th->base_block is information for (a) parsing, (b) compiling
and (c) setting up the frame to execute the program passed by
`eval' method. For example, (1) parser need to know up-level
variables to detect it is variable or method without paren.
Befor (a), (b) and (c), VM set th->base_block by passed bindng
(or previous frame information). After execute (a), (b) and (c),
VM should clear th->base_block. However, if (a), (b) or (c)
raises an exception, then th->base_block is not cleared.
Problem is that the uncleared value th->balo_block is used for
irrelevant iseq compilation. It causes SEGV or critical error.
I tried to solve this problem: to clear them before exception,
but finally I found out that it is difficult to do it (Ruby
program can be run in many places).
Because of this background, I set th->base_block before
compiling iseq and restore it after compiling.
Basically, th->base_block is dirty hack (similar to global
variable) and this patch is also dirty.
* bootstraptest/test_eval.rb: add a test for above.
* internal.h: remove unused decl.
* iseq.c (rb_iseq_compile_with_option): add base_block parameter.
set th->base_block before compation and restore it after
compilation.
* ruby.c (require_libraries): pass 0 as base_block instead of
setting th->base_block
* tool/compile_prelude.rb (prelude_eval): apply above changes.
* vm.c, vm_eval.c: ditto.
* vm_core.h: add comments.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@36179 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Before this commit:
`finish frame' was place holder which indicates that VM loop
needs to return function.
If a C method calls a Ruby methods (a method written by Ruby),
then VM loop will be (re-)invoked. When the Ruby method returns,
then also VM loop should be escaped. `finish frame' has only
one instruction `finish', which returns VM loop function.
VM loop function executes `finish' instruction, then VM loop
function returns itself.
With such mechanism, `leave' instruction (which returns one
frame from current scope) doesn't need to check that this `leave'
should also return from VM loop function.
Strictly, one branch can be removed from `leave' instructon.
Consideration:
However, pushing the `finish frame' needs costs because
it needs several memory accesses. The number of pushing
`finish frame' is greater than I had assumed. Of course,
pushing `finish frame' consumes additional control frame.
Moreover, recent processors has good branch prediction,
with which we can ignore such trivial checking.
After this commit:
Finally, I decide to remove `finish frame' and `finish'
instruction. Some parts of VM depend on `finish frame',
so the new frame flag VM_FRAME_FLAG_FINISH is introduced.
If this frame should escape from VM function loop, then
the result of VM_FRAME_TYPE_FINISH_P(cfp) is true.
`leave' instruction checks this flag every time.
I measured performance on it. However on my environments,
it improves some benchmarks and slows some benchmarks down.
Maybe it is because of C compiler optimization parameters.
I'll re-visit here if this cause problems.
* insns.def (leave, finish): remove finish instruction.
* vm.c, vm_eval.c, vm_exec.c, vm_backtrace.c, vm_dump.c:
apply above changes.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@36099 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* include/ruby/ruby.h: public symbols must have default visibility.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@36081 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
dfp (dynamic frame pointer) to ep (environment pointer).
This change make VM `normal' (similar to other interpreters).
Before this commit:
Each frame has two env pointers lfp and dfp. lfp points
local environment which is method/class/toplevel frame.
lfp[0] is block pointer.
dfp is block local frame. dfp[0] points previous (parent)
environment pointer.
lfp == dfp when frame is method/class/toplevel.
You can get lfp from dfp by traversing previous environment
pointers.
After this commit:
Each frame has only `ep' to point respective enviornoment.
If there is parent environment, then ep[0] points parent
envioenment (as dfp). If there are no more environment,
then ep[0] points block pointer (as lfp). We call such ep
as `LEP' (local EP). We add some macros to get LEP and to
detect LEP or not.
In short, we replace dfp and lfp with ep and LEP.
rb_block_t and rb_binding_t member `lfp' and `dfp' are removed
and member `ep' is added.
rename rb_thread_t's member `local_lfp' and `local_svar' to
`root_lep' and `root_svar'.
(VM_EP_PREV_EP(ep)): get previous environment pointer. This macro
assume that ep is not LEP.
(VM_EP_BLOCK_PTR(ep)): get block pointer. This macro assume
that ep is LEP.
(VM_EP_LEP_P(ep)): detect ep is LEP or not.
(VM_ENVVAL_BLOCK_PTR(ptr)): make block pointer.
(VM_ENVVAL_BLOCK_PTR_P(v)): detect v is block pointer.
(VM_ENVVAL_PREV_EP_PTR(ptr)): make prev environment pointer.
(VM_ENVVAL_PREV_EP_PTR_P(v)): detect v is prev env pointer.
* vm.c: apply above changes.
(VM_EP_LEP(ep)): get LEP.
(VM_CF_LEP(cfp)): get LEP of cfp->ep.
(VM_CF_PREV_EP(cfp)): utility function VM_EP_PREV_EP(cfp->ep).
(VM_CF_BLOCK_PTR(cfp)): utility function VM_EP_BLOCK_PTR(cfp->ep).
* vm.c, vm_eval.c, vm_insnhelper.c, vm_insnhelper.h, insns.def:
apply above changes.
* cont.c: ditto.
* eval.c, eval_intern.h: ditto.
* proc.c: ditto.
* thread.c: ditto.
* vm_dump.c: ditto.
* vm_exec.h: fix function name (on vm debug mode).
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@36030 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Current VM doesn't use this bit.
* vm_core.h (RUBY_VM_GET_BLOCK_PTR): added.
* eval.c (rb_block_given_p): use RUBY_VM_GET_BLOCK_PTR().
* vm_eval.c (rb_f_block_given_p): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@35908 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
(1) rename rb_location_t to rb_iseq_location_t.
(2) rename field names of rb_iseq_location_t to adjust
RubyVM::Backtrace::Location methods.
(2-1) filename -> path
(2-2) filepath -> absolute_path
(2-3) basename -> base_label
(2-4) name -> label
(3) rename filed name rb_iseq_location_t#line_no to
rb_iseq_location_t#first_lineno to clear purpose of this field.
(4) The field names rb_binding_t#(filename|line_no) are also renamed
to rb_binding_t#(path|first_lineno).
* compile.c: apply above changes.
* iseq.c: ditto.
* proc.c: ditto.
* vm*.c: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@35899 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
is no longer included from vm.c). I hope this separation reduce
compile time of vm.c.
* internal.h: ditto.
* vm.c, vm_core.h, vm_dump.c, vm_eval.c: ditto.
* vm_eval.c: some functions (callee, etc) moved to vm_backtrace.c.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@35871 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
(1) unify similar functions (rb_backtrace_each() and
backtrace_object()). backtrace_each() is a unified function.
variation:
a) backtrace_object(): create backtrace object.
b) vm_backtrace_str_ary(): create bt as an array of string.
c) vm_backtrace_print(): print backtrace to specified file.
d) rb_backtrace_print_as_bugreport(): print backtrace on
bugreport style.
(2) remove rb_backtrace_each(). Use backtrace_each() instead.
(3) chang the type of lev parameter to size_t.
a) lev == 0 means current frame (exception, etc use it).
b) lev == 1 means upper frame (caller(0) use it).
* vm_core.h, vm_dump.c, vm_eval.c: ditto.
* vm.c (backtrace_object(), vm_backtrace_str_ary()): fix to return a
correct size of caller(lev) array.
Let n be a "caller(0).size" then ln as caller(lev).size should be
(n - lev). However, the previous implementation returns a wrong
size array (ln > n - lev). [ruby-dev:45673]
* test/ruby/test_backtrace.rb: add tests for backtrace.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@35787 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
information.
rb_location_t#filename, filepath, name and line_no was moved from
rb_iseq_t. rb_location_t#basename is a new field which is
similar to `name' field without any decoration.
`name' field contains some decoration such as `block in foo'.
`basename' only contains `foo'.
rb_iseq_t contains memory object of rb_location_t.
* iseq.c: setup rb_location_t for each rb_iseq_t memory objects.
* compile.c, proc.c, vm.c, vm_dump.c, vm_eval.c, vm_insnhelper.c,
vm_method.c: support about it.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@35756 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Patch by Google Inc. [ruby-core:45073].
* configure.in (RUBY_NACL): New M4 func to configure variables for
NaCl.
(RUBY_NACL_CHECK_PEPPER_TYPES): New M4 func to check the old names
of Pepper interface types.
(BTESTRUBY): New variable to specify which ruby should be run on
"make btest". NaCl can run the built binary by sel_ldr, but it need
rbconfig.rb. So this variable is distinguished from $MINIRUBY.
* thread_pthread.c: Disabled some features on NaCl.
* io.c: ditto.
* process.c: ditto.
* signal.c: ditto.
* file.c: ditto.
* missing/flock.c: ditto.
* nacl/pepper_main.c: An example implementation of Pepper application
that embeds Ruby.
* nacl/example.html: An example of web page that uses the Pepper
application.
* nacl/nacl-config.rb: Detects variants of NaCl SDK.
* nacl/GNUmakefile.in: Makefile template for NaCl specific build
process.
* nacl/package.rb: script for packaging a NaCl-Ruby embedding
application.
* nacl/reate_nmf.rb: Wrapper script of create_nmf.py
* dln.c (dln_load): Added a hack to call on NaCl.
* util.c (ruby_getcwd): Path to the current directort is not available
on NaCl.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@35672 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread.c (rb_thread_schedule_limits): set th#yielding while
release GVL to yield CPU time.
* thread.c (timer_thread_function): skip timer interrupt when
th#yielding is true. This patch fixes r35480.
* thread.c (rb_threadptr_execute_interrupts_common): revert
a patch of r35480.
* ChangeLog: add an extended memo of r35480.
http://bugs.ruby-lang.org/projects/ruby-trunk/wiki/R35480_ExtendedMemo
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@35486 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
to solve an issue reported at [ruby-dev:44413] [Ruby 1.9 - Bug #5217].
Before this fix, each instruction has an information including
line number (iseq::iseq_insn_info_table). Instead of this data
structure, recording only line number changing places
(iseq::iseq_line_info_table).
The order of entries in iseq_line_info_table is ascending order of
iseq_line_info_table_entry::position. You can get a line number
by an iseq and a program counter with this data structure.
This fix reduces memory consumption of iseq (bytecode).
On my measurement, a rails application consumes 21.8MB for
iseq with this fix on the 32bit CPU. Without this fix, it
consumes 24.7MB for iseq [ruby-dev:44415].
* proc.c: ditto.
* vm_insnhelper.c: ditto.
* vm_method.c: ditto.
* vm.c (rb_vm_get_sourceline): change to use rb_iseq_line_no().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@33046 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
NoMemoryError is better than rb_bug, of course.
Patch by Eric Wong. [ruby-core:38572][ruby-core:38594].
* signal.c (rb_register_sigaltstack): ditto.
* vm_core.h: moved ALT_STACK_SIZE definition from signal.c.
* vm.c (thread_free): use xfree() instead of free().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@32749 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
(rb_gc_mark_unlinked_live_method_entries): fix SEGV bug.
rb_method_entry_t was free'd even when the method is still on the
stack if it is BMETHOD (i.e., Method#call). This is because
rb_method_entry_t is embedded in struct METHOD. This commit
separates them and marks the live method entries.
See [ruby-core:38449] in detail. fix [Bug #5047] [ruby-core:38171]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@32669 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
'inhibit_thread_createion' field.
* thread.c (rb_thread_terminate_all): set inhibit_thread_creation.
* thread.c (thread_s_new): don't permit to create new thread
if the VM is under destruction. Otherwise evil finalizer code
can make SEGV. [Bug #4992][ruby-core:37858]
* bootstraptest/test_objectspace.rb: new test for this fix.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@32492 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread.c (rb_threadptr_execute_interrupts_common): check signal
deliverly if it is main thread.
fixes [ruby-dev:44005] [Ruby 1.9 - Bug #4950]
* bootstraptest/test_fork.rb: add a test for above.
* signal.c (rb_get_next_signal): skip if signal_buff is empty.
(check signal_buff.size first)
* vm_core.h: remove unused variable rb_thread_t::exec_signal.
* thread.c (rb_thread_check_trap_pending): check
rb_signal_buff_size() because rb_thread_t::exec_signal
is no longer available.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@32345 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
no waiting thread. If there are 2 or more runnable threads,
the timer thread does polling. Avoid polling makes power save
for several computers (0.2W per a Ruby process, when I measured).
If outside-event such as signal or Thread#kill was occuerred
when the timer thread does not do polling, then wake-up
the timer thread using communication-pipe (the timer thread
waits this communication-pipe with select(2)).
The discussion about this modification can be found from the post
[ruby-core:33456] and other related posts.
Note that Eric Wong and KOSAKI Motohiro give us the huge
contributions for this modification. Thanks.
* thread_pthread.c (rb_thread_wakeup_timer_thread): add a function.
This function wakes up the timer thread using communication-pipe.
* thread.c (rb_thread_stop_timer_thread): add a parameter which
specify closing communication-pipe or not.
* thread.c (rb_thread_terminate_all): do not stop timer thread here
(ruby_cleanup() terminate timer thread).
* signal.c: wake up timer thread using
rb_thread_wakeup_timer_thread() from signal handler.
* eval.c (ruby_cleanup): use rb_thread_stop_timer_thread(1).
* process.c: use rb_thread_stop_timer_thread(0)
(reuse communication-pipe).
* thread_win32.c (rb_thread_wakeup_timer_thread): add a dummy
function.
* vm_core.h: add and fix decl. of functions.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@32244 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
ruby_vm_global_state_version and related macros
from vm_core.h to vm_insnhelper.h.
* vm.c (vm_clear_all_cache): added. This function is called
when ruby_vm_global_state_version overflows.
TODO: vm_clear_all_inline_method_cache() is only place holder.
We need to implement it ASAP.
* vm_method.c (vm_clear_global_method_cache): added.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@32056 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
based on the patch by Eric Wong. [ruby-core:36156]
* vm_core.h (rb_thread_call_with_gvl): don't declare here.
* thread.c: include internal.h.
(rb_thread_execute_interrupts): new function.
* internal.h (rb_thread_execute_interrupts): declared.
(rb_thread_call_with_gvl): declared.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@31971 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
CoreDataFromValue() because we need type check. Otherwise,
type mismatch can cause segmentation fault crash.
[ruby-core:35086] [Ruby 1.9-Bug#4367]
* vm.c (thread_data_type): remove static.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@30781 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
vm_insnhelper.c (caller_setup_args): reverted r30241 and r30243
except for the test.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@30283 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
