Now do_select() is free from unexpected hangup if
HAVE_RB_FD_INIT=1 [Bug #4636]
* thread.c (rb_thread_fd_select, rb_thread_wait_fd_rw):
adapt new argument type.
* thread.c (rb_thread_select): make dummy implementation.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@31396 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
after blocking region, because remaining waiting_fd might
cause unnecessary IOError.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@30873 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
blocking region with GIL released, for fd.
* thread.c (rb_thread_fd_close): implement. [ruby-core:35203]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@30852 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
resource at fork. Sadly this is purely bandaid. We need to
implement proper fix later. [Bug #4169] [ruby-core:33767]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@30272 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
APIs to modularize GVL implementation.
* thread_pthread.c, thread_pthread.h: Two GVL implementations.
(1) Simple locking GVL which is same as existing GVL.
(2) Wake-up queued threads. The wake-up order is simple FIFO.
(We can make several queues to support exact priorities, however
this causes some issues such as priority inversion and so on.)
This impl. prevents spin-loop (*1) caused on SMP environemnts.
*1: Only one Ruby thread acqures GVL again and again.
Bug #2359 [ruby-core:26694]
* thread_win32.c, thread_win32.h: Using simple lock
not by CRITICAL_SECTION but by Mutex.
Bug #3890 [ruby-dev:42315]
* vm.c (ruby_vm_destruct): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@29956 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
the func with normal state. a patch from Satoshi Shiba <shiba
AT rvm.jp> at [ruby-dev:42162]. [ruby-core:31783]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@29155 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
failed realloc by using xrealloc instead of not realloc. a patch
from Jim Meyering <meyering at redhat.com> in [ruby-core:30920]
[Bug #3489]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@28457 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
by switching the thread if gc_stress == true
* gc.c (ruby_gc_stress_start): dotto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@28078 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
call during GVL state transition. [Bug#3354] [ruby-dev:41435]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@28055 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
immediately.
* win32/win32.c (subtract): ditto.
based on a patch from Roger Pack in [ruby-core:27957].
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@27930 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Lower-priority thread may run even if there are higher-priority
threads. See [ruby-dev:40977].
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@27351 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
of event_flags. [ruby-core:25191]
* thread.c (rb_threadptr_exec_event_hooks): delete RUBY_EVENT_VM if
all event_hooks are removed.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@27033 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
before clean up. [ruby-core:26877]
* thread.c (rb_thread_atfork): no other threads to be joined.
* vm_core.h (rb_thread_lock_unlock, rb_thread_lock_destroy):
new functions.
* vm.c (ruby_vm_destruct): unlock and destroy global VM lock.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@26390 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* defs/dtrace.d: new file. defined a dtrace provider "ruby".
* include/ruby/ruby.h (LIKELY): moved from vm.c.
(UNLIKELY): ditto.
(OBJSETUP): probe "object-create".
(RUBY_EVENT_RESCUE): new event.
* vm_exec.c (DEBUG_ENTER_INSN): embeded a probe insn-entry into it.
(DEBUG_END_INSN): insn-return.
* vm.c (LIKELY): moved into ruby.h.
(UNLIKELY): ditto.
(Init_BareVM): embeded a probe "raise" into it.
* variable.c (rb_class2name_without_alloc): new utility function.
* tool/rbinstall.rb (install?(:ext, :arch, :'ext-arch')): installs
dtrace.d if necessary.
* thread_pthread.c (add_signal_thread_list): probe "raise".
(rb_thread_create_timer_thread): ditto.
* thread.c (rb_thread_schedule_rec): probes "thread-enter" and
"thread-leave",
(thread_start_func_2): ditto.
(thread_cleanup_func): probe "thread-term"
* lib/mkmf.rb: supports dtrace postprocessor on making an extension.
* iseq.c (rb_vm_insn_name): new utility function.
(rb_vm_insn_len): ditto.
* insns.def (hook): probes "method-etnry", "method-return", "line",
and "rescue".
* compile.c (iseq_compile_each): adds a trace op for "rescue" probe.
* gc.c (garbage_collect): probes "gc-begin" and "gc-end".
(obj_free): probe "object-free"
(garbage_collect_with_gvl): probe "raise"
(negative_size_allocation_error): ditto.
(rb_memerror): ditto.
* eval.c (rb_rescue2): probe "rescue"
(rb_longjmp): probe "raise"
* ext/probe/probe.c: new extension for application defined probes.
* ext/probe/extconf.rb: ditto.
* configure.in (--with-tracing-model): new option to choose a tracing
mechanism.
(DTRACE): new substitution. name of dtrace(1).
(RUBY_TRACING_MODEL): new substitution.
(DTRACE_OBJ): ditto.
(MINIDTRACE_OBJ): ditto.
(GOLFDTRACE_OBJ): ditto.
(LIBRUBY_DTRACE_OBJ): ditto.
(RUBY_DTRACE_POSTPROCESS): new macro. checks whether the dtrace on
the system needs postprocessing.
(RUBY_DTRACE_BSD_BROKEN): new macro. checks whether the dtrace
supports USDT.
* Makefile.in:
(DTRACE): new variable. name of dtrace(1).
(TRACING_MODEL): new variable. name of the chosen tracing mechanism.
(DTRACE_OBJ): same as the one in configure.in.
(MINIDTRACE_OBJ): ditto.
(GOLFDTRACE_OBJ): ditto.
(LIBRUBY_DTRACE_OBJ): ditto.
(CPPOUTFILE): new substitution. necessary for generating dtrace.d
(trace_none.h): new target for TRACING_MODEL=none
(RUBY_H_INCLUDES): appended a header for tracing.
(distclean-local): also removes preprocessed version of dtrace.d
($(LIBRUBY_A)): needs $(LIBRUBY_DTRACE_OBJ) if dtrace needs
postprocessing.
($(PROGRAM)): ditto.
(golf): ditto.
(miniruby): ditto.
($(arch_hdrdir)/ruby/dtrace.d): new target. preprocessed verson
of defs/dtrace.d. generated if necessary.
($(arch_hdrdir)/ruby/trace_dtrace.h): new target.
definition of probes.
($(LIBRUBY_DTRACE_OBJ)): new target. generated if dtrace needs
postprocessing.
($(DTRACE_OBJ)): ditto.
($(MINIDTRACE_OBJ)): ditto.
($(GOLFDTRACE_OBJ)): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@26235 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
at fork to get rid of deadlock. based on the patch from Hongli
Lai in [ruby-core:26783]. [ruby-core:26361]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@25842 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
object become dummy because of fork. [ruby-core:26744]
[ruby-core:26745]
* bootstraptest/test_thread.rb: add a test for above.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@25771 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
in this function.
(RB_GC_SAVE_MACHINE_CONTEXT): call rb_gc_save_machine_context and
save the stack pointer at caller side.
(rb_thread_schedule_rec): use RB_GC_SAVE_MACHINE_CONTEXT instead of
rb_gc_save_machine_context.
(blocking_region_begin): changed to a macro. use
RB_GC_SAVE_MACHINE_CONTEXT instead of rb_gc_save_machine_context.
[ruby-dev:39659]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@25701 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
description that no exception is allowed inside `func', instead.
see [ruby-dev:39582]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@25570 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* eval.c (get_errinfo, rb_rubylevel_thread_errinfo): Getter for current exception for a given thread
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@25278 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* test/ruby/test_thread.rb (test_recursive_outer): Test for above
* hash.c (rb_hash_hash): Return a sensible hash for in case of recursion [ruby-core:24648]
* range.c (rb_range_hash): ditto
* struct.c (rb_struct_hash): ditto
* array.c (rb_array_hash): ditto
* test/ruby/test_array.rb (test_hash2): test for above
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@24943 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
rb_exec_recursive_*. The functions recursive_push, pop and check now assume
a valid hash table as their first argument. Added documentation.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@24891 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
deadlock detection because the deadlock exception makes main thread
run. [ruby-dev:39142]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@24616 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
stack overflows in the signal handler, if sigaltstack is
available. On stack overflow (and with sigaltstack), the signal
handler is more likely to have room to create an exception
object. [ruby-core:23813]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@23720 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
"rb_thread_t *", the prefix to be rb_threadptr_ instead of
rb_thread_.
* thread.c (rb_thread_add_event_hook(), rb_thread_remove_event_hook):
change the parameter type from rb_thread_t * to VALUE.
* eval.c, eval_error.c, eval_intern.h, signal.c, vm_core.h, vm_eval.c:
ditto.
* include/ruby/intern.h: remove decl of rb_thread_signal_raise() and
rb_thread_signal_exit().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@23651 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
using rb_fdset_t.
* io.c (select_internal): use rb_thread_fd_select instead of
rb_thread_select. based on the patch from Kengo Matsuyama.
[ruby-dev:38221]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@23109 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Marcus Rueckert <darix AT opensu.se> at [ruby-core:21492].
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@21725 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
vm_get_ruby_level_next_cfp(), rb_vm_make_env_object(),
vm_stack_to_heap(), vm_make_proc(), vm_invoke_proc(),
vm_get_sourceline(), vm_cref(), vm_localjump_error(),
vm_make_jump_tag_but_local_jump(), vm_jump_tag_but_local_jump().
This changes may affect only core because most of renamed functions
require a pointer of not-exposed struct such as rb_thread_t or NODE.
In short, they are core functions.
* cont.c, eval.c, eval_intern.h, load.c, proc.c, thread.c,
vm_core.h, vm_dump.c, vm_eval.c, vm_exec.c, vm_insnhelper.c:
ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@21659 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
buffered, main thread should wait until timer thread delivers it.
* thread.c (timer_thread_function): should defer delivery of a signal
if main thread does not yet trap a previous one. [ruby-dev:37676]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@21582 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
should skip rb_str_new2() if rb_sourcefile() returns NULL.
rb_sourcefile() returns NULL if frame is toplevel of Fiber.
[ruby-core:21161] [Bug #985]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@21578 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
rb_fd_init, rb_fd_term, rb_fd_zero, rb_fd_set, rb_fd_clr, rb_fd_isset,
rb_fd_select, rb_fd_ptr, rb_fd_max, HAVE_RB_FD_INIT): new type,
functions, and macros for Windows.
* win32/win32.c (extract_fd, rb_w32_select): use rb_fdset_t to expand
fd_array if needed. [ruby-core:19946]
* win32/win32.c (copy_fd): new funcion for rb_w32_select().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@21487 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread.c (rb_thread_call_without_gvl): added as a alias of
rb_thread_blocking_region().
* thread.c (rb_thread_call_with_gvl): added.
* vm_core.h (rb_thread_t#blocking_region_buffer): added for
rb_thread_call_with_gvl().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@21185 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
a mutex that is locked by another thread, which makes the mutex
dangling in keeping_mutexes and causes [BUG] or stuck finally.
Now unlocking is performed as locking thread.
* thread.c (mutex_unlock, rb_mutex_unlock, rb_mutex_unlock_all):
mutex_unlock receives a thread.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@21105 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
thread.
* thread.c (thread_start_func_2): stops timer thread if forked in
the new thread. [ruby-core:19385]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@21097 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
loaded, to get rid of loading same library again. [ruby-core:19798]
* thread.c (rb_barrier_wait): can not wait destroyed barrier.
* thread.c (rb_barrier_destroy): destroys barrier so that no longer
waited.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@20223 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* util.c (freedtoa): used only when MULTIPLE_THREADS is not defined.
* win32/win32.c (rb_w32_pipe): serial is DWORD.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@19945 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread.c (rb_barrier_wait): achieves the lock if no thread was
waiting yet.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@19573 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* io.c: use VALUE of thead instead of rb_tread_t to check interrupts.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@19477 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
staff.
* signal.c (signal_buff): per process resouce now.
* signal.c (trap_list): moved to VM.
* signal.c (rb_get_next_signal): reverted.
* signal.c (rb_trap_exit): trap_pending_list was no longer used.
* thread.c (timer_thread_function): delivers buffered per-process
signals to each VMs.
* vm.c (rb_vm_mark): marks trap_list.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@19119 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
RUBY_UBF_IO and RUBY_UBF_PROCESS.
Because there is no default (universal) unblocking function.
* ext/socket/socket.c, file.c, io.c, process.c, thread.c: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@18985 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
this function may cause ruby's thread switching.
* include/ruby/intern.h: ditto.
* regint.h: use rb_thread_check_ints() instead of
RUBY_CHECK_INTS() directly.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@18572 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
kosaki
nobu
nagachika
akr
ko1
naruse
mame
tarui
usa
marcandre
wanabe
yugui
matz
kazu