=========================
2005-12-09 05:56:38 v0_06
=========================

  2005-12-09 05:50:54 (r41) by rcaputo
  .releaserc A; Client-Keepalive.pm M

    Add a .releaserc for Module::Release, and bump up the version. 

  2005-09-17 18:26:46 (r40) by rcaputo
  Makefile.PL M; Client-Keepalive.pm M

    Apply a semi-anonymous guest patch to make POE::Component::SSLify an
    optional dependancy. It defers errors regarding missing SSL support
    to runtime. This resolves rt.cpan.org ticket 14253. 

=========================
2005-09-13 15:52:00 v0_05
=========================

  2005-09-13 15:51:04 (r38) by rcaputo; Client-Keepalive.pm M

    Now we are 0.05! 

  2005-09-04 18:56:42 (r37) by rcaputo
  t/04_free_each.t M; Client-Keepalive.pm M

    The code to reclaim freed sockets tests to see whether the freed
    socket is still usable. It did not check whether the socket was
    actually open! This patch adds that check, hopefully robustifying the
    code a little. 

=========================
2005-07-29 16:02:20 v0_04
=========================

  2005-07-29 16:02:02 (r35) by rcaputo
  Makefile.PL M; Client-Keepalive.pm M

    Bump the version for release, and increase the version of the
    Client::DNS dependency to "the latest". 

  2005-07-29 15:54:39 (r34) by rcaputo; t/50_bisbee_timeout.t A

    I forgot to add Jeff Bisbee's regression test. Aaaaaugh! 

  2005-07-28 05:57:46 (r33) by rcaputo
  t/01_socket_reuse.t M; MANIFEST M; t/03_each_queue.t M;
  Client-Keepalive.pm M

    Jeff Bisbee discovered that Client::Keepalive may return multiple
    responses for a single request. This causes bizarre behavior (usually
    crashes) in the likes of POE::Component::Client::HTTP. This commit
    adds a regression test for the issue and patches Client::Keepalive to
    avoid redundant responses. 

  2005-07-25 18:50:37 (r32) by rcaputo; Client-Keepalive.pm M

    Replace tabs with an appropriate number of spaces. At some point
    we'll need to just go with tabs-indent; spaces-align behavior. 

  2005-06-23 02:13:42 (r31) by rcaputo; README M

    Update the README for the new documentation. TODO: Automate this for
    distributions where the README is merely the main module's perldoc
    directed to a text file. 

=========================
2005-06-23 01:11:36 v0_03
=========================

  2005-06-23 01:04:12 (r29) by rcaputo; Client-Keepalive.pm M

    Oops! Something wasn't indented properly! Gasp! 

  2005-06-22 14:40:07 (r28) by rcaputo; Client-Keepalive.pm M

    Bump up the version to 0.03 for release. Also destroy tabs in the
    file. 

  2005-06-21 17:36:19 (r27) by rcaputo; Conn-Keepalive.pm M

    Fix $VERSION to work with Subversion. 

  2005-06-21 15:14:28 (r26) by rcaputo; MANIFEST M; Makefile.PL M

    Remove META.yml from MANIFEST. It appears to be auto-generated into
    the distribution at "make dist" time. Replace cvs-log.perl with
    svn-log.perl in Makefile.PL since we're in Subversion now. 

  2005-05-06 16:18:30 (r25) by rcaputo; Makefile.PL M

    Depend on POE 0.31, and add POE::Component::SSLify as a dependency. 

  2005-05-06 16:00:19 (r24) by martijn
  t/11_dead_socket.t M; Client-Keepalive.pm M

    r5553@fangorn: martijn | 2005-05-06 17:59:52 +0200 Fix for the
    connection being closed on the server side, but not waiting for the
    EOF on the client side if the connection is closed before the
    Connection object is released. We don't reuse the socket in that case
    anymore. Now to try to find a way to do the same if the connection
    closing happens after the Connection object goes away. In this case,
    the socket may have already been reused, so the new client still gets
    an EOF 'error' as the first event, and nothing else. 

  2005-05-06 16:00:02 (r23) by martijn
  mylib/TestServer.pm M; t/11_dead_socket.t M

    r5552@fangorn: martijn | 2005-05-06 17:58:48 +0200 fix test and
    TestServer 

  2005-05-06 15:59:44 (r22) by martijn
  mylib/TestServer.pm M; MANIFEST M; t/11_dead_socket.t A

    r5551@fangorn: martijn | 2005-05-06 17:14:53 +0200 Add tests for bug
    where we throw away the connection right before we get EOF, and it
    gets reused and the new connection gets an EOF error event
    immediately. And now to figure out how to fix this 

  2005-05-06 15:59:29 (r21) by martijn; t/10_resolver.t A; MANIFEST M

    r5550@fangorn: martijn | 2005-05-06 17:01:48 +0200 add DNS tests 

  2005-05-06 15:59:18 (r20) by martijn; Client-Keepalive.pm M

    r5549@fangorn: martijn | 2005-05-06 16:59:39 +0200 use PoCo::SSLify
    when the scheme is https. an imap component might want us to check
    for imaps too for example, but we'll worry about that when it starts
    to exist :) 

  2005-05-06 15:59:07 (r19) by martijn; Client-Keepalive.pm M

    r5548@fangorn: martijn | 2005-05-06 16:58:35 +0200 remove timer for
    requests for which thea DNS lookup fails 

  2005-05-06 15:58:57 (r18) by martijn; Client-Keepalive.pm M

    r5547@fangorn: martijn | 2005-05-06 16:57:47 +0200 add some debugging
    output 

  2005-05-06 15:58:31 (r17) by martijn
  t/07_keep_alive.t M; t/01_socket_reuse.t M; t/05_errors.t M;
  t/06_activity.t M; Makefile.PL M; t/04_free_each.t M; t/03_each_queue.t
  M; t/08_quick_reuse.t M; Client-Keepalive.pm M

    r5546@fangorn: martijn | 2005-05-06 16:53:12 +0200 Make it use
    Client::DNS. and adjust the tests to use localhost instead of its IP
    (except for a few for better coverage) 

  2005-05-05 05:25:51 (r16) by rcaputo; mylib/TestServer.pm M

    Add a shutdown_clients() method so tests can shutdown all the clients
    at once. Also make the server more robust against race conditions by
    not calling put() on a disconnected client. 

  2005-05-05 05:24:45 (r15) by rcaputo; Conn-Keepalive.pm M

    Destroy the connection object's wheel before freeing the underlying
    socket. Otherwise there's overlap between it and the keepalive pool
    while they both have select_read() on the socket. 

=========================
2005-04-15 15:49:56 v0_02
=========================

  2005-04-15 15:49:56 (r13) by rcaputo
  t/07_keep_alive.t M; t/01_socket_reuse.t M; t/02_socket_queue.t M;
  t/09_timeout.t M; t/06_activity.t M; t/04_free_each.t M;
  t/03_each_queue.t M; t/08_quick_reuse.t M; Client-Keepalive.pm M

    Applied Martijn van Beers' patch to make Client::Keepalive never
    return the connection immediately. Rather, it always posts
    connections back, even if they can be immediately satisfied from the
    component's connection cache. This simplifies the module's general
    usage since the pattern was to post the immediately returned
    connection to a common got-a-connection handler anyway. 

  2004-10-06 02:41:06 (r12) by rcaputo
  t/07_keep_alive.t M; t/02_socket_queue.t M; t/05_errors.t M;
  t/04_free_each.t M; t/03_each_queue.t M; t/08_quick_reuse.t M;
  Client-Keepalive.pm M

    Remove the component's alias during shutdown(). This immediately
    kills off the component when it's done shutting down. The component's
    immediate destruction triggers some ASSERT errors due to sloppy
    cleanup in the tests. This commit fixes the tests so they stop dying
    and some junk. 

  2004-10-06 02:15:23 (r11) by rcaputo; Client-Keepalive.pm M

    Ah, there's no need to pass %args to a session initializer if they
    aren't being used there anyway. 

  2004-10-06 02:12:16 (r10) by rcaputo; Client-Keepalive.pm M

    Remove some parameters we're not using. This eliminates all use of
    HEAP in the code. Looks like object-based sessions are more powerful
    than I originally thought. 

  2004-10-06 01:47:03 (r9) by rcaputo; Makefile.PL M

    Well, technically, it doesn't need all these funky prerequisites.
    They're holdovers from when this Makefile.PL belonged to
    POE::Component::Client::HTTP. 

==============
End of Excerpt
==============