $cb->($_[0], $ref);
          1
       } elsif ($@) {
          # error case
          $json->incr_skip;
-         $self->{rbuf} = $json->incr_text;
+         $_[0]{rbuf} = $json->incr_text;
          $json->incr_text = "";
-         $self->_error (Errno::EBADMSG);
+         $_[0]->_error (Errno::EBADMSG);
          ()
       } else {
-         $self->{rbuf} = "";
+         $_[0]{rbuf} = "";
+         ()
+      }
+   }
+};
+=item cbor => $cb->($handle, $scalar)
+Reads a CBOR value, decodes it and passes it to the callback. When a parse
+error occurs, an C<EBADMSG> error will be raised.
+If a L<CBOR::XS> object was passed to the constructor, then that will be
+used for the final decode, otherwise it will create a CBOR coder without
+enabling any options.
+You have to provide a dependency to L<CBOR::XS> on your own: this module
+will load the L<CBOR::XS> module, but AnyEvent does not depend on it
+itself.
+Since CBOR values are fully self-delimiting, the C<cbor> read and write
+types are an ideal simple RPC protocol: just exchange CBOR datagrams. See
+the C<cbor> write type description, above, for an actual example.
+=cut
+register_read_type cbor => sub {
+   my ($self, $cb) = @_;
+   my $cbor = $self->{cbor} ||= cbor_coder;
+   my $data;
+   sub {
+      my (@value) = eval { $cbor->incr_parse ($_[0]{rbuf}) };
+      if (@value) {
+         $cb->($_[0], @value);
+         1
+      } elsif ($@) {
+         # error case
+         $cbor->incr_reset;
+         $_[0]->_error (Errno::EBADMSG);
+         ()
+      } else {
          ()
       }
    }
 };
 =cut
 register_read_type storable => sub {
    my ($self, $cb) = @_;
-   require Storable;
+   require Storable unless $Storable::VERSION;
    sub {
       # when we can use 5.10 we can use ".", but for 5.8 we use the re-pack method
       defined (my $len = eval { unpack "w", $_[0]{rbuf} })
          or return;
       # bypass unshift if we already have the remaining chunk
       if ($format + $len <= length $_[0]{rbuf}) {
          my $data = substr $_[0]{rbuf}, $format, $len;
          substr $_[0]{rbuf}, 0, $format + $len, "";
-         $cb->($_[0], Storable::thaw ($data));
+         eval { $cb->($_[0], Storable::thaw ($data)); 1 }
+            or return $_[0]->_error (Errno::EBADMSG);
       } else {
          # remove prefix
          substr $_[0]{rbuf}, 0, $format, "";
          # read remaining chunk
          $_[0]->unshift_read (chunk => $len, sub {
-            if (my $ref = eval { Storable::thaw ($_[1]) }) {
+            eval { $cb->($_[0], Storable::thaw ($_[1])); 1 }
-               $cb->($_[0], $ref);
-            } else {
-               $self->_error (Errno::EBADMSG);
+               or $_[0]->_error (Errno::EBADMSG);
-            }
          });
       }
       1
    }
 };
+=item tls_detect => $cb->($handle, $detect, $major, $minor)
+Checks the input stream for a valid SSL or TLS handshake TLSPaintext
+record without consuming anything. Only SSL version 3 or higher
+is handled, up to the fictituous protocol 4.x (but both SSL3+ and
+SSL2-compatible framing is supported).
+If it detects that the input data is likely TLS, it calls the callback
+with a true value for C<$detect> and the (on-wire) TLS version as second
+and third argument (C<$major> is C<3>, and C<$minor> is 0..3 for SSL
+3.0, TLS 1.0, 1.1 and 1.2, respectively).  If it detects the input to
+be definitely not TLS, it calls the callback with a false value for
+C<$detect>.
+The callback could use this information to decide whether or not to start
+TLS negotiation.
+In all cases the data read so far is passed to the following read
+handlers.
+Usually you want to use the C<tls_autostart> read type instead.
+If you want to design a protocol that works in the presence of TLS
+dtection, make sure that any non-TLS data doesn't start with the octet 22
+(ASCII SYN, 16 hex) or 128-255 (i.e. highest bit set). The checks this
+read type does are a bit more strict, but might losen in the future to
+accomodate protocol changes.
+This read type does not rely on L<AnyEvent::TLS> (and thus, not on
+L<Net::SSLeay>).
+=item tls_autostart => $tls[, $tls_ctx]
+Tries to detect a valid SSL or TLS handshake. If one is detected, it tries
+to start tls by calling C<starttls> with the given arguments.
+In practise, C<$tls> must be C<accept>, or a Net::SSLeay context that has
+been configured to accept, as servers do not normally send a handshake on
+their own and ths cannot be detected in this way.
+See C<tls_detect> above for more details.
+Example: give the client a chance to start TLS before accepting a text
+line.
+   $hdl->push_read (tls_detect => "accept");
+   $hdl->push_read (line => sub {
+      print "received ", ($_[0]{tls} ? "encrypted" : "cleartext"), " <$_[1]>\n";
+   });
+=cut
+register_read_type tls_detect => sub {
+   my ($self, $cb) = @_;
+   sub {
+      # this regex matches a full or partial tls record
+      if (
+         # ssl3+: type(22=handshake) major(=3) minor(any) length_hi
+         $self->{rbuf} =~ /^(?:\z| \x16 (\z| [\x03\x04] (?:\z| . (?:\z| [\x00-\x40] ))))/xs
+         # ssl2 comapatible: len_hi len_lo type(1) major minor dummy(forlength)
+         or $self->{rbuf} =~ /^(?:\z| [\x80-\xff] (?:\z| . (?:\z| \x01 (\z| [\x03\x04] (?:\z| . (?:\z| . ))))))/xs
+      ) {
+         return if 3 != length $1; # partial match, can't decide yet
+         # full match, valid TLS record
+         my ($major, $minor) = unpack "CC", $1;
+         $cb->($self, "accept", $major + $minor * 0.1);
+      } else {
+         # mismatch == guaranteed not TLS
+         $cb->($self, undef);
+      }
+      1
+   }
+};
+register_read_type tls_autostart => sub {
+   my ($self, @tls) = @_;
+   $RH{tls_detect}($self, sub {
+      return unless $_[1];
+      $_[0]->starttls (@tls);
+   })
+};
 =back
-=item AnyEvent::Handle::register_read_type type => $coderef->($handle, $cb, @args)
+=item custom read types - Package::anyevent_read_type $handle, $cb, @args
-This function (not method) lets you add your own types to C<push_read>.
+Instead of one of the predefined types, you can also specify the name
+of a package. AnyEvent will try to load the package and then expects to
+find a function named C<anyevent_read_type> inside. If it isn't found, it
+progressively tries to load the parent package until it either finds the
+function (good) or runs out of packages (bad).
-Whenever the given C<type> is used, C<push_read> will invoke the code
+Whenever this type is used, C<push_read> will invoke the function with the
-reference with the handle object, the callback and the remaining
+handle object, the original callback and the remaining arguments.
-arguments.
-The code reference is supposed to return a callback (usually a closure)
+The function is supposed to return a callback (usually a closure) that
-that works as a plain read callback (see C<< ->push_read ($cb) >>).
+works as a plain read callback (see C<< ->push_read ($cb) >>), so you can
+mentally treat the function as a "configurable read type to read callback"
+converter.
-It should invoke the passed callback when it is done reading (remember to
+It should invoke the original callback when it is done reading (remember
-pass C<$handle> as first argument as all other callbacks do that).
+to pass C<$handle> as first argument as all other callbacks do that,
+although there is no strict requirement on this).
-Note that this is a function, and all types registered this way will be
-global, so try to use unique names.
-For examples, see the source of this module (F<perldoc -m AnyEvent::Handle>,
+For examples, see the source of this module (F<perldoc -m
-search for C<register_read_type>)).
+AnyEvent::Handle>, search for C<register_read_type>)).
 =item $handle->stop_read
 =item $handle->start_read
 Note that AnyEvent::Handle will automatically C<start_read> for you when
 you change the C<on_read> callback or push/unshift a read callback, and it
 will automatically C<stop_read> for you when neither C<on_read> is set nor
 there are any read requests in the queue.
-These methods will have no effect when in TLS mode (as TLS doesn't support
+In older versions of this module (<= 5.3), these methods had no effect,
-half-duplex connections).
+as TLS does not support half-duplex connections. In current versions they
+work as expected, as this behaviour is required to avoid certain resource
+attacks, where the program would be forced to read (and buffer) arbitrary
+amounts of data before being able to send some data. The drawback is that
+some readings of the the SSL/TLS specifications basically require this
+attack to be working, as SSL/TLS implementations might stall sending data
+during a rehandshake.
+As a guideline, during the initial handshake, you should not stop reading,
+and as a client, it might cause problems, depending on your application.
 =cut
 sub stop_read {
    my ($self) = @_;
-   delete $self->{_rw} unless $self->{tls};
+   delete $self->{_rw};
 }
 sub start_read {
    my ($self) = @_;
-   unless ($self->{_rw} || $self->{_eof}) {
+   unless ($self->{_rw} || $self->{_eof} || !$self->{fh}) {
       Scalar::Util::weaken $self;
       $self->{_rw} = AE::io $self->{fh}, 0, sub {
          my $rbuf = \($self->{tls} ? my $buf : $self->{rbuf});
-         my $len = sysread $self->{fh}, $$rbuf, $self->{read_size} || 8192, length $$rbuf;
+         my $len = sysread $self->{fh}, $$rbuf, $self->{read_size}, length $$rbuf;
          if ($len > 0) {
             $self->{_activity} = $self->{_ractivity} = AE::now;
             if ($self->{tls}) {
                &_dotls ($self);
             } else {
                $self->_drain_rbuf;
             }
+            if ($len == $self->{read_size}) {
+               $self->{read_size} *= 2;
+               $self->{read_size} = $self->{max_read_size} || MAX_READ_SIZE
+                  if $self->{read_size} > ($self->{max_read_size} || MAX_READ_SIZE);
+            }
          } elsif (defined $len) {
             delete $self->{_rw};
             $self->{_eof} = 1;
             $self->_drain_rbuf;
    my ($self, $err) = @_;
    return $self->_error ($!, 1)
       if $err == Net::SSLeay::ERROR_SYSCALL ();
-   my $err =Net::SSLeay::ERR_error_string (Net::SSLeay::ERR_get_error ());
+   my $err = Net::SSLeay::ERR_error_string (Net::SSLeay::ERR_get_error ());
    # reduce error string to look less scary
    $err =~ s/^error:[0-9a-fA-F]{8}:[^:]+:([^:]+):/\L$1: /;
    if ($self->{_on_starttls}) {
 sub _dotls {
    my ($self) = @_;
    my $tmp;
-   if (length $self->{_tls_wbuf}) {
+   while (length $self->{_tls_wbuf}) {
-      while (($tmp = Net::SSLeay::write ($self->{tls}, $self->{_tls_wbuf})) > 0) {
+      if (($tmp = Net::SSLeay::write ($self->{tls}, $self->{_tls_wbuf})) <= 0) {
-         substr $self->{_tls_wbuf}, 0, $tmp, "";
+         $tmp = Net::SSLeay::get_error ($self->{tls}, $tmp);
+         return $self->_tls_error ($tmp)
+            if $tmp != $ERROR_WANT_READ
+               && ($tmp != $ERROR_SYSCALL || $!);
+         last;
       }
-      $tmp = Net::SSLeay::get_error ($self->{tls}, $tmp);
+      substr $self->{_tls_wbuf}, 0, $tmp, "";
-      return $self->_tls_error ($tmp)
-         if $tmp != $ERROR_WANT_READ
-            && ($tmp != $ERROR_SYSCALL || $!);
    }
    while (defined ($tmp = Net::SSLeay::read ($self->{tls}))) {
       unless (length $tmp) {
          $self->{_on_starttls}
       $self->{_tls_rbuf} .= $tmp;
       $self->_drain_rbuf;
       $self->{tls} or return; # tls session might have gone away in callback
    }
-   $tmp = Net::SSLeay::get_error ($self->{tls}, -1);
+   $tmp = Net::SSLeay::get_error ($self->{tls}, -1); # -1 is not neccessarily correct, but Net::SSLeay doesn't tell us
    return $self->_tls_error ($tmp)
       if $tmp != $ERROR_WANT_READ
          && ($tmp != $ERROR_SYSCALL || $!);
    while (length ($tmp = Net::SSLeay::BIO_read ($self->{_wbio}))) {
       $self->{wbuf} .= $tmp;
       $self->_drain_wbuf;
+      $self->{tls} or return; # tls session might have gone away in callback
    }
    $self->{_on_starttls}
       and Net::SSLeay::state ($self->{tls}) == Net::SSLeay::ST_OK ()
       and (delete $self->{_on_starttls})->($self, 1, "TLS/SSL connection established");
 =item $handle->starttls ($tls[, $tls_ctx])
 Instead of starting TLS negotiation immediately when the AnyEvent::Handle
 object is created, you can also do that at a later time by calling
-C<starttls>.
+C<starttls>. See the C<tls> constructor argument for general info.
 Starting TLS is currently an asynchronous operation - when you push some
 write data and then call C<< ->starttls >> then TLS negotiation will start
-immediately, after which the queued write data is then sent.
+immediately, after which the queued write data is then sent. This might
+change in future versions, so best make sure you have no outstanding write
+data when calling this method.
 The first argument is the same as the C<tls> constructor argument (either
 C<"connect">, C<"accept"> or an existing Net::SSLeay object).
 The second argument is the optional C<AnyEvent::TLS> object that is used
 context in C<< $handle->{tls_ctx} >> after this call and can be used or
 changed to your liking. Note that the handshake might have already started
 when this function returns.
 Due to bugs in OpenSSL, it might or might not be possible to do multiple
-handshakes on the same stream. Best do not attempt to use the stream after
+handshakes on the same stream. It is best to not attempt to use the
-stopping TLS.
+stream after stopping TLS.
+This method may invoke callbacks (and therefore the handle might be
+destroyed after it returns).
 =cut
 our %TLS_CACHE; #TODO not yet documented, should we?
    my ($self, $tls, $ctx) = @_;
    Carp::croak "It is an error to call starttls on an AnyEvent::Handle object while TLS is already active, caught"
       if $self->{tls};
+   unless (defined $AnyEvent::TLS::VERSION) {
+      eval {
+         require Net::SSLeay;
+         require AnyEvent::TLS;
+         1
+      } or return $self->_error (Errno::EPROTO, 1, "TLS support not available on this system");
+   }
    $self->{tls}     = $tls;
    $self->{tls_ctx} = $ctx if @_ > 2;
    return unless $self->{fh};
-   require Net::SSLeay;
    $ERROR_SYSCALL   = Net::SSLeay::ERROR_SYSCALL     ();
    $ERROR_WANT_READ = Net::SSLeay::ERROR_WANT_READ   ();
    $tls = delete $self->{tls};
    $ctx = $self->{tls_ctx};
    local $Carp::CarpLevel = 1; # skip ourselves when creating a new context or session
    if ("HASH" eq ref $ctx) {
-      require AnyEvent::TLS;
       if ($ctx->{cache}) {
          my $key = $ctx+0;
          $ctx = $TLS_CACHE{$key} ||= new AnyEvent::TLS %$ctx;
       } else {
          $ctx = new AnyEvent::TLS %$ctx;
    Net::SSLeay::CTX_set_mode ($tls, 1|2);
    $self->{_rbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
    $self->{_wbio} = Net::SSLeay::BIO_new (Net::SSLeay::BIO_s_mem ());
-   Net::SSLeay::BIO_write ($self->{_rbio}, delete $self->{rbuf});
+   Net::SSLeay::BIO_write ($self->{_rbio}, $self->{rbuf});
+   $self->{rbuf} = "";
    Net::SSLeay::set_bio ($tls, $self->{_rbio}, $self->{_wbio});
    $self->{_on_starttls} = sub { $_[0]{on_starttls}(@_) }
       if $self->{on_starttls};
 =item $handle->stoptls
 Shuts down the SSL connection - this makes a proper EOF handshake by
 sending a close notify to the other side, but since OpenSSL doesn't
-support non-blocking shut downs, it is not guarenteed that you can re-use
+support non-blocking shut downs, it is not guaranteed that you can re-use
 the stream afterwards.
+This method may invoke callbacks (and therefore the handle might be
+destroyed after it returns).
 =cut
 sub stoptls {
    my ($self) = @_;
-   if ($self->{tls}) {
+   if ($self->{tls} && $self->{fh}) {
       Net::SSLeay::shutdown ($self->{tls});
       &_dotls;
 #      # we don't give a shit. no, we do, but we can't. no...#d#
       if $self->{tls} > 0;
    delete @$self{qw(_rbio _wbio _tls_wbuf _on_starttls)};
 }
+=item $handle->resettls
+This rarely-used method simply resets and TLS state on the handle, usually
+causing data loss.
+One case where it may be useful is when you want to skip over the data in
+the stream but you are not interested in interpreting it, so data loss is
+no concern.
+=cut
+*resettls = \&_freetls;
 sub DESTROY {
    my ($self) = @_;
    &_freetls;
       push @linger, AE::io $fh, 1, sub {
          my $len = syswrite $fh, $wbuf, length $wbuf;
          if ($len > 0) {
             substr $wbuf, 0, $len, "";
-         } else {
+         } elsif (defined $len || ($! != EAGAIN && $! != EINTR && $! != WSAEWOULDBLOCK)) {
             @linger = (); # end
          }
       };
       push @linger, AE::timer $linger, 0, sub {
          @linger = ();
 sub AnyEvent::Handle::destroyed::AUTOLOAD {
    #nop
 }
+=item $handle->destroyed
+Returns false as long as the handle hasn't been destroyed by a call to C<<
+->destroy >>, true otherwise.
+Can be useful to decide whether the handle is still valid after some
+callback possibly destroyed the handle. For example, C<< ->push_write >>,
+C<< ->starttls >> and other methods can call user callbacks, which in turn
+can destroy the handle, so work can be avoided by checking sometimes:
+   $hdl->starttls ("accept");
+   return if $hdl->destroyed;
+   $hdl->push_write (...
+Note that the call to C<push_write> will silently be ignored if the handle
+has been destroyed, so often you can just ignore the possibility of the
+handle being destroyed.
+=cut
+sub destroyed { 0 }
+sub AnyEvent::Handle::destroyed::destroyed { 1 }
 =item AnyEvent::Handle::TLS_CTX
 This function creates and returns the AnyEvent::TLS object used by default
 for TLS mode.
 It is only safe to "forget" the reference inside EOF or error callbacks,
 from within all other callbacks, you need to explicitly call the C<<
 ->destroy >> method.
+=item Why is my C<on_eof> callback never called?
+Probably because your C<on_error> callback is being called instead: When
+you have outstanding requests in your read queue, then an EOF is
+considered an error as you clearly expected some data.
+To avoid this, make sure you have an empty read queue whenever your handle
+is supposed to be "idle" (i.e. connection closes are O.K.). You can set
+an C<on_read> handler that simply pushes the first read requests in the
+queue.
+See also the next question, which explains this in a bit more detail.
+=item How can I serve requests in a loop?
+Most protocols consist of some setup phase (authentication for example)
+followed by a request handling phase, where the server waits for requests
+and handles them, in a loop.
+There are two important variants: The first (traditional, better) variant
+handles requests until the server gets some QUIT command, causing it to
+close the connection first (highly desirable for a busy TCP server). A
+client dropping the connection is an error, which means this variant can
+detect an unexpected detection close.
+To handle this case, always make sure you have a non-empty read queue, by
+pushing the "read request start" handler on it:
+   # we assume a request starts with a single line
+   my @start_request; @start_request = (line => sub {
+      my ($hdl, $line) = @_;
+      ... handle request
+      # push next request read, possibly from a nested callback
+      $hdl->push_read (@start_request);
+   });
+   # auth done, now go into request handling loop
+   # now push the first @start_request
+   $hdl->push_read (@start_request);
+By always having an outstanding C<push_read>, the handle always expects
+some data and raises the C<EPIPE> error when the connction is dropped
+unexpectedly.
+The second variant is a protocol where the client can drop the connection
+at any time. For TCP, this means that the server machine may run out of
+sockets easier, and in general, it means you cannot distinguish a protocl
+failure/client crash from a normal connection close. Nevertheless, these
+kinds of protocols are common (and sometimes even the best solution to the
+problem).
+Having an outstanding read request at all times is possible if you ignore
+C<EPIPE> errors, but this doesn't help with when the client drops the
+connection during a request, which would still be an error.
+A better solution is to push the initial request read in an C<on_read>
+callback. This avoids an error, as when the server doesn't expect data
+(i.e. is idly waiting for the next request, an EOF will not raise an
+error, but simply result in an C<on_eof> callback. It is also a bit slower
+and simpler:
+   # auth done, now go into request handling loop
+   $hdl->on_read (sub {
+      my ($hdl) = @_;
+      # called each time we receive data but the read queue is empty
+      # simply start read the request
+      $hdl->push_read (line => sub {
+         my ($hdl, $line) = @_;
+         ... handle request
+         # do nothing special when the request has been handled, just
+         # let the request queue go empty.
+      });
+   });
 =item I get different callback invocations in TLS mode/Why can't I pause
 reading?
 Unlike, say, TCP, TLS connections do not consist of two independent
-communication channels, one for each direction. Or put differently. The
+communication channels, one for each direction. Or put differently, the
 read and write directions are not independent of each other: you cannot
 write data unless you are also prepared to read, and vice versa.
-This can mean than, in TLS mode, you might get C<on_error> or C<on_eof>
+This means that, in TLS mode, you might get C<on_error> or C<on_eof>
 callback invocations when you are not expecting any read data - the reason
 is that AnyEvent::Handle always reads in TLS mode.
 During the connection, you have to make sure that you always have a
 non-empty read-queue, or an C<on_read> watcher. At the end of the
    $handle->on_eof (undef);
    $handle->on_error (sub {
       my $data = delete $_[0]{rbuf};
    });
+Note that this example removes the C<rbuf> member from the handle object,
+which is not normally allowed by the API. It is expressly permitted in
+this case only, as the handle object needs to be destroyed afterwards.
 The reason to use C<on_error> is that TCP connections, due to latencies
 and packets loss, might get closed quite violently with an error, when in
-fact, all data has been received.
+fact all data has been received.
 It is usually better to use acknowledgements when transferring data,
 to make sure the other side hasn't just died and you got the data
 intact. This is also one reason why so many internet protocols have an
 explicit QUIT command.
 C<low_water_mark> this will be called precisely when all data has been
 written to the socket:
    $handle->push_write (...);
    $handle->on_drain (sub {
-      warn "all data submitted to the kernel\n";
+      AE::log debug => "All data submitted to the kernel.";
       undef $handle;
    });
 If you just want to queue some data and then signal EOF to the other side,
 consider using C<< ->push_shutdown >> instead.
 =item I want to contact a TLS/SSL server, I don't care about security.
 If your TLS server is a pure TLS server (e.g. HTTPS) that only speaks TLS,
-simply connect to it and then create the AnyEvent::Handle with the C<tls>
+connect to it and then create the AnyEvent::Handle with the C<tls>
 parameter:
    tcp_connect $host, $port, sub {
       my ($fh) = @_;
 When you have intermediate CA certificates that your clients might not
 know about, just append them to the C<cert_file>.
 =back
 =head1 SUBCLASSING AnyEvent::Handle
 In many cases, you might want to subclass AnyEvent::Handle.
 To make this easier, a given version of AnyEvent::Handle uses these
 =item * all members not documented here and not prefixed with an underscore
 are free to use in subclasses.
 Of course, new versions of AnyEvent::Handle may introduce more "public"
-member variables, but thats just life, at least it is documented.
+member variables, but that's just life. At least it is documented.
 =back
 =head1 AUTHOR
 Robin Redeker C<< <elmex at ta-sa.org> >>, Marc Lehmann <schmorp@schmorp.de>.
 =cut
-1; # End of AnyEvent::Handle
+1

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing AnyEvent/lib/AnyEvent/Handle.pm (file contents): Revision 1.182 by root, Thu Sep 3 12:35:01 2009 UTC vs. Revision 1.239 by root, Tue Dec 10 20:39:12 2013 UTC

Diff Legend

Comparing AnyEvent/lib/AnyEvent/Handle.pm (file contents):
Revision 1.182 by root, Thu Sep 3 12:35:01 2009 UTC vs.
Revision 1.239 by root, Tue Dec 10 20:39:12 2013 UTC