input

这是广泛使用的通信开源项目,对于大容量,高并发的通讯要求完全能够胜任,他广泛可用于网络游戏医学图像网关的高qos要求.更详细的内容可阅读相应的材料

        * ace/ace_dll.dsp:

          Unchecked the "separate types" option for debug linking.
          Thanks to Nick Pratt for this suggestion also.

Mon Mar 26 13:00:37 2001  Carlos O'Ryan  <[e-mail]>

        * ace/Node.h:
        * ace/Node.cpp:
        * ace/Array_Base.h:
        * ace/Array_Base.inl:
        * ace/Array_Base.cpp:
        * ace/Unbounded_Queue.h:
        * ace/Unbounded_Queue.inl:
        * ace/Unbounded_Queue.cpp:
        * ace/Unbounded_Set.h:
        * ace/Unbounded_Set.inl:
        * ace/Unbounded_Set.cpp:
        * ace/Containers_T.h:
        * ace/Containers_T.i:
        * ace/Containers_T.cpp:
          Move several containers to their own file.  This reduces the
          amount of code that needs to be parsed, improving compilation
          time (not to mention the fact that the code becomes more
          readable.)
          For backwards compatibility the Containers file still includes
          the new smaller files.

        * ace/Makefile:
        * ace/Makefile.am:
        * ace/Makefile.bor:
          Add new files to the install rules.
          Since they are templates they do not need to be compiled on MSVC
          so no changes to the project files are required.

        * ace/ARGV.h:
        * ace/ARGV.cpp:
        * ace/Future.h:
        * ace/Handle_Gobbler.h:
        * ace/Local_Name_Space.h:
        * ace/Malloc_Base.h:
        * ace/Memory_Pool.h:
        * ace/Name_Space.h:
        * ace/Priority_Reactor.h:
        * ace/Remote_Name_Space.h:
        * ace/Service_Config.h:
        * ace/Signal.cpp:
        * ace/Signal.h:
        * ace/Stats.h:
        * ace/Thread_Manager.h:
        * ace/Timeprobe_T.h:
        * ace/Timer_Heap_T.h:
        * ace/Timer_Queue_Adapters.h:
        * ace/TkReactor.cpp:
          To reduce compilation and recompilation times we only include
          the header file that includes the container used in each case,
          instead of the full containers class.

Fri Mar 23 12:40:34 2001  Douglas C. Schmidt  <[e-mail]>

        * apps/Gateway/Gateway/Event_Forwarding_Discriminator.cpp: Make sure
          to delete the Consumer_Dispatch_Set * that we get back from
          unbind().  Thanks to the ever vigilant Mike Winter
          <[e-mail]> for reporting this.

Thu Mar 22 10:18:24 2001  Douglas C. Schmidt  <[e-mail]>

        * ace/SString.h: Make the various data members protected
          rather than private so subclasses can access them.  Thanks to
          Martin Krumpolec <[e-mail]> for this suggestion.

        * tests/README: Clarified some things about the tests.
          Thanks to Eric Desamore <[e-mail]> for
          motivating this.

Mon Mar 26 11:03:22 2001  Balachandran  <[e-mail]>

        * ACE version 5.1.14 released.

Sat Mar 24 07:02:35 2001  Chris Cleeland  <[e-mail]>

        * ace/INET_Addr.i (get_host_addr): Added a const_cast to the
        VxWorks-only arm that casts away const on the this ptr so that
        inet_ntoa_b can access and fill in the buffer data member that's
        used to hold the result from inet_ntoa_b.  Thanks to Erik Johannes
        <[e-mail]> for point this out.

Fri Mar 23 18:01:05 2001  Balachandran Natarajan  <[e-mail]>

        * bin/auto_run_tests.lst: Some of the tests cannot be run in
          Single threaded configurations. Prevented such tests from
          running in the nightly builds.

Fri Mar 23 16:25:18 2001  Balachandran Natarajan  <[e-mail]>

        * tests/New_Fail_Test.cpp: Reverted the change from "Fri Mar 23
          06:34:33 2001  Balachandran Natarajan  <[e-mail]>". The
          change fails to comipile when exceptions are on.

Fri Mar 23 08:47:29 2001  Carlos O'Ryan  <[e-mail]>

        * apps/Gateway/Gateway/Options.cpp:
          Explicitly #include Log_Msg.h, on platforms that do not define
          ACE_TEMPLATES_REQUIRE_SOURCE the header file is not implicitly
          included.  This should fix compilation problems on Tru64/CXX and
          SunCC/4.2

Fri Mar 23 06:34:33 2001  Balachandran Natarajan  <[e-mail]>

        * tests/New_Fail_Test.cpp: Fixed a stupid warning in g++,
          No_AMI_Messaging  builds. Surprising that it showed up only in
          this build.

Thu Mar 22 21:33:37 2001  Chris Cleeland  <[e-mail]>

        * ace/INET_Addr.h: VxWorks' inet_ntoa doesn't behave like most
        others, and our use was causing memory leaks.  VxWorks prefers
        that you use its own inet_ntoa_b, which takes the buffer to fill
        as an argument.  Thus, for VxWorks, added a data member char buf
        of the appropriate length so that get_host_addr can use
        inet_ntoa_b to fill that in.

        * ace/INET_Addr.i (get_host_addr): For VxWorks only, changed this
        to use the preferred inet_ntoa_b.  This eliminate memory leaks
        under VxWorks.  Thanks to Erik Johannes <[e-mail]> for
        the suggested fix.

Wed Mar 21 19:20:02 2001  Balachandran Natarajan  <[e-mail]>

        * apps/Gateway/Peer/Peer.cpp (nonblk_put): Fixed a compile error
          in g++.

Wed Mar 21 17:28:26 2001  Steve Huston  <[e-mail]>

        * ace/Message_Queue.h:
        * ace/Message_Queue_T.h: Moved definition of typedef for
          ACE_DEFAULT_MESSAGE_QUEUE_TYPE from Message_Queue.h to
          Message_Queue_T.h, just after defining ACE_Message_Queue<>.
          Fixes the case where Message_Queue_T.h is included first,
          it includes Message_Queue.h before ACE_Message_Queue<> is
          defined, and that breaks the compile. IBM C/C++ hit this.

Wed Mar 21 16:42:19 2001  Steve Huston  <[e-mail]>

        * bin/auto_compile: Added an option: -notao to ignore the TAO aspects
          of the non-configurable things, like looking at the TAO/ChangeLog.
          Also, now looks for auto_run_tests.lst in the $BUILD/bin directory,
          not the $MODULE/bin directory - this allows some builds to specify
          their own lists.

Wed Mar 21 16:00:56 2001  Nanbor Wang  <[e-mail]>

          The followings are contributions from Scott Snyder
          <[e-mail]> to make ACE happy on Digital UNIX with KCC
          3.4d.

        * include/makeinclude/platform_tru64_kcc.GNU:
        * include/makeinclude/platform_osf1_4.0_kcc.GNU: Added KCCOPTS and
          support for tru64 unix.

        * ace/OS.h: On this platform, we ran into problems with sigwait
          being a macro.  The fix that had been added for this was keyed
          on __DECCXX_VER, so it isn't picked up on this platform.

        * ace/Sock_Connect.cpp: Enum_Interface_Test failed on this
          platform.  The code in Socket_Connect.cpp that implements this
          is conditional on `__unix' being defined.  However, with this
          compiler, we get `unix' and `__unix__' defined, but not
          `__unix'.  Fixed by testing for `__unix__' too.

        * ace/ACE.cpp:  Cache_Conn_Test was failing.  It looks like the
          code in ACE::out_of_handles doesn't have a case to handle osf1.

Wed Mar 21 12:08:35 2001  Nanbor Wang  <[e-mail]>

          The followings are contributions from Scott Snyder
          <[e-mail]> to make ACE happy on IRIX 6.x with KCC 3.4d.

        * include/makeinclude/platform_irix6.x_kcc.GNU:
          Disable the following warings and separate common options into
          $(KCCOPTS).

          280: declaration of a member with the same name as its class Get
          this from <netinet/in.h>.

          608: this pragma must immediately precede a declaration or
          statement Get this from <ucontext.h>.

        * ace/OS.h: Changed to only request RTLD_GROUP and RTLD_NODELETE
          if they are actually available.

        * ace/config-irix6.x-kcc.h: Undefined
          ACE_LACKS_PLACEMENT_OPERATOR_DELETE since KCC actually supports it.

          Config_Test was failing because with the default setup, ace
          tries to locate the shared memory pool at 0x04000000.  But it
          turns out that this is the same address at which irix likes to
          load the KCC shared library, libKCC.so.  So the memory pool ends
          up blowing away all the library code.  Fixed by shifting the
          default base address to 0x05000000 for this configuration.

          OS_Test was failing because on irix, kcc 4.0d miscompiles
          memchr_emulation if +K3 optimization is used.  I worked around
          this by just using the library version of memchr instead:

        * ace/ACE.cpp:  Cached_Conn_Test was failing because the code in
          ACE::out_of_handles doesn't have a case to handle irix.

        * tests/DLL_Test.cpp: It died trying to run object dtors. Fixed by
          turning off the close-on-destruction flag for this test.

Wed Mar 21 13:33:58 2001  Ossama Othman  <[e-mail]>

        * examples/IPC_SAP/SSL_SAP/SSL-server-fancy.cpp:

          Patched this file so that the example actually works.  It now
          correctly interacts with the SSL client in this example.  Thanks
          to Defang Zhou <[e-mail]> for contributing the
          patch/changes necessary to make it work.

Wed Mar 21 07:23:15 2001  Douglas C. Schmidt  <[e-mail]>

        * tests/SOCK_Connector_Test.cpp (find_another_host): Make the
          test a bit more resilent against errors with gethostbyname().
          Thanks to Lu Yunhai <[e-mail]> for this fix.

Tue Mar 20 17:31:21 2001  Carlos O'Ryan  <[e-mail]>

        * tests/run_test.lst:
          The Logging_Strategy_Test dynamically loads the Logger service
          (from netsvcs), therefore it does not work on static builds.

Tue Mar 20 18:17:24 2001  Douglas C. Schmidt  <[e-mail]>

        * ace/OS.i: Fixed the USYNC_PROCESS arm of CreateMutex() in
          ACE_OS::mutex_init() so that it calls
          ACE_OS::set_errno_to_last_error().  Thanks to Ram Ben-Yakir
          <[e-mail]> for reporting this.

Tue Mar 20 01:33:24 2001  Ossama Othman  <[e-mail]>

        * ace/SSL/SSL_SOCK_Acceptor.cpp (ssl_accept):
        * ace/SSL/SSL_SOCK_Connector.cpp (ssl_connect):
        * ace/SSL/SSL_SOCK_Stream.i (send, close):

          Wrap the underlying OpenSSL calls in a do-while(SSL_pending())
          loop.  I decided to wrap the SSL_write() calls just in case.
          This should fix the last of the non-blocking IO issues in ACE's
          SSL wrappers.  [Bug 393]

Tue Mar 20 00:40:43 2001  Ossama Othman  <[e-mail]>

        * ace/SSL/SSL_SOCK_Stream.i (recv):

          Fixed non-blocking IO support for this method.  It was necessary
          to ensure that a full SSL record was read before returning
          control to the caller.  Note that the send() method doesn't need
          modification since OpenSSL always writes a full record before
          returning.  [Bug 393]

Mon Mar 19 19:53:34 2001  Carlos O'Ryan  <[e-mail]>

        * include/makeinclude/ace_flags.bor:
          Define macros for the RT Event Service test library

Mon Mar 19 16:57:43 2001  Irfan Pyarali  <[e-mail]>

        * ace/TP_Reactor.cpp (handle_events): Return value from
          handle_events() after the Reactor has been deactivated should be
          -1.  We just need to remember that as illustrated in
          ACE_Reactor::run_reactor_event_loop(), this is not an error:

          if (result == -1 && this->implementation_->deactivated ())
            return 0;

        * tests/Thread_Pool_Reactor_Test.cpp (svr_worker): Changed this
          function to call ACE_Reactor::run_reactor_event_loop() instead
          of ACE_Reactor::handle_events().  In the previous revision, we
          were not handling return values from
          ACE_Reactor::handle_events() correctly when
          ACE_Reactor::deactivated() had been called.

Mon Mar 19 09:14:12 2001  Chad Elliott  <[e-mail]>

        * include/makeinclude/platform_chorus4.x_g++.GNU:

          Added an output option to the link command.  Linking would fail
          in cases where the output was specified in another directory.

Sun Mar 18 22:12:16 2001  Ossama Othman  <[e-mail]>

        * ace/SSL/SSL_SOCK_Stream.i (close):

          Commented out debugging output that was inadvertently committed
          to the repository.

          Don't bother invalidating the handle.  Doing so can potentially
          cause some side-effects in the underlying OpenSSL
          implementation since setting the handle in the SSL_SOCK_Stream
          also causes the OpenSSL SSL_set_fd() function to be called.

          Fixed a socket leak caused by the fact that the handle was
          invalidated before the underlying ACE_SOCK_Stream was closed.

Sun Mar 18 09:46:47 2001  Balachandran Natarajan  <[e-mail]>

        * tests/New_Fail_Test.cpp: Fixed warnings in g++.

Sun Mar 18 08:31:34 2001  Douglas C. Schmidt  <[e-mail]>

        * apps/Gateway/Gateway,
        * apps/Gateway/Peer: Added a number of fixes to the Gateway and Peer
          applications.  Thanks to Lu Yunhai <[e-mail]> for
          contributing these.