[#46105] [ruby-trunk - Feature #6687][Open] Enumerable#with — "merborne (kyo endo)" <redmine@...>

14 messages 2012/07/02

[#46133] [ruby-trunk - Feature #6688][Open] Object#replace — "prijutme4ty (Ilya Vorontsov)" <prijutme4ty@...>

24 messages 2012/07/03

[#46160] [ruby-trunk - Feature #6693][Open] Don't warn for unused variables starting with _ — "marcandre (Marc-Andre Lafortune)" <ruby-core@...>

15 messages 2012/07/04

[#46200] [ruby-trunk - Bug #6702][Open] Date should be either required or not — "rosenfeld (Rodrigo Rosenfeld Rosas)" <rr.rosas@...>

14 messages 2012/07/05

[#46296] [ruby-trunk - Feature #6717][Open] Method like #instance_eval that returns self (like #tap) — "alexeymuranov (Alexey Muranov)" <redmine@...>

10 messages 2012/07/10

[#46320] [ruby-trunk - Feature #6721][Open] Object#yield_self — "alexeymuranov (Alexey Muranov)" <redmine@...>

25 messages 2012/07/11

[#46339] [ruby-trunk - Bug #6724][Open] waaaaaaant! ( — "zenspider (Ryan Davis)" <redmine@...>

11 messages 2012/07/11

[#46377] [ruby-trunk - Feature #6727][Open] Add Array#rest (with implementation) — "duckinator (Nick Markwell)" <nick@...>

25 messages 2012/07/13

[#46492] [ruby-trunk - Feature #6737][Open] Add Hash#read and alias as #[]. — "trans (Thomas Sawyer)" <transfire@...>

12 messages 2012/07/15

[#46500] [ruby-trunk - Feature #6739][Open] One-line rescue statement should support specifying an exception class — Quintus (Marvin Gülker) <sutniuq@...>

22 messages 2012/07/15

[#46562] [ruby-trunk - Feature #6758][Open] Object#sequence — "merborne (kyo endo)" <redmine@...>

19 messages 2012/07/20

[#46574] [ruby-trunk - Feature #6762][Open] Control interrupt timing — "ko1 (Koichi Sasada)" <redmine@...>

39 messages 2012/07/20

[#46641] [ruby-trunk - Bug #6780][Open] cannot compile zlib module, when cross-compiling. — "jinleileiking (lei king)" <jinleileiking@...>

14 messages 2012/07/23

[#46659] [ruby-trunk - Bug #6783][Open] Infinite loop in inspect, not overriding inspect, to_s, and no known circular references. Stepping into inspect in debugger locks it up with 100% CPU. — "garysweaver (Gary Weaver)" <garysweaver@...>

8 messages 2012/07/23

[#46792] [ruby-trunk - Bug #6799][Open] Digest::*.hexdigest returns an ASCII-8BIT String — "Eregon (Benoit Daloze)" <redmine@...>

11 messages 2012/07/26

[#46799] [ruby-trunk - Feature #6801][Open] String#~ for a here document — "merborne (kyo endo)" <redmine@...>

12 messages 2012/07/27

[#46829] [ruby-trunk - Feature #6806][Open] Support functional programming: forbid instance/class variables for ModuleName::method_name, allow for ModuleName.method_name — "alexeymuranov (Alexey Muranov)" <redmine@...>

7 messages 2012/07/28

[#46832] [ruby-trunk - Bug #6807][Open] Can't compile ruby without ruby — "devcurmudgeon (Paul Sherwood)" <storitel@...>

13 messages 2012/07/28

[#46834] [ruby-trunk - Feature #6808][Open] Implicit index for enumerations — "trans (Thomas Sawyer)" <transfire@...>

15 messages 2012/07/28

[#46838] [ruby-trunk - Bug #6810][Open] `module A::B; end` is not equivalent to `module A; module B; end; end` with respect to constant lookup (scope) — "alexeymuranov (Alexey Muranov)" <redmine@...>

17 messages 2012/07/28

[#46896] (Half-baked DRAFT) new `require' framework — SASADA Koichi <ko1@...>

Hi,

22 messages 2012/07/31

[ruby-core:46870] Re: [ruby-trunk - Feature #2565] adding hooks for better tracing

From: Aaron Patterson <tenderlove@...>
Date: 2012-07-29 20:55:52 UTC
List: ruby-core #46870
On Sun, Jul 29, 2012 at 12:02:27PM +0900, SASADA Koichi wrote:
> Sorry for late response.
> 
> (2012/07/19 3:52), tenderlovemaking (Aaron Patterson) wrote:
> >   https://bugs.ruby-lang.org/projects/ruby/wiki/DTraceProbes
> 
> my comments:
> 
> > ruby:::function-entry(classname, methodname, filename, lineno);
> > ruby:::function-return(classname, methodname, filename, lineno);
> 
> set_trace_func spearates "call" and "c-call" (and return).
> No need to separate on dtrace?

I don't think we need to separate these.  Information like that hasn't
been useful to me when debugging or profiling my ruby programs.

> > ruby:::require-entry(requiredfile, filename, lineno);
> > ruby:::require-return(filename);
> > ruby:::load-entry(loadedfile, filename, lineno);
> > ruby:::load-return(filename);
> 
> Do we need `require' and `load' both needed? (It depends on usecase)

I'm not sure we need both.  What I *want* is a way to know when Ruby
starts and finishes loading a file.

> What happen on exception?

The `raise` probe is fired.

> (There is a same question on `function-entry')
> 
> > ruby:::object-create-start(classname, filename, lineno);
> > ruby:::object-create-done(classname, filename, lineno);
> 
> (1) How to detect object creation and finish of creation?

rb_obj_alloc has probes.  It wraps object allocation.

> (2) I can't accept your patch on insns.def on string and array.

Why?  You want specific string / array / hash probes (like below)?

> IIRC, you suggest that String creation trace and Array creation trace
> and so on.  I write examples:
> 
>   ruby:::string-create(filename, lineno, size)
>   ruby:::array-create(filename, lineno, size)
>   ruby:::hash-create(filename, lineno, size)
>   ruby:::object-create(filename, lineno, classname)
> 
>   size is string or array size.
>   classname is a classname of object.

These seem good.  I can implement them.

> I'm not sure we need string-modified (size) are needed.

I tried doing this once, but it seemed like the patch was too large.  I
wanted to start small. :)

> ruby:::object-collected(object_id) will be help.  To use it correctly,
> object-create needs to pass object_id.  But it will be complicated.
> 
> > ruby:::gc-begin();
> > ruby:::gc-end();
> 
> How to define GC begin and end?

Nari-san already has defined GC begin and end for GC::Profiler.  See the
GC_PROF_TIMER_START and GC_PROF_TIMER_STOP macros:

  https://github.com/ruby/ruby/blob/trunk/gc.c#L180
  https://github.com/ruby/ruby/blob/trunk/gc.c#L199

My patch just adds the DTrace probes to his existing macros.

> > ruby:::gc-sweep-begin();
> > ruby:::gc-sweep-end();
> 
> How to define GC sweep begin and end?

Same here.  Nari-san has already defined GC_PROF_SWEEP_TIMER_START and
GC_PROF_SWEEP_TIMER_STOP for GC::Profiler.  My patch just adds probes by
his existing macros.

> > ruby:::line(filename, lineno);
> 
> Your patch depends on the `trace' instruction.  I plan to remove `trace'
> instruction on default (if I can implement it).  It will conflicts with
> your proposed patch.
> Or I shouldn't make such optimizations?

I think we should remove this.  I added it for backwards compatibility
with Joyant's DTrace probes, but I never use this.  Besides backwards
compatibility, I don't think it's a useful probe.

-- 
Aaron Patterson
http://tenderlovemaking.com/

In This Thread

Prev Next