ruby.git/vm_trace.c, branch v2_1_5

merge revision(s) r46419,r46429: [Backport #9940]

2014-07-30T15:52:15+00:00

	* vm_trace.c: clear and restore recursive checking thread local data
	  to avoid unexpected throw from TracePoint.
	  [Bug #9940]

	* test/ruby/test_settracefunc.rb: add a test.

	* thread.c: adde
	  * rb_threadptr_reset_recursive_data(rb_thread_t *th);
	  * rb_threadptr_restore_recursive_data(rb_thread_t *th, VALUE old);

	* vm_core.h: ditto.

	* thread.c: added


git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/branches/ruby_2_1@47009 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

vm_trace.c: isolate exceptions

2013-12-17T07:19:25+00:00

* vm_trace.c (rb_postponed_job_flush): isolate exceptions in
  postponed jobs and restore outer ones.  based on a patch by
  tarui.  [ruby-core:58652] [Bug #9168]

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44260 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

* vm_trace.c (rb_suppress_tracing): Fix initialization of stack

2013-12-04T04:05:15+00:00

  allocated rb_trace_arg_t structure. Without this patch, sometimes
  INTERNAL_EVENT_GC would be skipped accidentally inside
  rb_threadptr_exec_event_hooks_orig().

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43987 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

* vm_trace.c: skip "exception check" and "reentrant check (only normal

2013-11-26T09:42:04+00:00

  events)  for internal events.
  Reentrant check for internal events are remaining.



git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43855 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

* vm_trace.c: prohibit to specify normal events and internal events

2013-11-26T08:41:44+00:00

  simultaneously.
  I will introduce special care for internal events later.
* ext/-test-/tracepoint/tracepoint.c: test this behavior.
* test/-ext-/tracepoint/test_tracepoint.rb: ditto.



git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43854 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

* vm_trace.c (symbol2event_flag): add secret feature.

2013-11-10T18:19:41+00:00

  add a_call/a_return events.
  a_call is call | b_call | c_call, and same as a_return.



git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43645 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

vm_trace.c: isolate errinfo in trace blocks

2013-11-07T00:39:56+00:00

* vm_trace.c (rb_threadptr_exec_event_hooks_orig): errinfo should not
  be propagated to trace blocks so that no argument raise does not
  throw internal objects.  [ruby-dev:47793] [Bug #9088]

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43561 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

* add RUBY_TYPED_FREE_IMMEDIATELY to data types which only use

2013-10-29T11:16:54+00:00

  safe functions during garbage collection such as xfree().
  On default, T_DATA objects are freed at same points as fianlizers.
  This approach protects issues such as reported by [ruby-dev:35578].
  However, freeing T_DATA objects immediately helps heap usage.
  Most of T_DATA (in other words, most of dfree functions) are safe.
  However, we turned off RUBY_TYPED_FREE_IMMEDIATELY by default
  for safety.
* cont.c: ditto.
* dir.c: ditto.
* encoding.c: ditto.
* enumerator.c: ditto.
* error.c: ditto.
* file.c: ditto.
* gc.c: ditto.
* io.c: ditto.
* iseq.c: ditto.
* marshal.c: ditto.
* parse.y: ditto.
* proc.c: ditto.
* process.c: ditto.
* random.c: ditto.
* thread.c: ditto.
* time.c: ditto.
* transcode.c: ditto.
* variable.c: ditto.
* vm.c: ditto.
* vm_backtrace.c: ditto.
* vm_trace.c: ditto.
* ext/bigdecimal/bigdecimal.c: ditto.
* ext/objspace/objspace.c: ditto.
* ext/stringio/stringio.c: ditto.
* ext/strscan/strscan.c: ditto.



git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43466 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

* vm_trace.c (tp_free): remvoed because empty free function.

2013-10-29T09:47:06+00:00

  Use RUBY_TYPED_NEVER_FREE instead.



git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43464 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

* vm_trace.c: exterminate Zombies.

2013-10-22T06:24:54+00:00

  There is a bug that T_ZOMBIE objects are not collected.
  Because there is a pass to miss finalizer postponed job
  with multi-threading. This patch solve this issue.
* vm_trace.c (rb_postponed_job_register_one): set
  RUBY_VM_SET_POSTPONED_JOB_INTERRUPT(th) if another same job
  is registered.
  There is a possibility to remain a postponed job without
  interrupt flag.
* vm_trace.c (rb_postponed_job_register_one): check interrupt
  carefully.
* vm_trace.c (rb_postponed_job_register_one): use additional space
  to avoid buffer full.
* gc.c (gc_finalize_deferred_register): check failure.
* thread.c (rb_threadptr_execute_interrupts): check
  `postponed_job_interrupt' immediately.  There is a possibility
  to miss this flag.



git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43383 b2dd03c8-39d4-4d8f-98ff-823fe69b080e