Feature #977: caller for all threads patch

• Status: Closed
• Assignee: ko1 (Koichi Sasada)
• Target version: 1.9.2

[rogerdpack (Roger Pack)]

=begin
Here is a patch which provides backtrace for all current threads, instead of just the current one.
http://ph7spot.com/articles/caller_for_all_threads
Author said it would be great to have it accepted upstream.
Thoughts?
-=r
=end

[rogerdpack (Roger Pack)]

=begin
oops that's a feature request not a bug--for some reason I thought it would default to a feature request since that's the view from whence I clicked "Submit issue"
My bad.
-=r
=end

[ko1 (Koichi Sasada)]

• Assignee set to ko1 (Koichi Sasada)
• Target version set to 1.9.2

=begin

=end

[ko1 (Koichi Sasada)]

=begin
I made a patch to Thread#caller(lev=1). It may be more flexible than
fetching "all" backtrace.
How about it? (not tested enough)

Index: vm_eval.c

--- vm_eval.c (リビジョン 23650)
+++ vm_eval.c (作業コピー)
@@ -1342,6 +1342,19 @@ rb_make_backtrace(void)
}

VALUE
+rb_thread_backtrace(VALUE thval, int lev)
+{

• rb_thread_t *th;
• GetThreadPtr(thval, th);
• if (th->status != THREAD_KILLED && GET_THREAD() != th) {
• lev--;
• }
• return vm_backtrace(th, lev);
  +}

+VALUE
rb_backtrace_each(rb_backtrace_iter_func *iter, void *arg)
{
return vm_backtrace_each(GET_THREAD(), -1, iter, arg);
Index: thread.c

--- thread.c (リビジョン 23651)
+++ thread.c (作業コピー)
@@ -3817,6 +3817,26 @@ ruby_suppress_tracing(VALUE (*func)(VALU
return result;
}

+VALUE rb_thread_backtrace(VALUE thval, int lev);
+
+static VALUE
+rb_thread_caller_m(int argc, VALUE *argv, VALUE thval)
+{

• VALUE level;
• int lev;
• rb_scan_args(argc, argv, "01", &level);
• if (NIL_P(level))
• lev = 1;
• else
• lev = NUM2INT(level);
• if (lev < 0)
• rb_raise(rb_eArgError, "negative level (%d)", lev);
• return rb_thread_backtrace(thval, lev);
  +}

/*

• +Thread+ encapsulates the behavior of a thread of
• execution, including the main thread of the Ruby script.
  @@ -3873,6 +3893,7 @@ Init_Thread(void)
  rb_define_method(rb_cThread, "abort_on_exception=", rb_thread_abort_exc_set, 1);
  rb_define_method(rb_cThread, "safe_level", rb_thread_safe_level, 0);
  rb_define_method(rb_cThread, "group", rb_thread_group, 0);

• rb_define_method(rb_cThread, "caller", rb_thread_caller_m, -1);

  rb_define_method(rb_cThread, "inspect", rb_thread_inspect, 0);

Roger Pack wrote::

Bug #977: caller for all threads patch
http://redmine.ruby-lang.org/issues/show/977

Author: Roger Pack
Status: Open, Priority: Normal

Here is a patch which provides backtrace for all current threads, instead of just the current one.
http://ph7spot.com/articles/caller_for_all_threads
Author said it would be great to have it accepted upstream.
Thoughts?
-=r

---

http://redmine.ruby-lang.org

--
// SASADA Koichi at atdot dot net

=end

[ko1 (Koichi Sasada)]

=begin
Hongli Lai wrote::

SASADA Koichi wrote:

I made a patch to Thread#caller(lev=1). It may be more flexible than
fetching "all" backtrace.
How about it? (not tested enough)

The ability to see all running threads' backtraces, without needing a
reference to each one of those threads, is caller_for_all_thread's main
advantage. It's very useful for debugging a live application. Replacing
it with Thread#caller would require one to maintain references to all
threads that one wants to inspect. Does Ruby already provide some way to
obtain a list of all running threads?

def caller_for_all_thread
Thread.list.map{|t| t.caller}
end

--
// SASADA Koichi at atdot dot net

=end

[ko1 (Koichi Sasada)]

=begin
Rocky Bernstein wrote::

One thing I think might be cool is rather than raising an error for a
negative Fixnum value is to count from the other end. So caller(-1) is the
least-recent call.

If you want me to try my hand at extending the below, let me know.

Check the following code.

VALUE
+rb_thread_backtrace(VALUE thval, int lev)
+{

• rb_thread_t *th;
• GetThreadPtr(thval, th);
• if (th->status != THREAD_KILLED && GET_THREAD() != th) {

• lev--;
  

• }
• return vm_backtrace(th, lev);
  +}

--
// SASADA Koichi at atdot dot net

=end

[ko1 (Koichi Sasada)]

=begin
Rocky Bernstein wrote::

I was suggesting that rather than raise an error here, treat this like array
indexes do and basically use size - level. (By the way, also suggests it
might be cool to add some sort of length or size function.)

I guess I missed something, but what?

I had misunderstood your suggestion. At first, you should suggest the
"Kernel.caller" specification, not the Thread#caller spec.

Regards,

// SASADA Koichi at atdot dot net

=end

[ko1 (Koichi Sasada)]

=begin
Roger Pack wrote::

I really like it.
Appears that it wants default to be level 0 [?]

Thank you for your notice.

I've change my thought. Thread#backtrace() is more proffered name.

• On Thread#caller(lev), nobody may use lev (!= 0)
• Deciding the semantics of lev except zero may be difficult

How about it?

--
// SASADA Koichi at atdot dot net

=end

[ko1 (Koichi Sasada)]

=begin
Roger Pack wrote::

Roger Pack wrote::

I really like it.
Appears that it wants default to be level 0 [?]
Thank you for your notice.

I've change my thought. Thread#backtrace() is more proffered name.

• On Thread#caller(lev), nobody may use lev (!= 0)
• Deciding the semantics of lev except zero may be difficult

That sounds better. Then the semantics for caller never change. So
this would be Thread#backtrace can have lev > 0?
Either way's good for me.

No. Same as Exception#backtrace.

How about it, matz?

--
// SASADA Koichi at atdot dot net

=end

[matz (Yukihiro Matsumoto)]

=begin
Hi,

In message "Re: [ruby-core:23812] Re: [Bug #977] caller for all threads patch"
on Fri, 12 Jun 2009 08:28:53 +0900, SASADA Koichi [e-mail] writes:

|No. Same as Exception#backtrace.
|How about it, matz?

I see no problem.

						matz.

=end

[rogerdpack (Roger Pack)]

=begin
You can close this one--thank you to Ko1 for implementing it for me.
=end

[shyouhei (Shyouhei Urabe)]

• Status changed from Open to Closed

=begin

=end