1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
|
Fri Mar 28 16:13:30 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Dynamic_Service_Base.h:
* ace/Service_Gestalt.h:
Updated following a code-review.
Thu Mar 27 21:04:11 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* THANKS:
* ace/Service_Gestalt.cpp:
Adding the patch suggested in bug# 3007. Thanks to Michael Carter
<mcarter at swri dot org> for reporting and debugging.
Thu Mar 27 13:42:23 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Service_Config.h:
Replacing the use of a "magic" number for the service repository
size with the appropriate #
define
* ace/String_Base.cpp:
Discarding an unnecessary change
Tue Mar 25 16:44:29 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* examples/ASX/CCM_App/ASX_CCM_App.mpc:
* examples/ASX/CCM_App/CCM_App.cpp:
The DLL names are case-sensitive on *nix. Minor layout changes.
* ace/Parse_Node.cpp:
Fixing ACE_Stream_Node::{apply,link} to ensure the modules are linked
together and initialized correctly - see bug#2916. Moving the
initialization code out of the yacc parser, here.
* ace/Service_Config.h:
Making singleton() a public method. Updated comments on static
methods to indicate that those are now deprecated and should be
avoided in favor of the coresponding instance methods.
* ace/Service_Gestalt.cpp:
Renaming a local variable
* ace/Service_Repository.cpp:
* ace/Service_Types.cpp:
Adding/improved logging to remove_i, fini, close. Layout changes.
* ace/Svc_Conf.h:
* ace/Svc_Conf.y:
* ace/Svc_Conf_y.cpp:
Adding an overloaded yyerror that takes just a string to comply
with the changed bison template. Moving the module initialization
code out of the parser. See ACE_Stream_Node class.
* ace/Svc_Conf_Tokens.h:
Cosmetic changes (generated file)
* ace/svcconf.mpb:
Updated for Bison 2.3
Sat Mar 22 04:27:10 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Service_Object.cpp:
* ace/Service_Repository.cpp:
* ace/Service_Types.cpp:
Improved logging and layout changes
Fri Mar 21 21:13:32 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
Trying to be frugal and save space by packing the SR after a call
to remove() causes problems with relocating dependent
services. For the relocation to work, we cache the index where the
next service will be inserted. If something, in the mean time
removes a service (and packs the storage) it might prevent some
services from relocation. In turn, it causes finalization problems
because the non-relocated services may end up in prematurely
unmapped memory. See TAO/tests/DLL_ORB for use case.
* ace/Service_Repository.cpp:
Changed remove() to eliminate the "packing" code and updated the
few other methods, which assumed there are no "gaps" in the
service storage.
Thu Mar 20 19:42:06 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
The legacy implementation of the SR ordered the services, based on
the time of insertion. Later this was changed because the new DLL
unloading functionality caused crashes as services dependent on a
DLL were being unloaded prematurely. The change was based on
Service_Type_Dynamic_Guard stack based instances, capturing the
intended position of a DLL-based service. As the DLL service
loaded, i
t was purposely inserted at the saved position. Combined
with a reversing the order of finalization, that _almost_ worked.
Later, the concept was enhanced with a mechanism that relies on
the DLL_Handle's reference counting to keep a DLL mapped in memory
until any dependent services were finalized. This eliminated the
need to artificially reorder the services, but the code to do that
remained active.
However, inserting the first DLL service before any dependents,
caused its finalization to be performed last. This prevents the
service itself from being able to refer to its own dependent
services (if any). The problem was illustrated by Bug #3251. This
change removes the artificial reordering of dependent services.
* ace/Service_Gestalt.h:
* ace/Service_Gestalt.cpp:
Simplified Service_Type_Dynamic_Guard.
* ace/Service_Repository.h:
* ace/Service_Repository.cpp:
Simplified relocate_i() and fixed an error that was causing it to
choose incorrect ranges of service indexes to relocate.
Eliminated the boolean static_only parameter as it was always
being set to true.
Tue Mar 11 04:08:21 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/OS_NS_unistd.cpp:
ACE_OS::argv_to_string is limited in its choice of what to quote
when quoting is required. It must quote not only when the arg
contains a space, but also if it is an empty string or some other
common control characters.
Wed Mar 5 19:31:36 UTC 2008 Iliyan Jeliazkov <fi326c@ociweb.com>
* ace/ARGV.h:
* ace/ARGV.cpp:
Introducing a ctor that takes the number of parameters in argv
thus eliminating the requirement to have argv 0-terminated. This
requirement can be a hard to satisfy in cases where the argv has
been "manualy constructed", i.e. not provided by the OS
environment.
* ace/OS_NS_unistd.h:
* ace/OS_NS_unistd.cpp:
Introducing new argv_to_string which takes an explicit argc
argument and relaxes the requirement on argv (to be 0-terminated).
Mon Mar 3 00:25:37 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Dynamic_Service.h:
* ace/Dynamic_Service.cpp:
* ace/Dynamic_Service_Base.h:
* ace/Dynamic_Service_Base.cpp:
* ace/Service_Config.h:
* ace/Service_Config.inl:
* ace/Service_Config.cpp:
* ace/Service_Gestalt.h:
* ace/Service_Gestalt.inl:
* ace/Service_Gestalt.cpp:
Expanding the interface of process_directives() and parse_args_i()
with a parameter used to return information whether the standard
svc.conf file must be ignored, due to explicitely specified
configuration file. The open_i() method now passes this piece of
information to process_directives(). Reversing the "naked" pointer
replacement as it causes unexpected reference counting issues with
stack-allocated instances on SC.
Fri Feb 29 03:07:36 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Dynamic_Service.h:
* ace/Dynamic_Service.cpp:
* ace/Dynamic_Service_Base.h:
* ace/Dynamic_Service_Base.cpp:
* ace/Service_Config.h:
* ace/Service_Config.inl:
* ace/Service_Config.cpp:
Replacing the "naked" ACE_Service_Gestalt* with the "smart"
equivivalent.
* ace/Service_Gestalt.h:
* ace/Service_Gestalt.inl:
* ace/Service_Gestalt.cpp:
Expanding the interface of parse_args_i() with a parameter used to
return information whether the standard svc.conf file must be
ignored, due to explicitely specified configuration file. The
open_i() method now passes this piece of information to
process_directives().
Wed Feb 27 14:57:50 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* tests/Bug_2980_Regression_Test.cpp:
* tests/tests.mpc:
Replaced the usage of ACE_WIN32 with the platform-specific WIN32
since the driver is ACE-unaware application. Updated the test to
compile cleanly on Windows.
Tue Feb 26 13:36:47 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* tests/Bug_2980_Regression_Dll.cpp:
* tests/Bug_2980_Regression_Test.cpp:
* tests/tests.mpc:
The driver is a non-ACE executable (by design) and it doesn't have
the "standard" output stream of an ACE test. Therefore it is not
producing a test log file, which the test framework interprets as
an error. Changed the ACE-based DLL part to do the proper magic.
Tue Feb 26 02:42:23 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* tests/Bug_2980_Regression_Test.cpp:
Fixing a compile problem on Windows
Sun Feb 24 23:11:04 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Unbounded_Queue.cpp:
Reverted the last change since no other member function could
account for making head_ a 0. Interestingly, after removing the
code explicitely settiong head_ to 0 from the dtor the problem
"went away". Now, the test mentioned previosly works. WTF?
Sat Feb 23 03:34:30 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Unbounded_Queue.cpp:
This one will be tough to explain easily. Adding a check for head
!= 0 before trying to dereference it. The background is that
TAO/orbsvcs/tests/Notify/Ordering test is using a static instance
of Routing_Slip_Queue (orbsvcs/Notify), which is being destroyed
upon process termination. This is bad in itself because it causes
indeterminism of the process behavior. The problem still lies with
Unbounded_Queue though, because it was assumming the head_ member
is _always_ != 0. This however doesn't hold true, for example
perhaps by calling dequeue_head() enough times.
* bin/diff-builds.py:
Simplified argument passing and parsing.
Fri Feb 22 13:29:29 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* bin/diff-builds.py:
Fixed a bug causing confusion with the datestamps available for
comparison
* tests/Bug_2980_Regression_Dll.cpp:
* tests/Bug_2980_Regression_Test.cpp:
* tests/run_test.lst:
* tests/tests.mpc:
Addded a test for bug 2980. Thanks to Lothar Werzinger <lothar at
tradescape biz> for contributing the code.
Wed Feb 20 15:05:01 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* tests/Intrusive_Auto_Ptr_Test.cpp:
Fixed a compiler warning about unused parameter to main()
* tests/Object_Manager_Flipping_Test.cpp:
ACE::{init,fini} may or may not destroy the default service
configuration context. The behavior is platform dependent, based
on how ACE has been initialized. Windows main() is modified to
automatically call ACE::{init,fini} while *nix (generally)
isn't. In the first case, the test's (second) ACE::init is
ignored. While in the latter case, the tests ACE::init and
ACE::fini actually perform work and completely destroy the Object
Manager.
Tue Feb 19 21:00:00 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Service_Config.h:
Reinstated the Service_Gestalt #include to placate the VC8
compiler on Windows, which apparently cannot typedef a template
instantiation without being able to see the full definition of the
participating types.
Tue Feb 19 02:38:02 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Base_Thread_Adapter.h:
* ace/Base_Thread_Adapter.cpp:
* ace/OS_Thread_Adapter.h:
* ace/OS_Thread_Adapter.cpp:
Moved the gestalt-setting code to the base thread adapter class
to ensure all descendants behave uniformly.
* bin/diff-builds.py:
Added file caching and debugging.
Sun Feb 17 23:43:43 UTC 2008 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/OS_Thread_Adapter.h:
* ace/OS_Thread_Adapter.cpp:
Added a pointer instance member, which the ctor intializes with
the "current" service configuration context (gestalt). Current
means the one active in the "parent" thread. The invoke() method,
which runs in the new thread and uses the stored value to
initialize its own configuration context. This scheme ensures any
newly spawned thread would inherit the spawning thread's gestalt.
* ace/Service_Config.h:
* ace/Service_Config.cpp:
Encapsulated thread-specific key operations within a new
ACE_Threading_Helper class. It is designed to be used as an
instance member in ACE_Service_Config.
* ace/Service_Gestalt.cpp:
Moved the functionality dealing with the default service
configuration file from ACE_Service_Config::open_i() to
ACE_Service_Gestalt::open_i() which fixes an issue with some
application that inadverently ignored svc.conf's content (unless
explicitely calling ACE_Service_Config::open())
* ace/Parse_Node.h:
* ace/Service_Config.inl:
Updated the #include directive.
Fri Dec 28 22:17:18 UTC 2007 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Dynamic_Service.h:
* ace/Dynamic_Service.cpp:
* ace/Dynamic_Service_Base.h:
* ace/Dynamic_Service_Base.cpp:
* ace/Dynamic_Service_Dependency.cpp:
* ace/Service_Config.h:
* ace/Service_Config.inl:
* ace/Service_Config.cpp:
* ace/Service_Gestalt.cpp:
* ace/Thread_Adapter.cpp:
Changed the Service Config interfaces back to using plain pointers
instead of auto pointers for passing the Service Gestalt instances
in and out. The lifecycle of SG's is determined by the lifecycle
of SC instance. Anyone else willing to keep a reference should use
the Intrusive_Auto_Ptr template to capitalize on and participate
in the reference counting of SG's
Thu Dec 27 15:06:37 UTC 2007 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Intrusive_Auto_Ptr.inl:
* ace/Service_Config.inl:
* ace/Service_Config.cpp:
* ace/Service_Gestalt.cpp:
Started refcount in SG at 0.
Tue Dec 25 14:04:22 UTC 2007 Iliyan Jeliazkov <iliyan@ociweb.com>
* ace/Intrusive_Auto_Ptr.inl:
* ace/Service_Config.inl:
* ace/Service_Config.cpp:
* ace/Service_Gestalt.h:
* ace/Service_Gestalt.cpp:
Refining the reference count mechanism for SC
Mon Dec 17 16:03:22 UTC 2007 Iliyan Jeliazkov <iliyan@ociweb.com>
* .:
Started tracking the trunk with svnmerge.
* ace/Service_Config.cpp:
* ace/Service_Gestalt.h:
* ace/Service_Gestalt.cpp:
* ace/String_Base.cpp:
Working on the reference count mechanism
Sun Dec 16 20:48:39 UTC 2007 Iliyan Jeliazkov <iliyan@ociweb.com>
Changed the Service Config to use the PIMPL pattern instead of
inheriting from Service Gestalt. This allowed the use of a
reference counting to manage the lifetime of a Gestalt. This
begins to solve the problem discussed in in bugzilla 2980, whereby
the root service configuration context (gestalt) may end up being
finalized either by the main thread, or by the thread specific
storage cleanup.
Added an intrusive auto pointer implementation. It is a reference
counted auto pointer that can be used for types with explicit
reference management implementations. Updated interfaces to use
the new ref-counted gestalt.
* ace/Dynamic_Service.h:
* ace/Dynamic_Service.inl:
* ace/Dynamic_Service.cpp:
* ace/Dynamic_Service_Base.h:
* ace/Dynamic_Service_Base.cpp:
* ace/Dynamic_Service_Dependency.cpp:
* ace/Intrusive_Auto_Ptr.h:
* ace/Intrusive_Auto_Ptr.inl:
* ace/Intrusive_Auto_Ptr.cpp:
* ace/Refcounted_Auto_Ptr.h:
* ace/Refcounted_Auto_Ptr.inl:
* ace/Service_Config.h:
* ace/Service_Config.inl:
* ace/Service_Config.cpp:
* ace/Service_Gestalt.h:
* ace/Service_Gestalt.cpp:
* ace/Thread_Adapter.cpp:
* ace/ace.mpc:
* tests/Intrusive_Auto_Ptr_Test.cpp:
Added a new test for the intrusive reference counted auto pointer.
* tests/Object_Manager_Flipping_Test.cpp:
* tests/tests.mpc:
Updated to use the intrusive auto pointer
Local Variables:
mode: change-log
add-log-time-format: (lambda () (progn (setq tz (getenv "TZ"))
(set-time-zone-rule "UTC") (setq time (format-time-string "%a %b %e %H:%M:%S %Z %Y" (current-time))) (set-time-zone-rule tz) time))
indent-tabs-mode: nil
End:
|