Mon Sep 20 10:34:51 1999 Steve Huston * ACE-INSTALL.html: Fixed broken link to Riverace's HP-UX info. Thanks to Doug for finding this. Sun Sep 19 12:48:31 1999 Douglas C. Schmidt * tests/Signal_Test.cpp: Only send outselves a signal if we're running the signal handler in a separate thread. Sun Sep 19 10:39:30 1999 Douglas C. Schmidt * tests/Signal_Test.cpp: Changed the type of shut_down indicator from ACE_Atomic_Op<> to sig_atomic_t. This prevents some nasty deadlocks. * tests/run_tests.lst: Reenabled Signal_Test so we can see if it's still working. * tests/Signal_Test.cpp: Changed the worker_child() function so that just before its thread exits, it will sent its process a signal to the signal handling thread to dislodge it if it hasn't shutdown on its own accord yet. Thanks to David Levine for reporting this. This fixes [BUGID:309]. * ACE/Handle_Set.h: Updated the documentation to point out that it's the *caller's* responsibility to ensure handles aren't added/cleared from the set if their values are larger than ACE_DEFAULT_SELECT_REACTOR_SIZE. This behavior is necessary to ensure efficiency. * tests/Handle_Set_Test.cpp (test_boundaries): Fixed the test so that it will correctly ignore any handle values >= ACE_DEFAULT_SELECT_REACTOR_SIZE. Thanks to David Levine for reporting this. This fixes [BUGID:315]. Sat Sep 18 18:49:31 1999 Jeff Parsons * ace/config-win32-borland.h: Undefined ACE_HAS_BROKEN_IMPLICIT_CONST_CAST as suggested by Christopher Kohlhoff . Sat Sep 18 17:31:52 1999 Douglas C. Schmidt * examples/QOS: Improved the conformance to ACE programming guidelines a bit. * ace/Message_Block.cpp: Whenever a ACE_Message_Block accepts a ACE_Data_Block that was produced by someone else, it should verify that the allocator is non-NULL and if it is, then it should allocate the ACE_ALLOCATOR_RETURN on it. Thanks to Pierre Grondin for reporting this. * docs/tutorials/006: Fixed the activate() method so that it uses THR_DETACHED. Thanks to Shafiek Savahl for reporting this. Sat Sep 18 16:17:34 1999 Carlos O'Ryan * bin/auto_compile: The messages include the build name in the subject line. Thanks to Darrell for suggesting this idea. Sat Sep 18 12:00:28 1999 Douglas C. Schmidt * ace/Message_Block.h: Changed all the ACE_Message_Block constructors so that the priority parameter defaults to ACE_DEFAULT_MESSAGE_BLOCK_PRIORITY, rather than 0. Thanks to Detlef Becker for this suggestion. * ace/OS.h: Added a new macro called ACE_DEFAULT_MESSAGE_BLOCK_PRIORITY, which defaults to 0 but which can be overridden in a config.h file to a different value so it's possible to transparently choose default priorities that are > 0. Thanks to Detlef Becker for this suggestion. Fri Sep 17 19:49:15 1999 Kirthika Parameswaran * tests/Hash_Map_Manager_Test.cpp: Repositioned template instantiations and static variables to make the debug=0 build happy with SunCC5.0. Fri Sep 17 15:34:46 1999 Nanbor Wang * ace/Malloc.cpp (dump): * ace/Malloc_T.cpp (dump): Enriched various dump methods a bit. Thu Sep 16 17:53:57 CDT 1999 Loren Rittle * bin/update-ace+tao.sh: Added force option. Thu Sep 16 17:43:02 CDT 1999 Loren Rittle * bin/update-ace+tao.sh: Improve portability. Thu Sep 16 11:08:45 1999 Ossama Othman * ace/Makefile: * tests/Makefile: Updated dependencies. Wed Sep 15 21:42:08 1999 Carlos O'Ryan * ace/OS.h: Removed extra "typedef" in the declaration of the ACE_Protocol_Info structure, it was causing a large number of warnings in several platforms. Wed Sep 15 15:08:33 1999 Ossama Othman * bin/run_all_list.pm: Run the newly added TAO/tests/ORB_init/run_test.pl script when executing the ORB_init test. Wed Sep 15 14:49:50 1999 Balachandran Natarajan * ace/Hash_Map_With_Allocator_T.h: * ace/Hash_Map_With_Allocator_T.i: Added a method unbind () which unbinds an entry by only taking a key. Wed Sep 15 13:19:36 1999 Nanbor Wang * ace/Service_Config.cpp (process_directive): Added ace_yyrestart to make sure the service config parser has a valid buffer to work on. Thanks to Lothar Werzinger for reporting this. Wed Sep 15 10:50:21 1999 James Hu * ace/TP_Reactor.{cpp,h,i} (mask_ops): Fixed a problem where cancel_wakeup would fail. Since an event handler is suspended for the duration of the up-call, the wait_mask_ was being over-written by the suspend_mask_ when the event handler was resumed. This fixes the problem by extending TP_Reactor with its own implementations of mask_ops that operate on both wait_mask_ and suspend_mask_. Wed Sep 15 10:14:21 1999 Ossama Othman * docs/exceptions.html: Corrected some ACE_TRY_THROW examples. The exception being thrown must have a parameter list, even if it is empty (e.g.: ACE_TRY_THROW (CORBA::BAD_PARAM ());) Tue Sep 14 22:10:24 1999 Kristopher Johnson * ace/Process_Manager.cpp (wait): changed array index from i to idx. Thanks to Dave Madden for confirming this fix. Reverted this change: Tue Sep 14 16:01:04 1999 Christopher Kohlhoff because it's no longer necessary. Tue Sep 14 21:00:45 1999 Vishal Kachroo * ace/OS.h: Added a typedef for ACE_Protocol_Info for non-Win32 platforms. Defined it as a struct (with limited fields) instead of u_long because applications using ACE_Protocol_Info on non-Win32 platforms expect this to be a structure instead of a u_long. * ACE_wrappers/examples/QOS/Makefile: Added this for the test to compile on non-Win32 platforms. * ACE_wrappers/examples/QOS/server.cpp: * ACE_wrappers/examples/QOS/client.cpp: * ACE_wrappers/examples/QOS/QosEvent.h: * ACE_wrappers/examples/QOS/Sender_QOS_Event_Handler.cpp: Removed the last remains of win32isms from the test. The QoS test now compiles clean on non-Win32 platforms. Tue Sep 14 17:49:33 1999 Luther J Baker * ace/Arg_Shifter.cpp (get_the_parameter): removed an errant return 0; Tue Sep 14 16:01:04 1999 Christopher Kohlhoff * ace/Process_Manager.cpp (wait): moved declaration of loop index i out of the loop, because it is used outside the scope of the loop. Tue Sep 14 15:03:21 1999 Ossama Othman * bin/run_all_list.pm: Added ORB_init test to the list of tests to run. * configure.in: * configure: Do not define ACE_USE_POLL, even if system poll() works. Tue Sep 14 11:02:06 1999 Balachandran Natarajan * ace/Hash_Map_With_Allocator_T.h: * ace/Hash_Map_With_Allocator_T.cpp:Added a find () method that would lookup at the underlying Hash_Map_Manager_Ex to do the find. Sat Sep 11 12:18:22 1999 Douglas C. Schmidt * ace/Pipe.cpp: Make sure to use SOCK_STREAM rather than SOCK_DGRAM when creating a socketpair. Thanks to Dave Madden for reporting this. Mon Sep 13 13:31:23 1999 Luther J Baker * ace/Arg_Shifter.cpp (get_the_parameter): continued -ORB parsing fixes. Allows one simple method to return a parameter value for defined flag is it exists. Mon Sep 13 11:01:24 1999 Nanbor Wang * examples/Shared_Malloc/test_persistence.cpp: Added comments on the use of scanf. Thanks to Wenli Bai for reporting the problem. Sun Sep 12 19:23:51 1999 Nanbor Wang * ACE version 5.0.4 released. Sat Sep 11 12:18:22 1999 Douglas C. Schmidt * ace/Process_Manager.cpp: Reverted the following change so that we just add a fake "friend" to keep G++ happy... Also, fixed a typo where we should have allocated an ACE_Managed_Process rather than an ACE_Process. Thanks to Dave Madden for reporting this. * ace/Process_Manager.cpp: Added the ACE_CLASS_IS_NAMESPACE method to ACE_Managed_Process to make G++ happy. Also, removed the trailing '_' since we don't need this because we've prefixed "ACE_" already. Fri Sep 10 16:58:34 1999 Nanbor Wang * ace/Malloc_T.cpp (open,shared_malloc): Added several placement new's if position independent malloc is enabled to avoid reading uninitialized memory. Fri Sep 10 16:39:13 1999 Vishal Kachroo * examples/QOS/README: * examples/QOS/server.cpp: * examples/QOS/client.cpp: * examples/QOS/Sender_QOS_Event_Handler.h: * examples/QOS/Sender_QOS_Event_Handler.cpp: Removed the Win32-isms from the above files and used cleaner ACE interfaces. Added a README file that describes the test. Fri Sep 10 16:38:51 1999 Nanbor Wang * tests/WinCE/*.dsp: VC 6 changed all the dsp files format. Fri Sep 10 11:20:47 1999 Douglas C. Schmidt * ace/SOCK_Acceptor.h, ace/SOCK_Connector.h: Updated the documentation to clarify what happens wrt blocking vs. non-blocking connects and accepts. Thanks to Jose Rubio for motivating this change. Thu Sep 9 16:27:56 1999 Douglas C. Schmidt * ace/Process_Manager, ace/Process: Added some new patches from Dave Madden to obtain greater control over the spawn() behavior. Fri Sep 10 14:30:17 1999 Vishal Kachroo * ace/SOCK_Dgram_Mcast.i: * ace/SOCK_Dgram_Mcast.h: Added two send () functions using overlapped I/O to the ACE_SOCK_Dgram_Mcast class. These simply call the parent class functions with the same signature. The recv () functions need not be added as there is currently no recv () function within SOCK_Dgram_Mcast class to hide any parent class recv () functions. Fri Sep 10 12:24:17 1999 Carlos O'Ryan * THANKS: Added Joerg Pommnitz to the list of contributors. Thu Sep 9 22:48:56 1999 Darrell Brunsch * ace/config-hpux-10.x-hpc++.h: * ace/config-hpux-11.x-hpc++.h: Fixed spelling of ACE_PREPROCESSOR_ARGS to ACE_CC_PREPROCESSOR_ARGS. Thu Sep 9 18:22:21 CDT 1999 Carlos O'Ryan * ace/Future.cpp: HP/aCC was giving bogus warning about function without a return statement (the function only has a throw). Thu Sep 09 14:56:29 1999 David L. Levine * ACE-INSTALL.html: we now fully support QNX Neutrino 2.0. * ACE-INSTALL.html: expanded repo=1 discussion. [Bug 248] Thu Sep 09 13:21:03 1999 Nanbor Wang * ace/ace_ce.dsw: * ace/ace_ce_dll.dsp: Changed to use the project file format for CE Toolkit for VC 6. * ace/config-WinCE.h: Added some macros definitons for POSIX error codes. They are removed for Palm-size CE v2.01. This configuration is still not working because its compilers don't support placement new, which I don't know how to work around. * ace/Process_Manager.cpp (wait): Applied ASYS_TEXT macros. * ace/OS.cpp (init,fini): Applied ACE_TEXT macros around error message print out. (print_error_message): Added a WinCE specialized implementation. Thu Sep 09 12:45:06 1999 David L. Levine * ace/config-qnx-neutrino.h: added ACE_HAS_SIGISMEMBER_BUG. Sigset_Ops_Test passes with that. Thu Sep 09 12:31:51 1999 David L. Levine * tests/Handle_Set_Test (handle_vector[]): limit the maximum handle in handle_vector to 31, 63, 127, and 254 if ACE_DEFAULT_REACTOR_SIZE is less than or equal to 32, 64, 128, and 256, respectively. [Bug 315] Thu Sep 9 09:39:46 1999 Kirthika Parameswaran * etc/purify.solaris2: Added suppression for MLK due to OS accept call. [Bug 227] Thu Sep 09 09:24:29 1999 David L. Levine * tests/run_tests.lst: disabled Signal_Test. [Bug 309] Thu Sep 9 08:46:58 1999 Douglas C. Schmidt * tests/Lazy_Map_Manager_Test.cpp (main): ACEified this a bit and fixed a typo for the argv argument in main(). * Updated the ACE-mail.html file to put the *-request addresses first. Thanks to Lorin Hochstein for this suggestion. Wed Sep 08 23:01:22 1999 Nanbor Wang * examples/Service_Configurator/Misc/Timer_Service.h: Added ACE_Svc_Export specifiers. * examples/Service_Configurator/Misc/Misc.dsw: * examples/Service_Configurator/Misc/Timer.dsp: Added a new project file for the dll file. Wed Sep 08 22:19:47 1999 David L. Levine * ace/config-qnx-neutrino.h: added some magic to allow compilation to succeed when optimization is enabled, for x86 targets. The egcs 1.1.2 string.h #includes x86/string.h, which has anti-social #defines for memcmp, memcpy, and memset. Therefore, they can't be used as C++ member functions, such as in ACE_OS. The hack is to disable optimization just when #including string.h. * include/makeinclude/platform_qnx_neutrino.GNU: default both optimize and repo to 1. Wed Sep 8 17:25:07 1999 Douglas C. Schmidt * ace/OS.h (ACE_FACTORY_DEFINE): It shouldn't be necessary to use ACE_dynamic_cast() or ACE_static_cast() for the ACE_FACTORY_DEFINE macro because ACE_Service_Object's destructor is virtual. Thanks to Carlos for pointing this out. Wed Sep 8 16:51:45 1999 Kirthika Parameswaran * ace/Acceptor.cpp (handle_close): Removed a unnecessary check for which was causing UMR to occur. Wed Sep 08 16:48:55 1999 David L. Levine * include/makeinclude/wrapper_macros.gnu (purify): moved -ignore-signals option to tests/Makefile, because that's where it's needed. * tests/Makefile (purify): added SIGHUP and SIGTERM to -ignore-signals, to prevent warnings from Signal_Test. [Bug 309] Wed Sep 8 15:58:07 1999 Douglas C. Schmidt * tests/Log_Msg_Test.cpp (test_ostream): Make sure to add 1 to the buffer we allocate dynamically so we can put a NUL at the end of it! Thanks to David Levine for reporting this. [Bug 312] * examples/Service_Configurator/Misc/main.cpp (main): Changed the test of the -f Service Configurator option so that it links two different svc.conf files to avoid reentrancy problems. * examples/Service_Configurator/Misc/svc.conf: Revised this svc.conf file (and the application) to illustrate the use of both static *and* dynamic service. Thanks to Amir Bahmanyari for motivating this. Wed Sep 8 14:05:00 1999 Kirthika Parameswaran * ace/Log_Msg.cpp: Shifted the instantiations to the top to keep SunCC5.0 debug=0 build happy. * tests/Cached_Accept_Conn_Test.cpp: Repositioned global declarations to ward off compilation errors on HP-UX. Wed Sep 8 13:13:18 1999 Douglas C. Schmidt * ace/Process.h (ACE_Process): Made the spawn() method virtual so code can be executed just before and/or after the actual fork(). Thanks to Dave Madden for this fix. Tue Sep 07 19:55:52 1999 Irfan Pyarali * ace/Map_Manager.h (ACE_Map_Manager): The old map manager (the one before the changes made to support the active map manager) used to allow deletion of entries from the map during iteration (note that this feature was accidental since typically changes to container invalidates the iterators). The new design did not support this "feature". However, some users had come to depend on this feature. The solution was to delay the movement of freed slots from the occupied list to the free list until we run out of free slots in the free list when binding new entries into the map. However, this change requires additional state in each entry that keeps track of whether this entry has been freed. Also, the time required for binding new entries is less predictable since the bind may require moving of freed slots from the occupied list to the free list because of the initial lazy unbind. Changes were also required to the Active_Map_Manager since it directly depends on the internals of the Map_Manager. Since this lazy feature is not required in the majority of cases and results in increased memory consumption, this feature is only support if ACE_HAS_LAZY_MAP_MANAGER is defined. A new test Lazy_Map_Manager_Test.cpp was added to check the new changes. Thanks for Dr. Schmidt for helping with this solution and to Murphy Ivan for pointing out this problem. Bug fixed: id 228. Tue Sep 7 15:19:33 1999 Douglas C. Schmidt * tests/Log_Msg_Test.cpp (test_ostream): NUL-terminated the buffer to avoid an array bounds overflow! Thanks to David Levine for reporting this. Tue Sep 7 12:38:37 1999 Darrell Brunsch * tests/Reactor_Timer_Test.cpp: Purify was showing an UMR in test_registering_one_handler(). This was because Time_Handler didn't have a constructor that initialized its timer_id_ member and it wasn't being set in this method (like the others). I added a default constructor to Time_Handler. [Bug 249] Tue Sep 7 08:04:41 1999 Douglas C. Schmidt * ace/OS.i (isatty): Fixed this code so that it doesn't use osf_handle() anymore on NT, but instead just casts the ACE_HANDLE to an int. Thanks to Mark Winrock for suggesting this fix and testing it out. * tests/run_tests.lst: Added Signal_Test. 1999-09-07 Douglas C. Schmidt * tests/Signal_Test.cpp (main): Fixed the test so that it won't try to run the "synchronous signal handler in a separate thread" on Linux because Linux threads are non-standard wrt signals. Mon Sep 6 21:19:27 1999 Douglas C. Schmidt * tests/Signal_Test.cpp (run_test): Added an ACE_UNUSED_ARG for the synchronous_signal_handler pointer-to-function for the case where ACE_HAS_THREADS is disabled. Otherwise, we get a warning. Mon Sep 6 18:47:18 1999 Darrell Brunsch * ace/Process_Manager.h: Moved handle_signal outside of a !defined ACE_WIN32 block and moved handle_input inside of it, and this fixes some errors on MSVC. Mon Sep 6 14:19:14 1999 Douglas C. Schmidt * tests/Reactor_Notify_Test.cpp (main): Added a hyphen to the ACE_DEBUG printout "with timeout =" so that it says "with time-out =" to avoid confusing the run_tests.sh script! Thanks to David Levine for reporting this. * ace/Process_Manager.cpp: Integrated more patches from Dave Madden so that the correct register_handler() method is used on Win32. Mon Sep 06 15:05:17 1999 David L. Levine * ace/Makefile: moved ACELIB setting (to null) to before include of rules.lib.GNU. That way, rules.lib.GNU can use the value of $(ACELIB) to determine whether the ACE library object files should be linked into the library, with repo=1. [Bug 248] * include/makeinclude/rules.lib.GNU: with repo=1 when building static libs, except for libACE, include all of the ACE lib's object files. The prevents multiple instantiations, by GNU repo, of templates that are used in the ACE library. See the gcc info pages, Template Instantiation node, for more information. [Bug 248] Mon Sep 6 08:45:31 1999 Douglas C. Schmidt * examples/Reactor/WFMO_Reactor: Cleaned up some of the coding in the examples. * ace/Profile_Timer.cpp: Reformatted the code a bit. * examples/Threads/process_manager.cpp (main): Fixed some warnings and errors. Thanks to David for reporting this. * ace/OS.i: Fixed ACE_UNUSED_ARG(options) to be ACE_UNUSED_ARG(wait_options). Thanks to David's build logs for reporting this. * tests/Signal_Test.cpp: Changed the types of parameters passed to run_test() from int to long so that we can safely cast them to and from void *'s. Thanks to David Levine for pointing this out. * ace/Process_Manager.cpp (notify_proc_handler): Removed the now unused parameter h. Thanks to David for pointing this out. Mon Sep 06 07:25:01 1999 David L. Levine * ace/OS.h: added ACE_Export to ACE_OS_Exit_Info class declaration. It shouldn't be necessary, but Dmitry Goldshtain reports that Builder 4 needs it. Mon Sep 06 03:31:34 1999 Nanbor Wang * tests/Signal_Test.cpp (run_test): Removed a redundant use of ACE_UNUSED_ARG (synchronous_signal_handler). Thanks to KCC 3.3e for pointing this out. Mon Sep 6 01:03:03 1999 Douglas C. Schmidt * examples/Threads/Makefile (BIN): Added process_manager.cpp back into the list of programs that get built automagically. * ace/Process_Manager: Added some new patches from Dave Madden that continue to improve the ACE_Process_Manager and make it portable between NT and UNIX. * examples/Threads: Readded process_manager.cpp now that Dave Madden has updated it. Sun Sep 05 23:52:16 1999 Nanbor Wang * ace/OS.h (WIFEXITED,WEXITSTATUS): On platforms that do not define these macros, we should always assume the child process exit properly because the return status does not tell us how the process exited. Likewise, the WEXITSTATUS macro should simply return the status we got from ACE_OS::wait because it only contains the exit status of child process. Sun Sep 5 23:40:25 1999 Douglas C. Schmidt * examples/QOS: Continued to improve the coding style so it'll look like normal "ACE" code. Sun Sep 05 21:32:40 1999 David L. Levine * ace/Synch.cpp: removed unnecessary #include of ace/Object_Manager.h. 1999-09-05 Vishal Kachroo Made the receiver of QOS test portable by removing Win32-isms from it. Files modified : * examples/QOS/QosEvent.h: * examples/QOS/server.cpp: Sun Sep 5 13:40:18 1999 Michael Kircher * ace/ACE.cpp: Made the status code to be of type ACE_exitcode instead of int. This is necessary because of the change on waitpid (). Sun Sep 5 10:59:18 1999 Douglas C. Schmidt * ace/Process.i (wait): Removed the ACE_static_cast() Michael had added recently because the change below makes it unnecessary. * ace/OS.h (ACE_OS): Modified the 3-arg wait() and the waitpid() methods to take an ACE_exitcode* rather than int* so this will work on Win32. * ace/Log_Msg.cpp: Added a new explicit template instantiation for template class ACE_Guard >; to fix compiler errors that showed up with EGCS. * tests/Signal_Test.cpp (run_test): Added ACE_UNUSED_ARG macros for several symbols that aren't used when the test is compiled single threaded. Sat Sep 05 02:05:00 1999 Michael Kircher * ace/OS.cpp: Fixed a missing "#else". * ace/Process.i, ace/Process_Manager.h, ace/Process_Manager.cpp: Fixed an "int" versus "ACE_exitcode" conversion problem. Sat Sep 4 23:37:29 1999 Douglas C. Schmidt * ace/config-win32-common.h: Missing a '&&' that was causing problems on Win32. Thanks to Michael Kircher for reporting this. Sat Sep 04 20:24:32 1999 David L. Levine * ace/Handle_Gobbler.i (close_remaining_handles,free_handles): moved the function definitions ahead of their first use in the file. Sat Sep 04 20:17:49 1999 David L. Levine * ace/Pipe.cpp (open): added ACE_UNUSED_ARG (buffer_size) for all platforms that #define ACE_LACKS_SOCKET_BUFSIZ. Sat Sep 4 14:43:36 1999 Douglas C. Schmidt * tests/Signal_Test.cpp: Finished off the test so that it now works for all the various combinations of synchronous and asynchronous tests. Now, to make it work on Linux... * ace/OS.h: The use of ACE_static_cast in the ACE_FACTORY_DEFINE macro didn't make any sense, so I've replaced it with ACE_dynamic_cast. * ace/Process_Manager.*, * ace/Process.*, * ace/OS.h: Created a new typedef called ACE_exitcode that factors out the difference between Win32 and UNIX exit codes from wait() operations. Thanks to Dave Madden for suggesting this. Sat Sep 04 12:22:38 1999 David L. Levine * ace/OS.i (shm_open,close,filesize,ftruncate,mmap): moved these functions up in the file so that they're defined before being used by ACE_OS::sema_init (). Fri Sep 3 11:57:24 1999 Douglas C. Schmidt * ace/Event_Handler.h (ACE_Event_Handler): Added the new handle_exit() method, which is called by the ACE_Process_Manager when a process exits. Thanks to Dave Madden for contributing this. * ace/Process.*, ace/Process_Manager.*: Integrated a new round of changes from Dave Madden that factor out most of the remaining differences between Win32 and UNIX. * ace/OS.h, * ace/config-win32-common.h: Added a #ifndef for __IBMCPP__ to make things work for Visual Age C++. Thanks to Boris Kaminer for reporting this. * ace/{ACE,OS}: Added ACE_HAS_UNICODE to conditionally compile the strnew() definition. Thanks to Boris Kaminer for reporting this. * ace: Added a new optimized ace-dll.icc file so that ACE will build better with IBM's Visual Age C++. Thanks to Boris Kaminer for contributing this. * ace/OS.h, * ace/config-win32-common.h: Added a #ifndef for __IBMCPP__ to make things work for Visual Age C++. Thanks to Boris Kaminer for reporting this. * ace/{ACE,OS}: Added ACE_HAS_UNICODE to conditionally compile the strnew() definition. Thanks to Boris Kaminer for reporting this. * ace: Added a new optimized ace-dll.icc file so that ACE will build better with IBM's Visual Age C++. Thanks to Boris Kaminer for contributing this. 1999-09-03 Vishal Kachroo * ace/OS.i: * ace/OS.h: Added a new overloaded ACE_OS::ioctl () function that takes in ACE_QoS as one of the parameters, constructs a windows QOS from it and passes it to WSAIoctl (). This function can be used when the I/O control code is SIO_SET_QOS. Fri Sep 3 14:58:51 1999 Darrell Brunsch * config-aix-4.x.h: * config-hpux-10.x-hpc++.h: * config-hpux-11.x-hpc++.h: * config-mvs.h: * config-win32-borland.h: * config-win32.h: Moved many of the preprocessor information for various platforms from TAO_IDL to here (in ACE_CC_PREPROCESSOR and ACE_CC_PREPROCESSOR_ARGS). Fri Sep 3 13:52:43 1999 Carlos O'Ryan * bin/auto_compile: Fixed the implementation of the -dont_run option. Fri Sep 3 09:05:20 1999 Kirthika Parameswaran * ace/Containers_T.cpp (delete_nodes): Removed the redundant check which was verifying that was always > 0. But itself is unsigned so thsi check was redundant causing warnings with egcs. Thu Sep 2 23:54:49 1999 Nanbor Wang * ace/Containers_T.h (class ACE_Double_Linked_List): Added more comments to stress that this is not a general purpose container class. Thanks to Rusty Conover for suggesting this. Thu Sep 2 18:39:23 1999 Douglas C. Schmidt * ace/Process_Manager.cpp (handle_close): Did a better job of fixing the use of ACE_UNUSED_ARG to make EGCS happy. Thu Sep 2 12:21:41 1999 Douglas C. Schmidt * ace/Select_Reactor_Base: Renamed the notify_set_, alloc_set_, and free_set_ to *_queue_ since that's what they really are! Thu Sep 2 17:42:22 1999 Carlos O'Ryan * ace/Malloc_T.cpp: Eliminated UMR problem with the position independent malloc support. Thanks to Doug for figuring this stuff out. Thu Sep 2 12:21:41 1999 Douglas C. Schmidt * tests/Log_Msg_Test.cpp (test_log_msg_features): Changed the use of %p to prevent an error message from being printed that'll confuse the run_tests script. * Just added the 700th non-DOC group individual who has contributed to ACE+TAO over the years. In the last year, we've gone from 400 to 700 contributors, which is great! * tests/CDR_File_Test.cpp (main): Fixed a signed/unsigned warning and removed the unreferenced argc/argv parameters. * tests{run_tests.lst,Makefile}: Added the new ACE_Log_Msg.cpp test. * tests: Added a new test called Log_Msg_Test.cpp. This exercises the many features of the ACE_Log_Msg class. * tests/Object_Manager_Test.cpp (main): Just reformatted some stuff. Thu Sep 2 14:42:31 1999 Kirthika Parameswaran * tests/Cached_Accept_Conn_Test.cpp: Repositioned ACE_HAS_BROKEN_EXTENDED_TEMPLATES macro before the inclusion of the header since thats where the typedefs had been shifted. Also, to keep both gcc2.7.3 and sunCC5.0 w.o debugging symbols happy it was necessary to move the definitions of methods before the instantiations. Thu Sep 2 12:21:41 1999 Douglas C. Schmidt * ace/OS.h: Added an ACE_static_cast in the ACE_FACTORY_DEFINE macro to let the compiler catch "bad casts," i.e., when X isn't allow to be cast from an ACE_Service_Object. Thanks to Jody Hagins for suggesting this. * ace/config-win32-common.h: Deleted #define ACE_LACKS_STRPTIME and added #define ACE_HAS_STRPTIME #define ACE_LACKS_NATIVE_STRPTIME In addition, I removed the wday and yday stuff in ACE_OS::strptime(). Thanks to Kevin Lyda for suggesting this. * ace/OS.h: Make sure to add a cast to ACE_Service_Object * in the ACE_FACTORY_DEFINE macro so that the offset will be corrected automagically. Thanks to Eric C. Newton for reporting this and suggesting a fix. * ace/Log_Msg: Modified the msg_callback(ACE_Log_Msg_Callback *) method to return the existing ACE_Log_Msg_Callback * so we can support "chaining". * ace/Log_Msg.cpp: Added explicit template instantiations for ACE_Reverse_Lock. * ace/Log_Msg.cpp: Changed the implementation described below to use an ACE_Reverse_Lock, which is cool ;-). Thanks to Carlos for explaining how this worked. * ace/Log_Msg.cpp: Don't hold the lock during a msg_callback() to avoid deadlock. Thanks to Matthew Davis for reporting this. * tests/Cached_Accept_Conn_Test.cpp: Removed a "static" from int debug so that the friggin' SunC++ compiler will link... Thu Sep 02 11:04:18 1999 David L. Levine * bin/check_build_logs: added /dev/null to egrep invocation so that it always shows the file name. Thanks to Doug for reporting this. Thu Sep 02 11:02:25 1999 David L. Levine * ace/config-qnx-neutrino.h: added ACE_LACKS_SOCKET_BUFSIZ. Thu Sep 2 08:56:39 1999 Jeff Parsons * ace/CDR_Stream.h: * ace/CDR_Stream.i: Added << and >> operators for ACE_CDR::Char. These are not to be used by TAO or the IDL compiler, they are solely for use with the ACE library, so that 1-byte values can be streamed without using the CORBAisms to_char and from_char. Otherwise, if a 1-byte value were on the rhs of a CDR operator, the compiler will probably promote the operator to the one using ACE_CDR::Short, which is not at all what we want. Wed Sep 1 21:37:59 1999 Douglas C. Schmidt * tests/Makefile (BIN): Added the new CDR_File_Test.cpp to the list of files that are automatically built. Also, added this to run_tests.lst. * tests: Added a new test called CDR_File_Test, which exercises the ACE_CDR_Stream features in conjunction with file I/O. Thanks to Giga Giguashvili . * tests/CDR_Test.cpp: Moved the main() function to the end of the file. * ace/OS.i: Moved the new ACE_OS::filesize() method down in this file until *after* the ACE_OS::close() method to prevent a "used before being inlined" error on DU. Thanks to David for reporting this. Wed Sep 1 20:41:03 1999 Carlos O'Ryan * tests/run_tests.lst: Disabled the Time_Service_Test, it was crashing the parent process. Wed Sep 01 19:46:30 1999 David L. Levine * bin/check_build_logs: allow log directory (-l option argument) to start with anything except -. Wed Sep 1 17:15:25 1999 Kirthika Parameswaran * ace/Arg_Shifter.cpp (get_current_parameter): Removed egcs warnings by making int and also cleaned up the code. Wed Sep 1 15:51:41 1999 Douglas C. Schmidt * examples/Timer_Queue/Thread_Timer_Queue_Test.cpp: Fixed a typo where ACE_reinterpret_cast() was misspelled. Thanks to David for reporting this. * examples/QOS/server.cpp: "ACE"-ified this example. * ace/OS: Added a new method called ACE_OS::filesize() that takes a LPCTSTR so that we can pass a file name to this. * ace/config-win32-borland.h: Changed ACE_CC_COMPILER to CPP32.EXE on the recommendation of Christopher Kohlhoff . * ace/Synch.h: Updated the documentation of ACE_Semaphore::remove() to clarify that it should only be called by one thread to avoid race conditions. Thanks to Clarence Bishop for mentioning this. 1999-09-01 Vishal Kachroo * ACE_wrappers/ace/OS.h: * ACE_wrappers/ace/OS.i: Added a wrapper ACE_OS::enum_protocols () for WSAEnumProtocols(). Tested this function with the ACE QOS test. * examples/QOS/client.cpp: * examples/QOS/server.cpp: Replaced the occurences of WSAEnumProtocols with ACE_OS::enum_protocols () and successfully ran the QOS test. Wed Sep 1 15:51:41 1999 Douglas C. Schmidt * ace/OS.i: Ensure that ACE_OS::localtime_r() does the "Right Thing"[TM], i.e., returns the pointer in the buffer provided by the user, even if there's no native support for localtime_r() on the OS platform. In addition, added an ACE_OS_GUARD for this use-case so that we don't have race conditions in MT programs. This necessitated moving the localtime_r() wrapper facade into the OS.cpp file. Thanks to Jody Hagins for these suggestions. * ace/SV_Semaphore_Complex: Improved the documentation a bit based on comments from Herbert . Wed Sep 1 15:15:00 1999 Luther Baker * ace/Arg_Shifter.h: Added function (get_current_parameter(offset) to return parameters that may be mangled with flags. * ace/Arg_Shifter.cpp: Implemented function above. Wed Sep 1 14:50:48 1999 Nanbor Wang * ace/OS.h: Added definitions for _O_BINARY, _O_TEXT, and _O_RAW if they are not defined. Thanks to Kevin Lyda for pointing this out. Wed Sep 1 14:30:58 1999 Darrell Brunsch * ace/Process_Manager.cpp: Typo fixed. Wed Sep 1 11:18:26 1999 Douglas C. Schmidt * tests/Reactor_Notify_Test.cpp: Fixed an erroneous use of = rather than != and added another pair of explicit template instantiations for ACE_Auto_Basic_Ptr<>. Thanks to Carlos for reporting this. * ace/Process_Manager.cpp (handle_close): Removed a warning about unused parameter. Thanks to David for reporting this. * tests/Reactor_Notify_Test.cpp: I needed to add #elif defined (ACE_HAS_TEMPLATE_INSTANTIATION_PRAGMA) rather than #else Wed Sep 01 11:55:31 1999 Steve Huston * ace/CORBA_Handler.{h cpp} (ACE_ST_CORBA_Handler): Made the ACE_ST_CORBA_Handler instance get cleaned up via object manager at rundown, giving it a chance to clean up it's reactor registrations and reset its Orbix callbacks. Also added a new capability of allowing the ACE_ST_CORBA_Handler to daisy-chain the Orbix callbacks - this is only enabled if ACE_TAKEOVER_ORBIX_CALLBACKS is set in the config.h file, and its use is highly discouraged unless you really know what you're doing and can account for all of the callbacks. Wed Sep 1 11:18:26 1999 Douglas C. Schmidt Thanks to David for reporting this. * tests/run_tests.lst: Disabled the Process_Manager_Test until it's stable. Thanks to David for reporting this. Wed Sep 1 00:05:04 1999 Douglas C. Schmidt * ace/config-win32-borland.h: Added a #define for ACE_CC_COMPILER so that things will work properly on Borland. Thanks to Goldshtain Dmitry and Adrian Mercieca for reporting this. Wed Sep 01 10:45:15 1999 David L. Levine * ace/OS.{h,cpp} (ACE_OS_Object_Manager::init,fini): replaced use of ACE_ERROR with a new, private method (print_error_message). If one of the operations in init () fails, then the Log_Msg instance might not work properly. Thanks to Alain Magloire for noting this. print_error_message also consolidates the common const char strings into one place. This entire changes reduces the size of OS.o by about 900 bytes. Wed Sep 1 00:05:04 1999 Douglas C. Schmidt * tests/Reactor_Notify_Test.cpp: Only run the iterations in this test for 2000 times, not 10000 times. Wed Sep 1 07:41:58 1999 Jeff Parsons * ace/Process_Manager.cpp (wait): An ACE_NEW_RETURN needed ACE_INVALID_PID for the return value instead of ACE_INVALID_HANDLE. Wed Sep 1 00:05:04 1999 Douglas C. Schmidt * tests/Process_Manager_Test.cpp: Made some minor changes in this test so that it'll compile. We still need to update it so it'll run correctly with the new version of ACE_Process_Manager, however. * ace/Process_Manager: Integrated the latest and greatest ACE_Process_Manager from Dave Madden . This should now work relatively transparently on all OS platforms that support processes. We'll be adding a new Process_Manager_Test.cpp shortly... * tests/Reactor_Notify_Test.cpp: Modified this test to exercise the new user-level notification queueing in the ACE_Reactor. Tue Aug 31 16:05:14 1999 Douglas C. Schmidt * ace/Process_Manager.cpp (handle_close): Added an ACE_UNUSED_ARG to keep the compiler from warning that the "handle" parameter isn't used. * ace/OS.h: Added a new #defined called ACE_REACTOR_NOTIFICATION_ARRAY_SIZE that designates the size of the array of ACE_Notification_Buffers used by the new Reactor user-level notification queue feature. * ace/Select_Reactor_Base: Added support for a user-level notification queue that can buffer very large amounts of pending notifications, i.e., well beyond the 64 kbyte limit imposed by the use of a socket. This feature is only enabled if ACE_HAS_REACTOR_NOTIFICATION_QUEUE is enabled in a config.h file. Thanks to Detlef Becker for suggesting this and contributing a prototype implementation. Tue Aug 31 20:10:49 1999 Yamuna Krishnamurthy * ace/OS.h: Added the ACE_HAS_NO_THROW_SPEC check in order to suppress the generation of throw specs to placate some compilers and make it easier for CORBA developers to by lazy wrt C++ exception throw specs ;-). Thank to Andreas Geisler for suggesting this addition 1999-08-31 Vishal Kachroo Modified the ACE GQoS test to use a single handler registered for QOS_MASK|READ_MASK instead of two separate handlers. Had to remove the following files and add other files. * ACE_wrappers/examples/QOS/QOS_Event_Handler.cpp * ACE_wrappers/examples/QOS/Read_Handler.h * ACE_wrappers/examples/QOS/QOS_Event_Handler.h * ACE_wrappers/examples/QOS/Read_Handler.cpp Added the following files : The Sender event handler is only registered for QOS events. It uses the ACE_OS::sendto () as soon as it receives a QOS event. * ACE_wrappers/examples/QOS/Sender_QOS_Event_Handler.cpp * ACE_wrappers/examples/QOS/Sender_QOS_Event_Handler.h The Receiver event handler is registered for both QOS as well as READ events since it has to initially wait for a PATH RSVP message from the sender (QOS event, finally!!) and then read from the same QOS enabled socket. * ACE_wrappers/examples/QOS/Receiver_QOS_Event_Handler.cpp * ACE_wrappers/examples/QOS/Receiver_QOS_Event_Handler.cpp The client and server have been modified to subscribe to the same multicast session. The RSVP SP on the receiver side sends RESV messages at the earliest indication of QOS parameters AND existence of a matching PATH state. This state is found only if a multicast socket has been created (by the sender) with a matching "multicast session" address. * ACE_wrappers/examples/QOS/server.cpp: * ACE_wrappers/examples/QOS/client.cpp: Modified the following to incorporate the removal and addition of files listed above. * ACE_wrappers/examples/QOS/QOS.dsw * ACE_wrappers/examples/QOS/client.dsp * ACE_wrappers/examples/QOS/server.dsp Tue Aug 31 16:05:14 1999 Douglas C. Schmidt * ace/Malloc_T, * ace/Malloc: For historical reasons, we'd had a nested class called ACE_Malloc_Control_Block within ACE_Malloc_Header. This was redundant and confusing, so I merged the fields of ACE_Malloc_Control_Block into ACE_Malloc_Header. In addition, I also added a constructor that initialized the fields to 0 so that we won't have problems with Purify complaining about uninitialized memory reads. Thanks to Carlos for reporting this. Tue Aug 31 15:54:56 1999 Douglas C. Schmidt * ace/Process_Manager.h: There was a bizarre default value for the pid parameter, which was helping to cause ambiguity warnings from the C++ compiler. I've removed this. Tue Aug 31 00:16:34 1999 Douglas C. Schmidt * build/egcs/tests/Cached_Accept_Conn_Test.h: Moved the template class into the header file so that IBM C++ will be happy. * tests/Cached_*_Test.h: Updated these files to include the proper descriptive information. * tests/Upgradable_RW_Test.h: Cleaned up the formatting for this test. * tests/Cached_Accept_Conn_Test.cpp, * tests/Cached_Conn_Test.cpp, * tests/Conn_Test.cpp, * tests/DLList_Test.cpp, * tests/Hash_Map_Manager_Test.cpp: Fixed some "order of definition" problems caused by the workarounds for SunC++ 5.0 -g bugs. Also replaced the use of ADDR and STRING with ACE_ADDR and ACE_STRING to avoid potential clashes with symbols in the global name space. Thanks to David for reporting this. 1999-08-31 Vishal Kachroo * ACE_wrappers/examples/QOS/server.cpp * ACE_wrappers/examples/QOS/client.cpp * ACE_wrappers/examples/QOS/QOS_Event_Handler.cpp * ACE_wrappers/examples/QOS/QosEvent.h * ACE_wrappers/examples/QOS/Read_Handler.h * ACE_wrappers/examples/QOS/QOS_Event_Handler.h * ACE_wrappers/examples/QOS/Read_Handler.cpp * ACE_wrappers/examples/QOS/QOS.dsw * ACE_wrappers/examples/QOS/client.dsp * ACE_wrappers/examples/QOS/server.dsp: These files test the new ACE GQoS features, which implement IntServ features over WinNT 2000. Tue Aug 31 14:47:54 1999 Kirthika Parameswaran * ACE-INSTALL.html: Added a note on building egcs with the -fsquangle option contributed by Skye Sweeney . Tue Aug 31 00:16:34 1999 Douglas C. Schmidt * bin/clone.cpp: Modified this program so that it'll work more flexibly with symbolic links. In particular, clone can now generate symbolic links with relative pathnames to the source. Thus, you can now call it like this: % clone -s ../../foo . and all of the cloned links will properly refer to ../../foo (adding the appropriate number of extra '..'s to handle directory nesting). Thanks to Jonathan Biggar for these fixes. * ace/Process.cpp: Need to #include "ace/Signal.h". * Added a series of fixes that allow ACE to build and run cleanly with SunC++ 5.0 and Solaris 7 even when debugging is NOT enabled. These changes require moving the explicit template instantiations from the BOTTOM of the *.cpp files to near the top. Thanks to Robert Shewan for reporting this. * ace/Process_Manager.cpp: Fixed some warnings. Thanks to Carlos for reporting this. Mon Aug 30 21:14:46 1999 Douglas C. Schmidt * ace/Process_Manager.cpp: Added the destructor for ACE_Process_Descriptor. Thanks to David for reporting this. Mon Aug 30 17:20:56 1999 David L. Levine * ace/config-qnx-neutrino.h: added ACE_LACKS_MUTEXATTR_PSHARED. * ace/OS.i (mutex_init): enable the ACE_UNUSED_ARG (type) statement with ACE_LACKS_MUTEXATTR_PSHARED. Mon Aug 30 17:10:00 1999 Kirthika Parameswaran * ACE-INSTALL.html: Added a note about the "-fsquangle" option to be used with egcs to shorten long symbol names. Mon Aug 30 16:12:36 1999 Douglas C. Schmidt * ace/Process.i: Oops, the implementation of gethandle() was in a Win32-specific part of the code... * ace/Msg_WFMO_Reactor.h: Fixed a problem that was causing ACE_Msg_WFMO_Reactor from being included in the documentation windex files. Thanks to Jody Hagins for reporting this. Mon Aug 30 16:16:46 1999 Ossama Othman * PROBLEM-REPORT-FORM: Updated to point to refer reports ace-bugs mailing list. Mon Aug 30 14:27:26 1999 David L. Levine * tests/Cached_Accept_Conn_Test.cpp,Cached_Conn_Test.cpp, Map_Manager.cpp: use the explicit template instantiations even with ACE_HAS_GNU_REPO. g++ 2.91.66 misses some of them with -frepo. Mon Aug 30 15:23:12 1999 Douglas C. Schmidt * ace/Process_Manager: Integrated the new features that make it possible to notify a Reactor when a process exits. Thanks to Dave Madden for contributing this. * ace/Process: Added a more portable timed wait() operation for UNIX platforms. However, this method has the following two limitations: (1) on UNIX platforms this function uses , i.e., it overwrites any existing alarm and (2) it steals all SIGCHLDs during the timeout period, which will break another ACE_Process_Manager in the same process that's expecting SIGCHLD to kick off process reaping. If you know a good solution to this please let us know. Thanks to Dave Madden for contributing this. * ace/Process: Added setgroup() and getgroup() methods to ACE_Process_Options so that we can manage groups of processes. Thanks to Dave Madden for contributing this. * examples/OS/Process/process.cpp: Added a new version that tests the various wait() methods on ACE_Process. Thanks to Dave Madden for contributing this. * config-chorus.h, config-cray.h, config-cygwin32-common.h, config-freebsd-pthread.h, config-freebsd.h, config-hpux-9.x.h, config-hpux-9.x.h, config-lynxos.h, config-netbsd.h, config-sunos4-g++.h, config-sunos4-lucid3.2.h, config-sunos4-sun3.x.h, config-sunos4-sun4.1.4.h, config-sunos4-sun4.x-orbix.h, config-sunos4-sun4.x.h, config-win32-common.h: Added ACE_LACKS_SETPGID on the assumption that platforms lacking getpgid() won't have setpgid() either. If this turns out to be incorrect, please let me know. * ace/OS: Added a new wrapper facade method for setpgid(). Thanks to Dave Madden for contributing this. * ace/OS.h: Added a new macro called ACE_INVALID_PID. Thanks to Dave Madden for contributing this. * ace/Local_Tokens.h: Added a comment explaining that these locking classes aren't intended as general-purpose synchronization mechanisms. Thanks to Brian Wright * examples/Misc/test_trace.cpp (main): Added a call to ACE_OS::atexit() to make sure we test this someplace. Mon Aug 30 12:11:06 1999 Carlos O'Ryan * ace/Strategies.i: Once a Recyclable object reaches the ACE_RECYCLABLE_CLOSED state its state cannot change. This eliminates a nasty race condition where one thread detects that the connection is closed (and sets the state accordingly), but another thread changes the state to IDLE, based on old information. Thanks to Mark L Boriack for reporting this problem and helping us debug it. Mon Aug 30 10:43:08 1999 Carlos O'Ryan * ace/config-irix6.x-common.h: Added ACE_HAS_SIGSUSPEND and ACE_HAS_TIMEDWAIT Mon Aug 30 10:18:52 1999 Ossama Othman * PROBLEM-REPORT-FORM: Fixed typo. * acconfig.h: * configure.in: Removed ACE_LACKS_ATEXIT test since the macro no longer exists. Added test for ACE_HAS_SIGSUSPEND. Mon Aug 30 07:38:11 1999 David L. Levine * tests/Message_Block_Test.cpp, examples/Reactor/Misc/test_event_handler_t.cpp: use the explicit template instantiations even with ACE_HAS_GNU_REPO. g++ 2.91.66 misses some of them with -frepo. Sun Aug 29 22:57:26 1999 Douglas C. Schmidt * tests/Signal_Test.cpp (worker_parent): Revised the test so that we transmit the parent process ID in argv so that the child doesn't have to "guess." This solves the annoying problems with Linux threads... Sun Aug 29 21:35:17 1999 David L. Levine * ace/config-g++-common.h: 1) Don't #define ACE_HAS_EXPLICIT_INSTANTIATION with ACE_HAS_GNU_REPO, so that -frepo can do its thing. [Bug 248] 2) #undef ACE_HAS_STRING_CLASS with ACE_HAS_GNU_REPO, because egcs 2.91.66 can't build IOStream test (due to an unresolved basic_string method instantiation) otherwise. * ace/config-aix-4.x.h,ace/config-cygwin32-common.h, ace/config-hpux-10.x-g++.h,ace/config-irix5.3-g++.h, ace/config-irix6.x-g++.h,ace/config-linux-common.h, ace/config-lynxos.h,ace/config-osf1-4.0.h,ace/config-sunos4-g++.h, ace/config-sunos5.4-g++.h,ace/config-sunos5.5.h, ace/config-unixware-2.*-g++.h: moved ACE_HAS_STRING_CLASS #define before #include of config-g++common.h, so that config-g++-common.h can #undef it with ACE_HAS_GNU_REPO. * include/makeinclude/rules.lib.GNU: when ace_lib_prelink is enabled, add a link step before building a library. This is used, for example, to insert template instantiations into object files. [Bug 248] * include/makeinclude/wrapper_macros.GNU: set ace_lib_prelink to 1 when repo is enabled. Added ACE_NUL and ACE_NUL_STDERR macros, for internal use to support the prelink step. [Bug 248] Sun Aug 29 07:53:33 1999 Douglas C. Schmidt * tests/Signal_Test.cpp: Linux threads are problematic because getppid() isn't the process id of your parent PROCESS, it's the "process" id of your parent THREAD. Yikes! Therefore, I had to rearrange this test so that it'll work on Linux. * ace/config-qnx-neutrino.h, * ace/config-vxworks5.x.h, * ace/config-osf1-4.0.h, * ace/config-sunos5.5.h, * ace/config-linux-common.h: Added support for the ACE_HAS_SIGTIMEDWAIT and ACE_HAS_SIGSUSPEND macros. * ace/OS: Added a new sigsuspend() wrapper to class ACE_OS. This implementation is "smarter" than the default behavior of sigsuspend(2) because it'll handle NULL sigset_t *'s gracefully. * ace/Signal: Make the use of ACE_Sig_Set & parameters const-correct. * ace/Signal: Added a new method for ACE_Sig_Action that'll register an ACE_Sig_Set's worth of signals simultaneously. * ace/Signal: Updated the documentation to explain what the various constructors do. * ace/Process.h: Updated the documentation for ACE_Process::spawn() to explain what the return values are. * tests/Makefile: Added Signal_Test to the list of tests that are built automatically. * tests: Added a new test, Signal_Test.cpp, that exercises the various signal handling capabilities supported by ACE on various OS platforms. * ace/OS.h, * ace/README, * ace/config-vxworks5.x.h: Removed all mention of the ACE_LACKS_ATEXIT feature. This is now the default behavior because we ACE_OS_Object_Manager::at_exit() to get consistent semantics, as described below. Thanks to David for pointing this out. * ace/OS.i: Changed the implementation of ACE_OS::atexit() to always use the ACE_OS_Object_Manager::instance ()->at_exit() method so that we get consistent semantics across platforms. Sun Aug 29 18:27:38 1999 Ossama Othman * ACE.ifnames: * ace-diff-config.in: Removed ace-diff-config.in and ACE.ifnames from the repository. They are only used for internal ACE Configuration Project testing. * Makefile.am: * configure.in: Removed references to ace-diff-config.in and ACE.ifnames. * acconfig.h: * configure.in: Added entries and tests for the following macros: ACE_AUTO_PTR_LACKS_RESET ACE_HAS_LLSEEK ACE_HAS_EXPLICIT_KEYWORD ACE_HAS_MUTABLE_KEYWORD ACE_HAS_STD_TEMPLATE_SPECIALIZATION ACE_HAS_STD_TEMPLATE_METHOD_SPECIALIZATION ACE_HAS_SIGTIMEDWAIT ACE_LACKS_ATEXIT ACE_LACKS_GETPPID * configure.in: Fixed auto_ptr test. auto_ptr was being used incorrectly. Sun Aug 29 07:53:33 1999 Douglas C. Schmidt * tests/Process_Manager_Test.cpp: Changed the names of the signal handlers to one_child_synchronous_signal_handler() and multiple_children_synchronous_signal_handler() to reflect how they are used. * ace/OS.[hi]: Added a new wrapper for sigtimedwait() so that we can encapsulate this. To make programming more consistent, the encapsulation uses an ACE_Time_Value, rather than a struct timespec. * ace/config-linux-pthread.h, * ace/config-linux-lxpthreads.h: Removed the #define ACE_HAS_SIGWAIT macro since this was already defined in config-linux-common.h. * tests/Malloc_Test.cpp: Improved the test by printing out additional diagnostic information. Also validated that it works when ACE_DEFAULT_BASE_ADDR == 0 on Solaris using SunCC (with purify) and EGCS. * ace/Malloc: Added a dump() method to ACE_Malloc_Header. * ace/Process_Manager.cpp (ACE_Process_Descriptor): Changed a -1 to a 0 to prevent a compiler warning. Thanks to David Levine and DEC UNIX for reporting this. Sun Aug 29 00:14:09 1999 Luther Baker * ace/Arg_Shifter.h (cur_arg_strncasecmp): Removed and replaced earlier method 'is_or_contains_ignore_case' with 'cur_arg_strncasecmp.' This method does less and it is more self-contained, adding protection and greater flexibility for the program program using it. * ace/Arg_Shifter.cpp (cur_arg_strncasecmp): Implementation corresponding with changes in ACE_Arg_Shifter class. Sun Aug 29 00:13:35 1999 Douglas C. Schmidt * ace/Process_Manager.i: Moved the destructor for ACE_Process_Descriptor out of the *.i file since it's always called by vector delete and therefore inlining it is pointless. Sat Aug 28 23:46:55 1999 Douglas C. Schmidt * ace/Process_Manager.cpp (resize): The loop should run from 0 < this->current_count_ rather than 0 < this->max_process_table_size_. Sat Aug 28 16:28:33 1999 David L. Levine * ace/Synch_T.i: reordered some methods to avoid use before definition. Fri Aug 27 18:52:00 1999 Luther Baker * ace/Arg_Shifter.h (is_or_contains_ignore_case): added method to facilitate CORBA spec in ORB_init. * ace/Arg_Shifter.cpp (is_or_contains_ignore_case): implemented the method added. Part of a larger project to implement CORBA spec regarding -ORB parsing in ORB_init. Fri Aug 27 18:58:50 1999 Douglas C. Schmidt * tests/run_tests.lst: Reenabled the Process_Manager_Test since it now seems to work fine with EGCS after the fix noted below. Fri Aug 27 18:39:14 1999 Douglas C. Schmidt * ace/Process_Manager.cpp (wait): Fixed a mistake in the coding logic whereby we were returning from the function in a loop even when there was no error. Thanks to David Levine and DEC UNIX for pointing this out. Fri Aug 27 16:05:49 1999 David L. Levine * ace/OS.*,Log_Msg.cpp,Synch_T.i: reverted changes of Fri Aug 27 11:29:21 1999 David L. Levine to see if they caused problems on NT. Thanks to Michael and Jeff for reporting this. Fri Aug 27 12:01:37 1999 David L. Levine * ace/OS.cpp (mktime): added static cast of -1 return value to time_t to avoid warnings on VxWorks. Fri Aug 27 11:38:36 1999 Nanbor Wang * ace/Service_Config.cpp (process_directives): Remember to close down the svc.conf file pointers once they are processed. Thanks to Lothar Werzinger for pointing this out. Fri Aug 27 11:29:21 1999 David L. Levine * ace/OS.*: moved ACE_OS_Recursive_Thread_Mutex_Guard from OS.cpp to OS.h so that Log_Msg.cpp can use it. * ace/Log_Msg.cpp: replaced use of ACE_Guards with ACE_OS_Thread_Recursive_Mutex. [Bug 262] * ace/OS.h (ACE_OS_Recursive_Thread_Mutex_Guard::lock): added this method. * ace/OS.cpp (ACE_OS_GUARD,ACE_TSS_CLEANUP_GUARD, ACE_TSS_BASE_GUARD): fixed to use ACE_OS_Recursive_Thread_Mutex_Guard::lock. Added _RETURN versions. * ace/Synch_T.i: reordered some methods to avoid use before definition. Fri Aug 27 09:53:09 1999 Joe Hoffert * ace/ATM_Stream.{h,cpp}: added get_vpi_vci method. * examples/IPC_SAP/ATM_SAP/CPP-{client,server}.cpp: Changed tests to print out VPI and VCI after connected. Fri Aug 27 10:50:10 1999 David L. Levine * ACE version 5.0.3 released. Fri Aug 27 09:32:26 1999 David L. Levine * bin/make_release: added some diagnostics to see why diffs aren't being created. Fri Aug 27 07:17:58 1999 Toshio Hori * ACE-INSTALL.html: fixed a typo, missing ; after < and >. Thu Aug 26 21:01:02 1999 Jim Rucker * ace/Thread_Manager.cpp (spawn_i): on VxWorks, if t_handle is non-0, set it with thr_handle. Thu Aug 26 16:33:42 1999 David L. Levine * ace/Log_Msg.cpp: removed unnecessary #include "ace/Synch_T.h". [Bug 262] Thu Aug 26 13:49:00 1999 David L. Levine * examples/Threads/Makefile: removed process_manager from BIN. See Sat Aug 21 12:36:40 1999 Douglas C. Schmidt. Thu Aug 26 12:22:55 1999 David L. Levine * ace/config-ghs-common.h: on Sun, added ACE_HAS_NONSTATIC_OBJECT_MANAGER. It's necessary to prevent seg fault on startup. Thanks to David Sunwall for reporting this problem. Thu Aug 26 12:11:09 1999 David L. Levine * docs/ACE-guidelines.html: added guidelines to avoid default arguments, and favor narrow interfaces. Thu Aug 26 11:44:26 1999 David L. Levine * tests/Time_Value_Test.cpp (main): added static casts of the doubles to longs. g++ 2.7.2.3 warned about passing a double to a long :-) Thu Aug 26 11:20:55 1999 David L. Levine * tests/Time_Value_Test.cpp (main): added a test showing that ACE_Time_Value (2.5) == ACE_Time_Value (2). The 2.5 gets coerced to a long with value 2. Beware . . . Thu Aug 26 10:43:04 1999 Carlos O'Ryan * ace/config-sunos5.6.h: egcs-2.95 under Solaris 5.6 does not need ACE_HAS_STL_QUEUE_CONFLICT, thanks to Krishnakumar B. for reporting this problem. This fixes [BUGID:264] Thu Aug 26 07:29:42 1999 David L. Levine * ace/OS.{h,i} (ACE_Time_Value): reverted Wed Aug 25 22:05:54 1999 David L. Levine change. It caused overload ambiguity when the types of sec and usec didn't match, and on Sun CC 4.2/SunOS 5.6. Thanks to Andy for reporting that. Wed Aug 25 22:05:54 1999 David L. Levine * ace/OS.{h,i} (ACE_Time_Value): added four new constructors. One takes a double, to avoid surprises (truncation) when constructing an ACE_Time_Value from a double. The other three constructors take two unsigned longs, ints, and unsigned ints. They avoid the ambiguity introduced by the new double constructor. Thanks to Dave Meyer for starting this :-) The double constructor is wrapped with #ifndef ACE_LACKS_FLOATING_POINT. * tests/Time_Value_Test.cpp (main): removed cast of 2.0 to long. Wed Aug 25 16:15:07 1999 David L. Levine * docs/ACE-subsets.html: removed spurious LOGGING_FILES text. Wed Aug 25 15:55:05 1999 David L. Levine * ace/OS.{h,cpp},Object_Manager.*,Signal.i (default_mask): moved the default_mask () static method from ACE_Object_Manager to ACE_OS_Object_Manager, so that it can be used by Log_Msg.cpp without layer inversion. (Left a deprecated, inline ACE_Object_Manager::default_mask (), for backward compatibility.) default_mask () now returns a sigset_t * instead of an ACE_Sig_Set *, again to avoid layer inversion. * ace/Log_Msg.cpp (log): replaced use of ACE_Sig_Guard with a bare-bones ACE_Log_Msg_Sig_Guard. [Bug 262] Wed Aug 25 14:53:58 1999 Carlos O'Ryan * bin/run_all.pl: * bin/auto_compile: The script did not start with the canonical trick to run perl from [t]csh, [ba]sh or perl itself. Wed Aug 25 14:45:06 1999 Ossama Othman * tests/Process_Manager_Test.cpp: Explicitly include "ace/Signal.h" to make declaration of ACE_Sig_Set class visible. return statement at end of multiple_children_sig_handler() function is never reached. It was commented out to prevent compiler warnings. Wed Aug 25 13:34:07 1999 Narendra Ravi * ACE-INSTALL.html: added information about -j GNU make option. Wed Aug 25 12:26:49 1999 Carlos O'Ryan * ace/OS.h: On Borland C++ (and other platforms) we should use std::bad_alloc instead of bad_alloc for the ACE_bad_alloc exception. We use the ACE_USES_STD_NAMESPACE_FOR_STDCPP_LIB macro to detect this. Thanks to Jody Hagins for this fix. * tests/run_tests.lst: Disabled the Token test, it is more trouble that it is worth. It hangs and tends to freeze the complete testsuite. Wed Aug 25 12:09:32 1999 Carlos O'Ryan * Makefile: Include (conditionally) the $TAO_ROOT/tao/tao_targets.GNU file, that defines the targets required to compile just TAO. Tue Aug 24 21:47:44 1999 David L. Levine * ace/Synch_Options.{h,cpp} (timeout): added const to ACE_Time_Value & argument, because it's not modified. Thanks to John Lindal for suggesting this. Tue Aug 24 10:02:51 1999 Douglas C. Schmidt * tests/Process_Manager_Test.cpp: Fixed a number of warnings caused by unused args and signed/unsigned mismatches. Thanks to David Levine for reporting this. * ace/Service_Repository.h, ace/Service_Config.h, ace/OS.h: Changed the default number of entries in the ACE_Service_Repository from ACE_DEFAULT_SELECT_REACTOR_SIZE to ACE_DEFAULT_SERVICE_REPOSITORY_SIZE, which defaults to 1024. Thanks to Ivan Murthy for suggesting this. Tue Aug 24 16:23:07 1999 Carlos O'Ryan * bin/make_release: Use an explicit temporary file to create the ChangeLog, i had some problems with the previous scheme, but it could be my shell. When tagging TAO we have to change directories. Tue Aug 24 12:19:29 1999 Carlos O'Ryan * tests/Makefile: Updated dependencies * performance-tests/TCP/tcp_test.cpp: Fixed argument parsing. Tue Aug 24 10:37:02 1999 David L. Levine * bin/cvslog (find_changelogs): one more time: append /$file to `pwd`, so that we can find ChangeLogs in child directories. Always call basename (), again. Tue Aug 24 09:10:58 1999 David L. Levine * bin/cvslog (find_changelogs): don't call basename () immediately after calling `pwd`, so that ChangeLogs in current directory will be found. Tue Aug 24 07:11:52 1999 David L. Levine * ace/OS.*,Object_Manager.{h,cpp},README: with ACE_LACKS_ATEXIT, use ACE_OS_Object_Manager::at_exit () to emulate atexit (). [Bug 246] * tests/Object_Manager_Test.cpp,tests/Makefile, tests/run_pharlap_tests.bat,tests/run_tests.bat, tests/run_tests.lst,tests/run_tests.psosim, tests/run_tests.vxworks,tests/t.icc,tests/icc.bat: added Object_Manager_Test. Mon Aug 23 22:03:24 1999 Carlos O'Ryan * include/makeinclude/rules.lib.GNU: * include/makeinclude/rules.local.GNU: When compiling both shared and static libraries only one of them was installed. Thanks to Paul Calabrese for providing this fix. Mon Aug 23 21:41:43 1999 Carlos O'Ryan * tests/run_tests.pl: We search for "Starting" instead of "starting" because that's what the tests print out now. Mon Aug 23 21:20:28 1999 Carlos O'Ryan * docs/ACE-guidelines.html: Added new guidelines relevant to ACE+TAO. Mon Aug 23 16:57:13 1999 Carlos O'Ryan * performance-tests/RPC/README: * performance-tests/RPC/client.cpp: Changed the options to make it more consistent with other tests like this one. Added a README file Mon Aug 23 16:53:00 1999 Carlos O'Ryan * performance-tests/Makefile: * performance-tests/TCP/README: Added a README file for the test * performance-tests/TCP/tcp_test.cpp: Added options to pace the events and to control the message size on the server side. * performance-tests/RPC/Makefile: * performance-tests/RPC/client.cpp: * performance-tests/RPC/ping.x: * performance-tests/RPC/server.c: Added a simple RPC test to compare sockets (ACE), RPC and TAO. Mon Aug 23 13:30:32 1999 Nanbor Wang * include/makeinclude/platform_sunos5_kcc.GNU: The optimization flag for KCC should be +K3, not -K2. Mon Aug 23 10:56:40 1999 David L. Levine * bin/cvslog: added this wrapper around cvs log. It expands ChangeLogTags to their ChangeLog entries. Thanks to Luther for helping greatly with it. Mon Aug 23 08:46:37 1999 David L. Levine * ace/Object_Manager.{h,cpp}: replaced the use of ACE_Unbounded_Queue with a new ACE_Cleanup_Info_Node private helper class. The reduces the size of libACE by over 4 Kb, and is the first step towards moving the Object_Manager's atexit () support to OS. [Bug 246] Mon Aug 23 07:28:02 1999 David L. Levine * tests/run_tests.check: changed "starting" to "Starting", to match a recent change to test_config.h. * tests/test_config.h: changed one remaining "starting" to "Starting", to be consistent. Mon Aug 23 07:08:17 1999 David L. Levine * tests/run_tests.lst: disabled Process_Manager_Test, because it never terminates when built with egcs on both Solaris 7 and Linux. Sun Aug 22 21:15:13 1999 David L. Levine * ace/Makefile: with Green Hills for x86 target, use $(MAKE) instead of make to build .obj/gethrtime.o. Thanks to Cliff_H_Campbell for reporting this. Sun Aug 22 21:02:26 1999 David L. Levine * ace/config-vxworks5.x.h: added ACE_LACKS_ATEXIT. Sun Aug 22 21:01:18 1999 David L. Levine * ace/README: added ACE_LACKS_ATEXIT. Sun Aug 22 14:51:56 1999 Vishal Kachroo * ace/OS.i (join_leaf): Constructed a winsock2 QOS structure within ACE_OS::join_leaf () from the corresponding ACE QOS classes. Passed this QOS structure to WSAJoinLeaf (). Sun Aug 22 12:18:42 1999 Nanbor Wang * ace/Process.i (wait): Fixed a typo. * ace/Process.cpp (ACE_Process): Removed an extra underscore. * ace/OS.i (getppid): Removed undefined variable . Sat Aug 21 12:36:40 1999 Douglas C. Schmidt * ace/OS.h: Moved the ACE_OS::set_exit_hook() static method into the OS.i file so be more consistent with the other ACE_OS methods. Also, changed the signature of extern "C" void ACE_OS_Object_Manager_Internal_Exit_Hook () to void ACE_OS_Object_Manager_Internal_Exit_Hook (void) so that this will match the type declaration for ACE_EXIT_HOOK. * ace/Process_Manager.cpp: Use the new ACE_OS::wait() method, which is portable to Win32, rather than ACE_OS::waitpid(). * tests/Process_Manager_Test.cpp: Completed the first major enhancement to the test that exercises the new ACE_Process_Manager features. This test builds and runs cleanly on Solaris using CC and should run on other versions of UNIX that support threads and processes. It would be great if someone could improve it to run on Win32, as well. * ace/Process_Manager: Added support for the spawn_n() method, which creates N new processes. * tests: Cleaned up the use of ACE_INIT_LOG/ACE_APPEND_LOG in Process_Manager_Test.cpp and Malloc_Test.cpp so we don't lose information. Both these tests now write their child processes' output to "-child.log". * ace/Thread_Manager.h: Added a default value of 0 to the exit() method since this is a common return status. * ace/config-win32-common.h: Added ACE_LACKS_GETPPID. * ace: Added a new wrapper for getppid(). For platforms that don't support this system call, please add ACE_LACKS_GETPPID. * tests: Moved the $ACE_ROOT/examples/Threads/process_manager.cpp example into the $ACE_ROOT/tests directory, renamed it to Process_Manager_Test.cpp, and revised the test to illustrate the new ACE_Process_Manager features. * tests/run_tests.lst: Added the Process_Manager_Test to the list of tests to run. * tests/Malloc_Test.cpp (main): Added some necessary ASYS_TEXT macros. * ace/Process.i: Rather than replicate the #ifdef logic in ACE_OS::kill() within ACE_Process:kill(), we just forward to ACE_OS::kill() and let it sort this out! * ace/Process: Moved several of the short wait() methods from the *.cpp file to the *.i file and inlined them. * ace/Process_Manager.cpp: The logic for the implementation of terminate() was incorrect, i.e., the process table entry should be removed *if* ACE::terminate_process() succeeds but *not* if it fails! * ace/OS.h: Added a new method called wait() that works portably on Win32 and UNIX/POSIX platforms. * ace/Process: Added an optional "options" parameter to the former "blocking" wait() method on ACE_Process. This allows clients to use the WNOHANG option to avoid blocking if the child process hasn't exited yet. Also changed the type of the return value to pid_t rather than int. Finally, added an optional status pointer to the "timed" wait() method (which only works on Win32). * ace/OS.h: On platforms that do not define WNOHANG natively, changed #ifdef of WNOHANG to 0100 (which can be used in bitmask comparisons) rather than Douglas Adams favorite constant: 42. Sat Aug 21 14:13:46 1999 Carlos O'Ryan * ACE version 5.0.2 released. Sat Aug 21 11:52:51 1999 Carlos O'Ryan * Makefile: Fixed small problem in the releaseall target. * bin/make_release: The shell did not like the embedded script to print the ChangeLog entry. Fri Aug 20 15:40:58 1999 Douglas C. Schmidt * ace/Process_Manager: Added a new method called wait() that's just syntactic sugar for ACE_Process_Manager::reap(). Thanks to Dave Madden for suggesting this. * ace/OS: Added support for ACE_OS::atexit(), which necessitated changing the extern "C" signature for ACE_EXIT_HOOK to typedef void (*ACE_EXIT_HOOK) (void) so that this will be consistent with the signature expected by the ISO/ANSI C atexit() call. At the moment, ACE_OS::atexit just calls ::atexit() (unless a platform's config.h file #defines ACE_LACKS_ATEXIT). Ultimately, we might want to have this hook into the ACE_OS_Object_Manager's at_exit() support, which may be more portable and reliable. * ace/ACE_Process_Manager.h: Made the destructor virtual so it's possible to subclass from this class. * ace/Process_Manager.h: Updated the comments on the wait() method to indicate that a signal handler must be used to keep track of the number of remaining child processes. Thanks to Dave Binder for motivating this. Thu Aug 19 18:45:00 1999 Chris Gill * ace/Auto_Ptr.h: modified ACE_AUTO_PTR_RESET macro, to pass the type of the pointer. This is needed for the case where ACE_AUTO_PTR_LACKS_RESET is defined, in order to explicitly convert the pointer into an auto_ptr for assignment. Thu Aug 19 17:26:50 1999 John Heitmann * docs/ACE-SSL.html: Cleaned up the auto generated errors. Thu Aug 19 16:47:21 1999 Douglas C. Schmidt * ACE-INSTALL.html: Added a reference to $ACE_ROOT/docs/ACE-SLL.html. * docs: Added a new ACE-SSL.html file that explains how to participate in the ACE SSL project. Thanks to John Heitmann for adding this. Thu Aug 19 15:31:58 1999 Nanbor Wang * ace/OS.h (ACE_OSTREAM_TYPE): Removed the redundant trailing * for the defintion when ACE_LACKS_IOSTREAM_TOTALLY is defined. Thu Aug 19 14:16:37 1999 David L. Levine * include/makeinclude/rules.nested.GNU: added ACE_NESTED_COMMAND hook. Users can set it to whatever they like (preferably in include/makeinclude/platform_macros.GNU). One example is to set it to: ACE_NESTED_COMMAND := || break $? to terminate building if a build in one subdirectory fails. Thanks to Lothar Werzinger for suggesting this addition. Thu Aug 19 13:06:00 1999 Michael Kircher * ace/Read_Buffer.cpp: Fixed a bug in rec_read () method which prohibited reading more than 1024 bytes at once. Thu Aug 19 13:03:07 1999 Douglas C. Schmidt * ace/ACE.h (ACE): Fixed a "cut & paste" error in the comment -- basename should have been dirname. Thanks to Mark Hyett for reporting this. * ace/ACE.cpp: Fixed a bug in ACE::dirname() that was causing problems if the pathname didn't have a delim character. Thanks to Russ Noseworthy for reporting this and providing a fix. Thu Aug 19 12:59:56 1999 Carlos O'Ryan * ace/Message_Block.i: Moved ACE_Message_Block::replace_data_block() down in the file because it was using inline functions that were defined after it was. Thanks to David for pointing this out. Thu Aug 19 10:53:50 1999 David L. Levine * ace/OS.i (setuid): added ACE_UNUSED_ARG (uid) for VxWorks, pSOS, WIN32, and Chorus. Thanks to Garry Brother for reporting this. Thu Aug 19 10:04:12 1999 David L. Levine * include/makeinclude/platform_vxworks5.x_g++.GNU: added trap to cause builds to fail if the exceptions flag is enabled. The VxWorks libraries don't support exceptions. And, exceptions give the (Tornado II) compiler fits, e.g., with ACE_wrappers/TAO/tao/IORManipulation.cpp. Thanks to Erik Johannes for this nifty preemptive solution, and to Nick Sawadsky for motivating its implementation. Thu Aug 19 09:34:41 1999 Douglas C. Schmidt * ace/WFMO_Reactor.h, * ace/WFMO_Reactor.cpp, * ace/Select_Reactor_T.h, * ace/Select_Reactor_T.cpp, * ace/Reactor_Impl.h, * ace/Reactor.h, * ace/Reactor.cpp: Added a pair of restart() methods that can be used to set/get the restart_ flag, which controls whether the handle_events() event loop is restarted automatically when select() is interrpted. Thanks to Erik Koerber for suggesting this. * ace/Service_Config: Updated the failure result comment for open() and cleaned up the return value and errno propagation for open_i() when process_commandline_directives() fails. Thanks to Jerry D. De Master for reporting the bug and suggesting a fix. * ace/Service_Config.h: Moved the process_directives() method used to parse the svc.conf files out of the protected part of the class to the public part so that application programmers can call this directly to reconfigure services from the svc.conf files. Thanks to John Mink for suggesting this. Wed Aug 18 22:45:24 1999 Douglas C. Schmidt * ace/WFMO_Reactor.h, * ace/WFMO_Reactor.cpp, * ace/XtReactor.h, * ace/XtReactor.cpp, * ace/TkReactor.h, * ace/TkReactor.cpp, * ace/FlReactor.h, * ace/FlReactor.cpp, * ace/Select_Reactor_T.h, * ace/Select_Reactor_T.cpp, * ace/Reactor_Impl.h, * ace/Reactor.h, * ace/Reactor.cpp: * ace/Timer_Hash_T.h: * ace/Timer_Heap_T.h: * ace/Timer_List_T.h: * ace/Timer_Queue_T.h: * ace/Timer_Wheel_T.h: * ace/Timer_Hash_T.cpp: * ace/Timer_Heap_T.cpp: * ace/Timer_List_T.cpp: * ace/Timer_Wheel_T.cpp: Changed all uses of const long to long... Wed Aug 18 22:15:58 1999 Carlos O'Ryan * tests/run_tests.sh: It was running the DISABLED tests, thanks to David for pointing this out. Wed Aug 18 21:48:52 1999 David L. Levine * ACE-INSTALL.html: added Toshio Hori suggestions (in a new section) on building just what's needed for TAO, and his tips for reducing disk space usage on Unix platforms. Wed Aug 18 21:32:56 1999 Ossama Othman * docs/usage-bugzilla.html: Added a simple document explaining basic usage of our installation of bugzilla. * ACE-INSTALL.html: * PROBLEM-REPORT-FORM: Updated with information about the bugzilla usage documentation. Wed Aug 18 20:28:23 1999 David L. Levine * bin/check_build_logs: check for just Stop, not [Ss]top, to avoid snagging top output. Gnu make says "Stop" when it can't find a dependency. Wed Aug 18 19:08:08 1999 David L. Levine * ace/OS.{h,cpp},Object_Manager.{h,cpp},Log_Msg.cpp: moved ACE_LOG_MSG_INSTANCE_LOCK from ACE_Object_Manager to ACE_OS_Object_Manager, so that Log_Msg.cpp needn't to depend on Object_Manager for it. Wed Aug 18 18:20:11 1999 Carlos O'Ryan * tests/SOCK_Send_Recv_Test.cpp: Fixed typo. Wed Aug 18 17:58:06 1999 Carlos O'Ryan * THANKS: * ace/README: * ace/config-irix6.x-common.h: Added new macro (ACE_HAS_BROKEN_DGRAM_SENDV) to handle platforms where using sendmsg with exactly IOV_MAX elements in the iovec fails (normally the number of elements can be up to and including IOV_MAX). Thanks to Tobin Bergen-Hill for tracking this down. Wed Aug 18 16:08:46 1999 Douglas C. Schmidt * tests/SOCK_Send_Recv_Test.cpp: Reformatted this test so that it conforms to ACE programming guidelines. Wed Aug 18 13:49:30 1999 David L. Levine * ace/Makefile: if ACE_COMPONENTS = FOR_TAO, don't include Filecache in UTILS_FILES. TAO doesn't need Filecache. And, its ACE_Hash and ACE_Equal_To explicit template instantiations collide with those in ACE_wrappers/TAO/tao/POA.cpp. That precludes building a shared libTAO for VxWorks, for example. And, it costs 14 Kb. Thanks to Erik Johannes for helping to find this solution. And reminding me to implement it :-) Wed Aug 18 13:16:09 1999 David L. Levine * ACE-INSTALL.html: clarify that PERL_PATH must include the full path, including the perl.exe filename. Thanks to Walter Welzel for reporting this. Wed Aug 18 13:11:30 1999 Carlos O'Ryan * ace/OS.h: * ace/OS.i: The new methods to manipulate condition variable attributes were not compiling on single threaded platforms. Wed Aug 18 13:09:36 1999 David L. Levine * tests/Makefile: don't add libDLL_Test.$(SOEXT) to BUILD on VxWorks. Thanks to Walter Welzel for reporting that it did get build for VxWorks with shared libs enabled, but not successfully (or usefully). Wed Aug 18 10:36:16 1999 Nanbor Wang * ACE-INSTALL.html: Updated all references to MSVC 4.x. We no longer support the compiler. Thanks to Welzel Walter for pointing this out. Wed Aug 18 10:42:13 1999 Douglas C. Schmidt * ace/Timer_List_T.cpp (reset_interval): Added a constcast to make SunC++ 4.2 happy. Thanks to Andy for reporting this. * ace/WFMO_Reactor.i (reset_timer_interval): The return type was incorrect... Thanks to Andy and Jeff for reporting this. Wed Aug 18 09:47:51 1999 Carlos O'Ryan * tests/run_tests.pl: The disabled tests were executed anyway, thanks to David for pointing this out. Wed Aug 18 09:34:46 1999 Carlos O'Ryan * ace/FlReactor.cpp: * ace/XtReactor.cpp: Fixed typos in the code, this is only used in a few platforms, that's why it didn't show up before. Tue Aug 17 20:02:55 1999 Douglas C. Schmidt * tests/Reactor_Timer_Test.cpp: Enhanced this test to exercise the new reset_timer_interval() feature that Darrell added. It appears to work quite well. Also, make the test easier to understand by adding more structure. * ace/WFMO_Reactor.h, * ace/WFMO_Reactor.cpp, * ace/XtReactor.h, * ace/XtReactor.cpp, * ace/TkReactor.h, * ace/TkReactor.cpp, * ace/FlReactor.h, * ace/FlReactor.cpp, * ace/Select_Reactor_T.h, * ace/Select_Reactor_T.cpp, * ace/Reactor_Impl.h, * ace/Reactor.h, * ace/Reactor.cpp: Exposed the new reset_interval() method Darrell added to all the Timer mechanisms via a method in each of these Reactor's called reset_timer_interval(). * ace/Timer_Hash_T.h: * ace/Timer_Heap_T.h: * ace/Timer_List_T.h: * ace/Timer_Queue_T.h: * ace/Timer_Wheel_T.h: * ace/Timer_Hash_T.cpp: * ace/Timer_Heap_T.cpp: * ace/Timer_List_T.cpp: * ace/Timer_Wheel_T.cpp: Changed the return values of to be (1) consistent (i.e., to return a value through all paths in the method) and (2) to return -1 on failure and 0 on success. Also replaced a number of C-style casts with ACE_reinterpret_casts. Tue Aug 17 20:51:35 1999 Carlos O'Ryan * ace/TP_Reactor.h: * ace/TP_Reactor.cpp: The TP Reactor also experiences a performance hit if signal masking is enabled. Therefore, we can now make this optional. Tue Aug 17 19:10:00 1999 Darrell Brunsch * ace/OS.cpp: Fixed typo. * ace/Stats.cpp: Removed some warnings showing up in MSVC when assigning ACE_UINT64 values to ACE_UINT32 variables. Used ACE_U64_TO_U32 to remove these warnings. * ace/Timer_Hash_T.h: * ace/Timer_Heap_T.h: * ace/Timer_List_T.h: * ace/Timer_Queue_T.h: * ace/Timer_Wheel_T.h: * ace/Timer_Hash_T.cpp: * ace/Timer_Heap_T.cpp: * ace/Timer_List_T.cpp: * ace/Timer_Wheel_T.cpp: Added a new method, which can be used to change the interval of a timer represented by a timer_id. Tue Aug 17 18:43:53 1999 Jeff Parsons * ace/OS.h: * ace/OS.i: Added wscmp and wsncmp functions to go with the existing wslen and wscpy functions. These all use WChar*, so arg size is independent of platform. These are to be used for all manipulations of CORBA::WString. The two that were added were needed for wstring tests in Param_Test. WChar* versions of the other string manipulation functions will be added as needed. Tue Aug 17 17:49:33 1999 Douglas C. Schmidt * ace/config-sunos5.5.h: If (_LARGEFILE_SOURCE) || (_FILE_OFFSET_BITS==64) are enabled then #undef ACE_HAS_PROC_FS and ACE_HAS_PRUSAGE_T to work around problems with Solaris 7. Thanks to Sridhara Rao Dasu for reporting this. * ace/OS.i: Updated the readdir_r() conditional compilation so that it'll work when _FILE_OFFSET_BITS == 64. Thanks to Sridhara Rao Dasu for reporting this. Tue Aug 17 17:17:13 1999 Carlos O'Ryan * ace/OS.h: * ace/OS.i: * ace/OS.cpp: Added new data type to handle condition variable attributes (ACE_condattr_t). New methods to manipulate that data type (ACE_OS::cond_attr_init and destroy) Add new method to create a condition variable from a pre-initialized attribute. * ace/Synch.h: * ace/Synch.i: * ace/Synch.cpp: Added new wrapper for the ACE_condattr_t structure. * ace/Token.h: * ace/Token.cpp: Use ACE_Condition_Attributes to minimize the number of memory allocations and locks in the critical path of the TP_Reactor. Tue Aug 17 15:55:17 1999 Nanbor Wang * performance-tests/TCP/TCP.{dsw,dsp}: * performance-tests/TCP/Makefile: * performance-tests/TCP/tcp_test.cpp: Added the new test modified by Carlos and a bit further by me. The test measures TCP performance of a platform. We can also use it to test the performance of various reactors. Tue Aug 17 14:52:40 1999 Darrell Brunsch * ace/config-win32.h: Added ACE_CC_COMPILER definition for MSVC. ACE_CC_COMPILER is a string containing the compiler program. Tue Aug 17 13:48:27 CDT 1999 Loren Rittle * bin/update-ace+tao.sh: Use authoritative name list instead of ad hoc list to update file. Tue Aug 17 06:33:05 1999 Douglas C. Schmidt * ace/Task.h: Clarified the comment regarding open() and close() on ACE_Task_Base. Thanks to Jeff Franks for suggesting this. Tue Aug 17 11:56:33 1999 Carlos O'Ryan * ace/Select_Reactor_T.h: * ace/Select_Reactor_T.cpp: Added option in the constructor to disable the signal handling features in the select reactor. This eliminates a couple of kernel level locks that are not needed in all applications. * ace/Stats.h: * ace/Stats.cpp: Now we keep track of the iteration where the minimum and maximum latencies were achieved. Tue Aug 17 06:33:05 1999 Douglas C. Schmidt * bin: Added a new shell script called update-ace+tao.sh that updates the ACE/TAO source tree to the latest numeric-assigned version. Thanks to Loren Rittle for contributing this. Mon Aug 16 20:09:12 1999 Carlos O'Ryan * ace/CDR_Stream.h: * ace/CDR_Stream.i: * ace/CDR_Stream.cpp: Added new method to transfer the contents of one CDR stream into another. This is used in the implementation of the ORB to minimize the number of data copies, memory allocations and locks. * ace/Message_Block.h: * ace/Message_Block.i: Added new method to replace the Data_Block, this method returns the old one, letting the user decide how to deallocated the Data block, possibly transfering it to another Message_Block. Mon Aug 16 15:58:12 1999 Steve Huston * ace/config-sunos5.5.h: * include/makeinclude/platform_sunos5_sunc++.GNU: Added support for using "compat=4" mode for Sun CC 5. To use this, do "make compat4=1", or equivalent. Compat 4 uses -compat=4 with its old iostreams, but the newer Sun CC 5 scheme of explicit template instantiation. Mon Aug 16 11:59:18 1999 Vishal Kachroo * ace/SOCK_Dgram_Mcast.cpp (subscribe): Constructed and passed a complete sockaddr_in to the ACE_OS::join_leaf () call. This was just being passed sockaddr_in.sin_addr.s_addr. Sat Aug 14 18:01:27 1999 Carlos O'Ryan * tests/run_tests.pl: Fixed some output messages that were going to stdout. Sat Aug 14 15:20:27 1999 Douglas C. Schmidt * ace/SOCK_Dgram_Mcast.cpp (subscribe): It looks like there was a bug where we were passing an s_addr to ACE_OS::join_leaf() rather than a sockaddr *. Fixing this also cleaned up some code that was #ifdef'd for UNICOS. Thanks to Vishal for reporting this. * ace/Process.h: Added a virtual to ~ACE_Process so that it can be used for subclassing more readily. Thanks to Zoran Ivanovic for suggesting this. Sat Aug 14 14:11:25 1999 Everett Anderson * ace/Remote_Tokens.cpp (tryacquire): Fixed so that it uses ACE_Token_Request::TRY_ACQUIRE instead of RELEASE. Fri Aug 13 18:55:51 1999 Douglas C. Schmidt * ace/Handle_Set.h (ACE_Handle_Set): Clarified the meaning of the nbits_ array. Thanks to Hao Ruan for motivating this. Fri Aug 13 21:18:40 1999 David L. Levine * ace/OS.i (llseek): on Linux, moved extern "C" llseek () declaration out of the function. Thanks to Tal Lev-Ami for this fix. Fri Aug 13 20:50:02 1999 Nanbor Wang * ace/Select_Reactor_T.cpp (handle_events): Added a deactivation check for non-threaded case. The original check was mixed with thread owner check and thus got disabled under single-threaded configurations. Thanks to Kirthika for reporting the problem. Fri Aug 13 13:11:16 1999 David L. Levine * bin/make_release: moved ACE_wrappers/TAO addition to $release_files so that it's only done for the ACE+TAO kit. Fri Aug 13 12:44:08 1999 David L. Levine * Makefile,bin/make_release: added support (and documentation) for creating kits with or without new tags, or using an old tag; generated man pages; installation to the public ftp/http site on ace; and zip files. Fri Aug 13 07:58:12 1999 David L. Levine * ace/OS.{h,i},README,config-osf1-4.0.h (llseek): added ACE_HAS_LLSEEK support. Thanks to Tal Lev-Ami for providing this. Thu Aug 12 22:05:37 1999 Douglas C. Schmidt * ace/OS.i (sema_wait, rw_tryrdlock, rw_trywrlock): Replaced more ad hoc uses of the int error = 0; ... error = errno; ... errno = error; idiom with the ACE_Errno_Guard. * ace/Synch.h: Updated the comment to remind folks who would like to have a working ACE_Process_Condition to contact us if they know how to do this. Thanks to Craig Anderson for motivating this. * tests/Process_Strategy_Test.cpp (open): Clarified the behavior of the code that's run when fork() is used. Thanks to Umar Syyid for motivating this. Thu Aug 12 21:30:09 1999 Alexander Babu Arulanthu * include/makeinclude/platform_vxworks5.x_g++.GNU: added SIMNT support. Thu Aug 12 20:54:40 1999 David L. Levine * ace/config-osf1-4.0.h: added ACE_HAS_USING_KEYWORD for cxx and Rational compilers. [Bug 163] Thu Aug 12 20:47:13 1999 Carlos O'Ryan * tests/Cached_Accept_Conn_Test.cpp: Made the and variables global instead of static, because otherwise Sun/CC automatic template instantiation cannot find them at link time. The variables are used in template code that is defined and used in the same file, but Sun/CC does not seem to understand that. Thu Aug 12 17:14:35 1999 Jeff Parsons * ace/OS.h: Changed typedef of WChar to ACE_UINT16. The CDR operations are already using write_2, read_2 and SHORT_SIZE, and in the process of implementing wchars and wstrings in IDL, it's going to be important to optimize for the Unicode common case. Thu Aug 12 16:24:16 1999 David L. Levine * bin/make_release: chdir up a level before removing the staging directory. Thanks to Carlos for pointing out that it's necessary, because we had chdir'd into it. And for noticing that $build_commands got clobbered in create_kit (). Obey $generate_man_pages setting. Thu Aug 12 16:03:03 1999 David L. Levine * ACE version 5.0.1 released. Thu Aug 12 15:49:00 1999 David L. Levine * Makefile,bin/make_release: build kits in /tmp staging area, and copy to ace's ftp pub/ACE directory. Thu Aug 12 11:36:46 1999 Carlos O'Ryan * ace/OS.h: * ace/Svc_Handler.cpp: * ace/Future.cpp: Some platforms require arguments in the exception thrown from operator new. The standard (std::bad_alloc) does not require any. Added a new macro (ACE_throw_bad_alloc) that takes care of these differences, including platforms that don't throw exceptions but return something. Wed Aug 11 17:00:00 1999 Kirthika Parameswaran * tests/Cached_Accept_Conn_Test.cpp: Removed run_event_loop and end_event_loop and put in handle_events with a flag to stop the loop. This is a bug fix for Bug183. Wed Aug 11 15:49:37 1999 David L. Levine * docs/ACE-guidelines.html: cleaned up ACE doc tool discussion. Added Carlos' inlining heuristics. Wed Aug 11 13:53:20 1999 David L. Levine * ace/config-win32.h,config-win32-borland.h: added ACE_HAS_MUTABLE_KEYWORD. Thanks to Jody for this request. Wed Aug 11 13:15:26 1999 David L. Levine * ace/config-g++-common.h,config-win32.h,config-win32-borland.h: added ACE_HAS_EXPLICIT_KEYWORD (for egcs only, in config-g++-common.h). Thanks to Jody Hagins for this update. * ace/config-g++-common.h: added ACE_HAS_MUTABLE_KEYWORD. Wed Aug 11 13:10:32 1999 David L. Levine * ace/OS.h,README: added ACE_HAS_MUTABLE_KEYWORD support. Thanks to Jody Hagins for providing it. Wed Aug 11 12:55:23 1999 Carlos O'Ryan * ace/Svc_Handler.cpp: * ace/Future.cpp: The throw expressions require a constructor, not just the type, thanks to David for pointing this out. Wed Aug 11 10:12:11 1999 David L. Levine * ace/OS.h,Auto_Ptr.h,SString.h,README: added support for ACE_HAS_EXPLICIT_KEYWORD. Thanks to Jody Hagins for suggesting this. Wed Aug 11 10:07:36 1999 Carlos O'Ryan * ace/Future.cpp: * ace/Local_Tokens.cpp: * ace/Log_Msg.cpp: * ace/Log_Record.cpp: * ace/Svc_Handler.cpp: * netsvcs/lib/Token_Handler.cpp: Fixed several warnings and errors with gcc-2.95 * include/makeinclude/wrapper_macros.GNU: Add -fcheck-new to the command-line of g++ when compiling with exceptions disabled. In that case new operators could return 0 and without this option the constructor will surely be invoked on invalid data. * bin/auto_compile: Use LOGNAME instead of USER to get the username because it is more portable. Tue Aug 10 23:47:32 1999 Kirthika Parameswaran * ace/Cached_Connect_Strategy_T.h: * ace/Caching_Utility_T.h: * ace/Caching_Strategies_T.h: * ace/Cleanup_Strategies_T.h: * ace/Cache_Map_Manager_T.h: * ace/Hash_Cache_Map_Manager_T.h: Defined macros to shorten lengthy class names for linkers which couldnt grok long names (example: Sun Linker). Thanks to Dr.Schmidt for this cool idea. Tue Aug 10 19:07:32 1999 Carlos O'Ryan * tests/run_tests.sh: * tests/run_tests.pl: * tests/run_tests.lst: Move the list of tests to a separate file. This file is used to drive both a perl version of the script and the old shell version. The new perl version is able to terminate programs that hang, but still has problem with processes that fork children who then hang for some reason. The perl script has to be ported to Win32 too. * bin/run_all_list.pm: Use the new tests/run_tests.pl script on Unix. * bin/run_all.pl: Run all the test scripts in ACE and TAO, much like run_all_win32 but for unix systems. * bin/auto_compile: Remember to remove the disable file in case of errors. Tue Aug 10 18:08:30 1999 Jeff Parsons * ace/CDR_Stream.h: * ace/CDR_STream.i: Added all the necessary stuff for wstrings. Tue Aug 10 15:36:02 1999 David L. Levine * tests/Malloc_Test.cpp (main): removed unused argc declaration, with ACE_LACKS_FORK && ! ACE_HAS_WIN32. Tue Aug 10 15:26:43 1999 David L. Levine * ace/OS.cpp (strptime): removed unreachable break after return statement, in default case of switch statement. Tue Aug 10 14:46:12 1999 David L. Levine * ace/OS.{h,cpp},README: Changed support for ACE_OS::strptime () to be configurable. VxWorks, LynxOS, and Chorus, for example, don't have a native ::strptime (). The current emulation cost 2 Kb, which will be objectionable to some embedded systems developers. Replaced ACE_LACKS_STRPTIME with a combination ACE_HAS_STRPTIME/ACE_LACKS_NATIVE_STRPTIME. ACE_HAS_STRPTIME enables ACE_OS::strptime () support. ACE_LACKS_NATIVE_STRPTIME lets ACE know that there's no native ::strptime (), so it then uses its emulation. This has two advantages: it minimizes footprint in the default case, and doesn't break builds on platforms that don't have ::strptime (). It easily allows developers to configure ACE identically, with respect to strptime, on their Unix, NT, and embedded platforms. [Bug 219] * ace/config-sunos5.5.h,config-linux-common.h,config-osf1-4.0.h, config-irix6.x-common.h: added #define ACE_HAS_STRPTIME. [Bug 219] Tue Aug 10 12:28:54 1999 Jeff Parsons * bin/run_all_list.pm: Added IDL test to the list. Mon Aug 09 16:49:45 1999 Nanbor Wang * ace/Process.{h,cpp} (wait): Re-visited the implementation so it behaves the same on both UNIX and Win32. Thanks to Doug for suggesting this. Mon Aug 09 16:30:24 1999 David L. Levine * docs/ACE-guidelines.html: added note to follow the ACE documentation tools guidelines and restrictions. Thanks to Doug for pointing this out. Mon Aug 09 16:06:01 1999 Nanbor Wang * include/makeinclude/wrapper_macros.GNU: Applied smart eg++ recognizer. Thanks to Craig Anderson for providing the patch. Mon Aug 9 11:29:36 1999 Douglas C. Schmidt * bin: Fixed html-windex so that it works when class names contain the digit '3' in them, e.g., ACE_WIN32_Asynch_Accept, etc. Thanks to Jody Hagins for reporting this problem. Mon Aug 09 15:10:27 1999 Nanbor Wang * ace/Process.h (wait): Rewrote comments. * ace/Service_Config.h (open): Revised the comments for several service_config flags. Thanks to Doug for pointing this out. Mon Aug 9 11:29:36 1999 Douglas C. Schmidt * examples/Reactor/Misc/test_demuxing.cpp: Added a destructor to ~STDIN_Handler that will remove the event handler for STDIN and the registered timer. Mon Aug 9 10:45:24 1999 Douglas C. Schmidt * ace/Memory_Pool.cpp (handle_signal): When working with an MMAP memory pool, a "real" segmentation fault (not related to updating the memory mapping) will cause the process to fail but not write a core file. This is because that on linux (and possibly other flavours of UNIX) core file writing is a default behaviour which is disabled once you register a signal handler. Even if the signal handler returns a bad return value which crashes the program, a core file will not be written. The fix for this is to modify the mmap() signal handler so that it de-registers the signal handler when detecting a segfault with the mapping up to date and return an "ok" answer. the program will attempt to access the illegal address again and dump core. Thanks to Joseph Weihs for contributing this fix. * ace/Signal: Changed the new handle_i() and register_handler_i() methods to be static member functions so they can be called from the static dispatch() member function. Thanks to Andreas Geisler for reporting this. Mon Aug 9 10:33:50 1999 Douglas C. Schmidt * examples/ASX/CCM_App/SC_Server.cpp: Make sure to enable the appropriate parameter for ACE_Service_Config::open() so that static services will be enabled. Thanks to Michael Roth for reporting this. Mon Aug 09 10:21:37 1999 David L. Levine * include/makeinclude/wrapper_macros.GNU (INCLDIRS): removed "-I.". It's not necessary, because the preprocessor will first look for headers in the current directory, when appropriate. It causes trouble, for example, on NT for VxWorks with Green Hills. Instead of including the system's Signal.h, ACE's signal.h gets #included with "-I.". Thanks to Dale Wood for helping to track this down. [Bug 35] * include/makeinclude/platform_vxworks5.x_ghs.GNU: no longer need to override INCLDIRS, now that it doesn't contain "-I.". [Bug 35] Mon Aug 09 10:09:57 1999 David L. Levine * ace/OS.h,streams.h: moved ACE_OSTREAM_TYPE #define from streams.h to OS.h, so that it is seen with ACE_LACKS_IOSTREAM_TOTALLY. And, provide default value for ACE_DEFAULT_LOG_STREAM. [Bug 208] * ace/Log_Msg.cpp: changed a couple remaining ostream occurrences to ACE_OSTREAM_TYPE. Replaced conditional use of ACE_DEFAULT_LOG_STREAM/&cerr with ACE_DEFAULT_LOG_STREAM, now that it's always defined. [Bug 208] Mon Aug 09 09:47:06 1999 David L. Levine * ace/Log_Msg.{h,cpp}: replaced the ACE_Log_Msg_key_created_ and ACE_Log_Msg_key_ file static variables with static class members ACE_Log_Msg::key_created_ and ACE_Log_Msg::log_msg_tss_key_, respectively. Only declare ACE_Log_Msg::log_msg_tss_key_ if the platform supports TSS (native or emulated). Replaced the ACE_Log_Msg_message_queue file static variable with ACE_Log_Msg_Manager::message_queue_. Cleaned up Log_Msg.cpp. [Bug 86] Thanks to Umar Syyid for suggesting that we remove the file static variables. Mon Aug 09 09:33:43 1999 David L. Levine * ace/OS.h: #define ECOMM (to ECONNABORTED) if it's not already #defined. This allows support on NT. Thanks to Andy for reporting this. Mon Aug 9 07:45:19 1999 Douglas C. Schmidt * ace/Signal.h: Modified the behavior of the ACE_Sig_Handler::dispatch() method so that on Win32 it'll automagically re-register the signal handler after the handler_signal() method returns with a non -1 value. Thanks to Andreas Geisler for suggesting this. * tests/Malloc_Test.cpp: Win32 didn't seem to like a 32k offset for the virtual address. Let's try 64k instead. Thanks to Nanbor for reporting this. * tests/Malloc_Test.cpp: Changed the sense of the conditional compilation test from && !defined (ACE_WIN32) to || defined (ACE_WIN32) so that this will run on Win32. Thanks to Nanbor for reporting this. * ace/OS.i: Cleaned up the coding style for the sigset* wrapper functions. Sun Aug 08 21:19:42 1999 David L. Levine * ace/config-osf1-4.0.h: removed unreferenced_function_param warning suppression. It's no longer needed with the fixes to avoid unreferenced function parameters. [Bug 217] Fri Aug 06 15:20:00 1999 Kirthika Parameswaran * ACE-INSTALL.html: added information about compilation on SunOS5.6 using egcs1.1.2. Fri Aug 06 13:47:51 1999 David L. Levine * ACE-INSTALL.html: added note about disabling optimization for some files with Tornado II. Fri Aug 06 13:39:48 1999 David L. Levine * ACE-INSTALL.html: updated VxWorks build information, specifically, the instructions for building on NT host. And, added information about linking libACE.so and libTAO.so into the VxWorks kernel. Thanks to Erik Johannes and Stephane Chatre for helping to clarify and verify. Fri Aug 06 12:10:19 1999 Nanbor Wang * ace/config-win32-common.h: Added ACE_HAS_POSITION_INDEPENDENT_MALLOC. * ace/OS.cpp (strptime_getnum): getnum should be strptim_getnum. * ace/config-win32-common.h: Added ACE_LACKS_STRPTIME. Thanks to Andy for pointing this out. Fri Aug 06 11:13:27 1999 David L. Levine * ace/streams.h,OS.h,Log_Msg.h: with ACE_LACKS_IOSTREAM_TOTALLY, removed the hack that re-#defined ostream. Replaced that with an ACE_OSTREAM_TYPE macro. [Bug 208] Fri Aug 06 10:49:34 1999 David L. Levine * ACE-INSTALL.html: updated egcs versions that we use (1.1.1 and 1.1.2). Fri Aug 6 08:43:22 1999 Douglas C. Schmidt * examples/Log_Msg/test_log_msg.cpp (main): Changed this test so that it no longer abort()s, but instead calls ACE_OS::_exit(). This should help avoid confusion when it's run. Thanks to Kevin Lyda for reporting this. This fixes [BUGID:211]. * ace/SOCK_Dgram_Mcast: The QoS-enabled subscribe() call must call the QoS-enabled open() call, rather than the "regular" open() call because otherwise QoS information isn't initialized! Thanks to Vishal for pointing this out. Fri Aug 06 08:22:00 1999 David L. Levine * ace/Object_Manager.cpp: with ACE_SHOULD_MALLOC_STATIC_OBJECT_LOCK, made the ACE_Static_Object_Lock_lock instance static, to match the non-ACE_SHOULD_MALLOC_STATIC_OBJECT_LOCK behavior. Thanks to Jody for this change. Also, consolidated the code to alway use the ACE_Static_Object_Lock_Type typedef. Fri Aug 06 06:55:49 1999 Jody Hagins * include/makeinclude/rules.local.GNU (depend.local): added ACE_DEPEND_SED_CMD hook, so that users can add sed scripts. An example is to use relative instead of absolute paths in dependencies. I put this into it to get the desired (site-specific) functionality: USER_DEPEND_SED_CMD=-e "s;$(PRISM_ROOT);\$$(PRISM_ROOT);g" Thu Aug 5 22:26:18 1999 Douglas C. Schmidt * ace/OS.h (ACE_OS): Added support for strptime() for platforms that lack it. Thanks to Kevin Lyda for contributing this. Note that if a platform fails to compile because it doesn't support this function, please enable ACE_LACKS_STRPTIME in the config.h file and send us email. Thu Aug 05 21:37:05 1999 David L. Levine * ace/Object_Manager.cpp (ACE_Static_Object_Lock_lock, ACE_Static_Object_Lock::instance, cleanup_lock): added support for ACE_SHOULD_MALLOC_STATIC_OBJECT_LOCK. It's not (currently) used by ACE. But, applications may find it useful for avoiding recursive calls if they have overridden operator new. Thanks to Jody Hagins for contributing it. Thu Aug 05 21:04:20 1999 David L. Levine * ace/streams.h: removed #includes of istream.h and ostream.h with ACE_HAS_STD_CPP_LIBRARY and ACE_USES_OLD_IOSTREAMS. Those headers aren't available with Sun CC. I think that they might be unnecessary with egcs for Tornado II. Thanks to Doug for reporting this. Ref. Thu Aug 05 14:31:48 1999 David L. Levine . Thu Aug 5 16:43:29 1999 Douglas C. Schmidt * tests/Malloc_Test.cpp: Rearranged the #ifdefs so that we don't do much of anything if the platform doesn't support processes. Thanks to David Levine for reporting this. Thu Aug 5 16:27:15 1999 Carlos O'Ryan * include/makeinclude/rules.lib.GNU: * include/makeinclude/wrapper_macros.GNU: Added new macro (ACE_SHLIBS) to control the final link line of shared libraries. Use this macro to list the -l options required by your shared library that are *not* included in the system LIBS macro, for example "-lTAO -lACE"; the system will automagically change that to be "-lTAOshr -lACEshr" on AIX. This also decouples the shared library link line from the binaries, which use LDLIBS and ACELIB, allowing the application developers to easily mix shared libraries and binaries in the same directory, even if the binaries depend on the generated shared library. Thu Aug 05 15:44:39 1999 David L. Levine * include/makeinclude/platform_vxworks5.x_g++.GNU: on WIN32 host, assume that $(CC) is on the user's path (or is a full pathname). That way, we don't need to determine ACE_CC dynamically, which causes trouble with the WIN32 shell. Thanks to Erik Johannes for reporting this. [NOTE: this had been committed on a branch, but never on the main line.] Thu Aug 05 15:35:21 1999 Nanbor Wang * apps/gperf/src/gperf.dsp: * apps/gperf/src/gperf_lib.dsp: * ace/ace_lib.dsp: * ace/ace_dll.dsp: Changed the default configuration to Win32 Debug so if anyone doesn't know what he is doing, he'll most likely to success. Thu Aug 05 15:30:32 1999 David L. Levine * ace/streams.h: with ACE_LACKS_IOSTREAMS_TOTALLY, always #define ostream FILE *. [Bug 208] Thu Aug 05 15:13:01 1999 David L. Levine * ace/config-lynxos.h: added note on why we don't #define ACE_HAS_POSIX_SEM for LynxOS. [Bug 164] Thu Aug 05 14:53:44 1999 David L. Levine * ace/config-vxworks5.x.h: added support for the Diab compiler, version 4.2a or later. Compilation succeeds but links still have unresolved symbols related to new and delete. [Bug 207] * ace/OS.h,README: added support for ACE_LACKS_NEW_H. [Bug 207] * ace/OS.h: for VxWorks, use the ANSI prototypes for inet_ntoa.h, etc., for all VxWorks compilers, not just Green Hills. This allows compilation to succeed with Diab 4.2a. [Bug 207] Thanks to Erik Johannes for helping with this quick port to Diab 4.2a for VxWorks. Thu Aug 05 14:31:48 1999 David L. Levine * ace/iosfwd.h: #include iosfwd with ACE_HAS_MINIMUM_IOSTREAMH_INCLUSION as well as without ACE_HAS_OLD_IOSTREAMS. This allows builds to succeed for Tornado II with its egcs. [Bug 206] * ace/streams.h: added #includes of istream.h, ostream.h, and iomanip.h with ACE_HAS_STD_CPP_LIBRARY and ACE_USES_OLD_IOSTREAMS. It's necessary for builds for Tornado II, but should be harmless and helpful on other platforms. [Bug 206] Thanks to Matthias Schumann for first reporting the solution to the build problem for Tornado II. Thu Aug 5 13:03:28 1999 Douglas C. Schmidt * ace/Malloc_T.cpp (shared_free): Looks like ACE_reinterpret_cast isn't the right thing to do... Let's just use good ol' fashioned C-style casts instead. Thanks to Bala for reporting this. Thu Aug 05 12:05:42 1999 David L. Levine * include/makeinclude/wrapper_macros.GNU: moved minimum_corba make flag support from wrapper_macros.GNU to TAO/rules.tao.GNU. Thanks to Bala for noticing this. [Bug 204] Thu Aug 05 12:01:14 1999 Nanbor Wang * ace/Malloc_T.cpp (shared_free): Applied ACE_reinterpret_cast (ACE_Malloc_Header*, yadda) to serveral pointer comparisons. MSVC was complaining because the pointers are of different types. Thu Aug 5 11:57:23 1999 Balachandran Natarajan * include/makeinclude/wrapper_macros.GNU (YACC): Removed the definition of interface_repo a command line option with 'make' to $TAO_ROOT/rules.tao.GNU as per Dr.Levine's suggestions. Thu Aug 05 08:43:27 1999 David L. Levine * include/makeinclude/platform_vxworks5.x_diab.GNU: added. Thanks to Erik Johannes for helping to figure out its contents. Note that compiles succeed with Diab 4.2a for VxWorks, but links do not, yet. Thu Aug 05 08:37:09 1999 David L. Levine * ACE-INSTALL.html: added Kirk Davies' notes for running the ACE tests on Tornado II, without NFS installed. Thu Aug 5 00:15:45 1999 Douglas C. Schmidt * ace/Based_Pointer_T: Changed the type of index for operator[] and operator+= to be int instead of long so that it'll work correctly when constant values are used. Thanks to Nanbor and KCC for pointing this out! Wed Aug 4 19:46:33 1999 Carlos O'Ryan * tests/run_tests.sh: Disabled the New_Fail_Test because it breaks other compilations. Wed Aug 4 15:56:51 1999 Douglas C. Schmidt * tests/Malloc_Test.cpp: Make this test work even if ACE_HAS_POSITION_INDEPENDENT_MALLOC isn't set for a particular platform. * tests/Malloc_Test.h: Added a #include for "ace/Based_Pointer_T.h". Thanks to Carlos for pointing out the need for this. Wed Aug 4 17:07:12 1999 Balachandran Natarajan * include/makeinclude/wrapper_macros.GNU : Made the interface repository files to be compiled by default. If the user does not want these files to be compiled he can turn it off by issuing interface_repo=1 with the 'make' command. Wed Aug 4 15:56:51 1999 Douglas C. Schmidt * ace/Task.h (ACE_Task_Base): Changed the parameter name to , which is more descriptive. Thanks to John Forest for motivating this. Wed Aug 4 15:54:51 1999 Carlos O'Ryan * netsvcs/clients/Naming/Client/Makefile: * netsvcs/clients/Naming/Dump_Restore/Makefile: If the libraries are specified as dependencies then they don't need to be listed in LDLIBS Wed Aug 4 13:20:54 1999 Douglas C. Schmidt * ace/Malloc.h: Ok, let's try this again ;-). Removed the forward declarations and typedefs and replaced them with #defines. Hopefully, this will make all the compilers happy! Wed Aug 04 12:58:53 1999 Nanbor Wang * tests/*.dsp: Fixed Alpha configurations. Wed Aug 4 12:26:53 1999 Carlos O'Ryan * ace/Malloc.h: Added more forward declarations for the compilations that enabled PIM. Wed Aug 4 11:59:57 1999 Carlos O'Ryan * ace/Malloc.h: The previous fix did not work, i used forward declarations to make it work. Wed Aug 4 09:28:35 1999 Douglas C. Schmidt * ace/Malloc.h: Replaced the use of typedefs for macros to define ACE_MALLOC_HEADER_PTR ACE, ACE_NAME_NODE_PTR, ACE_CHAR_PTR to try to avoid "incomplete type" messages from certain versions of G++. Macros are "later binding" than typedefs for this situation. * ace/Asynch_Acceptor.cpp (handle_accept): Fixed an expression that had incorrect parentheses. Thanks to David Digby for reporting this. Wed Aug 04 02:40:45 1999 Nanbor Wang * tests/run_tests.bat: * tests/tests.dsw: * tests/Malloc_Test.dsp: * tests/version_tests/Malloc_Test.dsp: Added project files for the Malloc_Test and added the test in run_test.bat. Tue Aug 3 22:39:34 1999 Douglas C. Schmidt * tests/{Makefile,run_tests.sh}: Added support for the new Malloc_Test. * tests/Malloc_Test.cpp: Added a new test that exercises the new position-independent ACE_Malloc with ACE_MMAP_Memory_Pool and ACE_Process_Mutex. * ace/Malloc_T: Added an accessor called mutex() to allow applications to acquire()/release() the lock used to provide mutual exclusion to an allocator. * tests/SV_Shared_Memory_Test.cpp: Reformatted this test to conform to ACE programming guidelines. * ace/Memory_Pool: The void * and char * arguments to the various ACE_*_Memory_Pool_Options constructors should be const; now they are! * examples/Shared_Malloc/test_position_independent_malloc.cpp (main): Added auto_ptr support to this test, as well. * examples/Shared_Malloc/test_multiple_mallocs.cpp: Enhanced this test so that it uses auto pointers and also now takes advantage of position-independent malloc, when possible. * examples/Shared_Malloc/test_position_independent_malloc.cpp: Finished integrating all the features of this test. It's pretty cool now! * examples/Shared_Malloc/Makefile: Updated the LSRC target so 'make depend' works correctly... * ace/Select_Reactor_Base.cpp (find): Don't set errno = ENOENT if we find a valid mapping of handle to Event_Handler. Thanks to Hao Ruan for reporting this. * ace/Based_Pointer_T: Added support for operator= (const ACE_Based_Pointer_Basic &). * ace/Based_Pointer_T: Added definitions for ACE_Based_Pointer::operator= and ACE_Based_Pointer_Basic::operator= to make EGCS happy. * ace/Malloc.h: Factored out the common code for ACE_Based_Pointer<> and the underlying types, thereby reducing the number of #ifdef's. * ace/Based_Pointer_T: Added an operator CONCRETE *() method * ace/Based_Pointer_Repository.cpp: The original implementation of this classes stored a pointer to the size of each mapped address, rather than just the value. This was causing problems when used for shared memory because the memory was allocated in one process and freed in another. This is now fixed by not storing a pointer, but by storing the value instead. Thanks to Nanbor for reporting this. Tue Aug 3 23:43:47 1999 Carlos O'Ryan * netsvcs/clients/Naming/Client/Makefile: * netsvcs/clients/Naming/Dump_Restore/Makefile: Fixed dependencies so parallel compilations (using the -j option) will work correctly. This fixes [BUGID:187] Tue Aug 3 23:23:18 1999 Kirthika Parameswaran * ace/Cached_Connect_Strategy_T.{h,cpp} (cleanup): Added a specialised method to prevent memory leaks with old versions of g++. This method calls the underlying iterator and does the cleanup. * ace/run_tests.sh: Removed purify check over Cached_Accept_Conn_Test and Cached_Conn_Test since they no longer leak memory when built using old version of g++. Tue Aug 3 22:23:18 1999 Balachandran Natarajan * include/makeinclude/wrapper_macros.GNU: Added an ifeq flag for the interface repository. Compiling with interface_repo=1 will compile the files for interface repository. Tue Aug 03 21:17:00 1999 Nanbor Wang * tests/SOCK_Connector_Test.cpp (host_is_up): Added a time out value of 5 second to spped up the test. Thanks to Jody Hagins for submitting the patch. Tue Aug 03 18:30:47 1999 Kirthika Parameswaran * ace/Strategies.h (ACE_Connection_Recycling_Strategy): Added a new pure virtual method which will provide a non-locking implementation of the method. * ace/Strategies_T.h (ACE_Cached_Connect_Strategy): * ace/Cached_Connect_Strategy_T.h (ACE_Cached_Connect_Strategy_Ex): Made public. * ace/Caching_Utility_T.{h,cpp} (ACE_Refcounted_Recyclable_Caching_Utility): Added this class which caters to purging of refcounted and recyclable handlers. * ace/Cleanup_Strategy_T.{h,cpp}: (ACE_Refcounted_Recyclable_Cleanup_Strategy): Added this class which caters to the cleanup of handlers which are refcounted and recyclable. Heres where the necessity of making the non-locking version of public arose in the first place. * ace/OS.h (ACE_Recyclable_State): Removed hardcoded values and left the enum values to the discretion of the compiler. Tue Aug 03 16:58:08 1999 Joe Hoffert * ace/ATM_Acceptor.{h,cpp}: * ace/ATM_Connector.{h,cpp}: * ace/ATM_Stream.{h,i,cpp}: * ace/ATM_Params.cpp: * ace/ATM_QoS.cpp: * ace/XTI_ATM_Mcast.{h,i,cpp}: Fixed inlining errors. Tue Aug 03 16:15:00 1999 Chris Gill * ace/Auto_Ptr.h: created ACE_AUTO_PTR_RESET macros, which depend on whether or not ACE_AUTO_PTR_LACKS_RESET is defined. * ace/config-kcc-common.h: defined ACE_AUTO_PTR_LACKS_RESET, because KCC native auto_ptr support is difficult to disable, and implements an older version of auto_ptr (without the reset method). Tue Aug 03 13:26:47 1999 Nanbor Wang * ace/Service_Repository.cpp (instance): Check to see if the ACE_Object_Manager is being shutting down before creating a new instance of Service_Repository. Otherwise, we may be creating a new service repository just for shutting it down. Thanks to Doug for noticing this. * ace/Service_Config.cpp (fini_svcs): Since it is now possible for ACE_Service_Repository::instance to return 0 during program shutdown, we must check its value before using it. Notice that the change assumes people don't do stupid things like initializing the Service_Config during program shutdown. Mon Aug 2 23:21:01 1999 Carlos O'Ryan * bin/auto_compile: Added several flags to: disable test execution, specify the list of build directories on the command line, send email on successful compilation, keep a more detailed history, etc. Mon Aug 02 18:38:37 1999 Joe Hoffert * ace/ATM_Acceptor.{h,i,cpp} : * ace/ATM_Connector.{h,i,cpp} : * ace/ATM_Params.{h,i,cpp} : * ace/ATM_QoS.{h,i,cpp} : * ace/ATM_Stream.{h,i,cpp} : * ace/XTI_ATM_Mcast.{h,i,cpp} : * ace/ATM_Addr.{h,cpp}: * ace/Makefile: * ace/TLI_Connector.i: Added new ATM wrapper files to make the mechanism of sockets or XTI transparent when using the ATM protocol. Modified existing files to take advantage of these wrapper classes. * examples/IPC_SAP/ATM_SAP : * examples/IPC_SAP/ATM_SAP/CPP-client.cpp : * examples/IPC_SAP/ATM_SAP/CPP-server.cpp : * examples/IPC_SAP/ATM_SAP/Makefile : Added new test directory and programs to test ATM wrapper classes. * examples/IPC_SAP/TLI_SAP/CPP-ATM-client.cpp: Modified existing test program to utilize new ATM_QoS class. Mon Aug 02 17:24:05 1999 Nanbor Wang * ace/OS.h (ACE_DL_TYPE): Only WinCE uses wchar_t DL type, we'll just deal with this in ACE_OS::dlsym instead of defining different types here. * ace/OS.i (dlsym): GetProcAddress is defined differently on Windows CE and other Win32 platforms so we need to treat them differently. * ace/Malloc.h (ACE_Malloc_Header): Added a dummy member function to prevent egcs from complaining "all member functions are private." * ace/Based_Pointer_Repository.cpp: Fixed the use of map iterators. * examples/Shared_Malloc/test_position_independent_malloc.cpp (main): Delete the allocator before exiting from main. * ace/Malloc.h (ACE_Malloc_Header,ACE_Name_Node,ACE_Control_Block): Added an unimplemented assignment operator for this class. Otherwise, some compilers will try to use ACE_Based_Pointer_Base::operator= which is not implemented. Mon Aug 2 09:58:20 1999 Ossama Othman * netsvcs/lib/Name_Handler.cpp (init): * netsvcs/lib/Server_Logging_Handler_T.cpp (init): * netsvcs/lib/TS_Clerk_Handler.cpp (init): Changed cast to ACE_SignalHandler to allow it to compile under gcc 2.95. Sun Aug 1 15:58:39 1999 James CE Johnson * docs/tutorials/colorize: Added 
 wrapper around generated output.

        * docs/tutorials/Chap_03/Makefile:
        * docs/tutorials/Chap_03/README:
        * docs/tutorials/Chap_03/mm.cpp:
        * docs/tutorials/Chap_03/ex03.html: mm.cpp (and ex03.html) are a
          slightly modified version of ex02.html that uses a memory mapped
          file instead of SYSV SHMEM.  To work around the lack of fork()
          in Win32, the app must be executed once in server mode
          (argv[1] == 's') and again in client mode.

Sun Aug  1 15:58:39 1999  Douglas C. Schmidt  

        * ace/Memory_Pool.cpp: If the base_addr is 0 for the
          ACE_MMAP_Memory_Pool_Options, then we will disable the
          "use_fixed_addr" option since it's going to cause big trouble to
          mmap() at location 0!  Plus, when 0 is used it's intended as an
          indication that the application wants the OS to choose the
          appropriate mapping.  This minor change in semantics works
          particularly well with the new
          ACE_HAS_POSITION_INDEPENDENT_MALLOC feature, because we can now
          let the OS choose the mapping location with out fear of begin
          mapped to a different location!

        * ace  /Memory_Pool.cpp: Conditionally include
          ace/Based_Pointer_Repository.h when
          ACE_HAS_POSITION_INDEPENDENT_MALLOC is defined.

        * ace/config-sunos5.5.h: Enabled the
          ACE_HAS_POSITION_INDEPENDENT_MALLOC macro by default.  Let's see
          whether this breaks lots of stuff ;-).

        * ace/OS.h: Moved the definition of ACE_IPPROTO_TCP down
          further in this file so that the IPPROTO_TCP macro would be
          defined.

        * ace/Based_Pointer_T.cpp: Removed an inappropriate use of
          ACE_INLINE in the *.cpp file.  Thanks to DEC UNIX C++ for
          reporting this!

Sun Aug 01 02:41:44 1999  Nanbor Wang  

        * examples/Shared_Malloc/Shared_Malloc.dsw
        * examples/Shared_Malloc/position_independent_malloc.dsp: Added
          this new project file for position independent malloc test.