Permalink
Commits on Sep 29, 2011
  1. Merge branch 'close_schedule_fix'

    raggi committed Sep 29, 2011
    * close_schedule_fix:
      Expose EM.num_close_scheduled, and utilise in TimeTilNextEvent to avoid long waits for unbinds.
  2. whitespace

    tmm1 committed Sep 29, 2011
  3. Merge branch 'master' into close_schedule_fix

    raggi committed Sep 29, 2011
    * master: (146 commits)
      Fix reconnect on unix-domain connection #261
      Fix rake-compiler cross compiling issues for win32 gems
      Another compile warning
      Fix compile warning
      README grammar fix
      bump rake-compiler dependency
      Remove dependency on readbytes.rb for ruby 1.9 (fixes #167, #234)
      Fix compilation on MSVC2008 (fixes #253)
      fix code samples in iterator docs
      pull pause/resume logic down into EventableDescriptor
      test EM::Connection#pause on popen connections
      print out ssl errors to stdout before exploding
      Renaming EventMachine_t Closing to Deregister
      Watched sockets should still receive Closing
      Tweaking the patch to only affect socket shutdown and close. Ref: #242
      Alias Ruby C API methods for 1.8/1.9 compat
      Add pure-ruby implementations of EM::Connection#set_sock_opt and #get_sock_opt
      Implement EM::Connection#set_sock_opt
      Never close STDIN, STDOUT and STDERR (i.e. only close when fd > 2) Should solve #183
      Segfault fix for pipes
      ...
    
    Conflicts:
    	ext/em.cpp
    	tests/test_basic.rb
Commits on Sep 17, 2011
  1. Merge pull request #262 from dre3k/master

    tmm1 committed Sep 17, 2011
    Fix reconnect on unix-domain connection #261
Commits on Sep 16, 2011
  1. Another compile warning

    tmm1 committed Sep 16, 2011
  2. Fix compile warning

    tmm1 committed Sep 16, 2011
Commits on Sep 15, 2011
Commits on Sep 12, 2011
  1. Merge pull request #256 from mblair/master

    tmm1 committed Sep 12, 2011
    Here's a grammar fix for the readme.
  2. README grammar fix

    mblair committed Sep 12, 2011
Commits on Sep 10, 2011
  1. bump rake-compiler dependency

    tmm1 committed Sep 10, 2011
Commits on Sep 9, 2011
  1. Merge pull request #232 from dturnbull/master

    tmm1 committed Sep 9, 2011
    return self
Commits on Sep 7, 2011
Commits on Sep 5, 2011
  1. fix code samples in iterator docs

    tmm1 committed Sep 5, 2011
Commits on Sep 1, 2011
  1. Merge pull request #246 from rtomayko/popen-pause

    tmm1 committed Sep 1, 2011
    Add pause/resume support for PipeDescriptor
  2. pull pause/resume logic down into EventableDescriptor

    tmm1 committed with rtomayko Sep 1, 2011
    Fixes pause/resume support on PipeDescriptor and possibly other
    connection types. See #245 for more info.
Commits on Aug 25, 2011
  1. Merge pull request #242 from vishnugopal/never-close-stdin-out-err

    tmm1 committed Aug 25, 2011
    Never close STDIN, STDOUT and STDERR (i.e. only close when fd > 2) Should
  2. Merge pull request #240 from bklang/master

    tmm1 committed Aug 25, 2011
    Implement EM::Connection#set_sock_opt
Commits on Aug 24, 2011
Commits on Aug 20, 2011
  1. Merge pull request #239 from pbozeman/pipe_fix

    tmm1 committed Aug 20, 2011
    Segfault fix for pipes
  2. Segfault fix for pipes

    Patrick Bozeman committed Aug 20, 2011
    EventMachine relies on the fact that when close(fd)
    is called that the fd is removed from any
    epoll event queues.
    
    However, this is not *always* the behavior of close(fd)
    
    See man 4 epoll Q6/A6 and then consider what happens
    when using pipes with eventmachine.
    (As is often done when communicating with a subprocess)
    
    The pipes end up looking like:
    
    ls -l /proc/<pid>/fd
    ...
    lr-x------ 1 root root 64 2011-08-19 21:31 3 -> pipe:[940970]
    l-wx------ 1 root root 64 2011-08-19 21:31 4 -> pipe:[940970]
    
    This meets the critera from man 4 epoll Q6/A4 for not
    removing fds from epoll event queues until all fds
    that reference the underlying file have been removed.
    
    If the EventableDescriptor associated with fd 3 is deleted,
    its dtor will call EventableDescriptor::Close(),
    which will call ::close(int fd).
    
    However, unless the EventableDescriptor associated with fd 4 is
    also deleted before the next call to epoll_wait, events may fire
    for fd 3 that were registered with an already deleted
    EventableDescriptor.
    
    Therefore, it is necessary to notify EventMachine that
    when an EventableDescriptor is closing so that it can
    remove it from the event loop.
Commits on Aug 11, 2011
  1. Merge pull request #236 from nahi/remove-rdoc-commands

    tmm1 committed Aug 11, 2011
    Remove RDoc document control commands