From normalperson at yhbt.net  Thu Dec  3 03:22:51 2009
From: normalperson at yhbt.net (Eric Wong)
Date: Thu, 3 Dec 2009 00:22:51 -0800
Subject: [Rev-talk] [PATCH/RFC] timer_watcher_spec: ensure interval has
	passed before	running
In-Reply-To: <20091019023837.GA17116@dcvr.yhbt.net>
References: <20091019023837.GA17116@dcvr.yhbt.net>
Message-ID: <20091203082251.GA17596@dcvr.yhbt.net>

Eric Wong <normalperson at yhbt.net> wrote:
> There was a timer_watcher event_callback defined for the
> IOWatcher class in the timer_watcher.rb file.  I assume this is
> a copy-and-paste error and a grep of the source shows no
> "timer_watcher" methods ever being dispatched.  So change this
> to allow the "on_timer" method to be overidden when given a
> block.
> 
> Also added a spec to show it works.
> ---

Actually, I think the spec I added at the time was always
broken under 1.9...

Or I'm understanding Rev::Loop.run_once incorrectly.

I thought this was broken by my recent changes to make
Rev::Loop.run_once work better under 1.8, but reverting back[1] to the
original commit that introduced this spec reveals it was always broken
under 1.9.

I've pushed this out to git://yhbt.net/rev , but please take
this with a grain of salt since it's probably too late for
me at night to be thinking clearly.

[1] - that means "git clean -f -d -x" in my working copy and any
      gem installations purged.   I even tried it on a clean box
      that's never seen rev.git before.

>From ccb25d3f2a97c234ebf613757f2ee59179dda249 Mon Sep 17 00:00:00 2001
From: Eric Wong <normalperson at yhbt.net>
Date: Wed, 2 Dec 2009 23:58:23 -0800
Subject: [PATCH] timer_watcher_spec: ensure interval has passed before running

Otherwise Rev::Loop.run_once may not have fired
---
 spec/timer_watcher_spec.rb |   12 +++++++++---
 1 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/spec/timer_watcher_spec.rb b/spec/timer_watcher_spec.rb
index dde994c..b7305cd 100644
--- a/spec/timer_watcher_spec.rb
+++ b/spec/timer_watcher_spec.rb
@@ -2,12 +2,15 @@ require File.dirname(__FILE__) + '/../lib/rev'
 
 describe Rev::TimerWatcher do
 
+  interval = 0.010
+
   it "can have the on_timer callback defined after creation" do
-    @watcher = Rev::TimerWatcher.new(0.01, true)
+    @watcher = Rev::TimerWatcher.new(interval, true)
     nr = '0'
     @watcher.on_timer { nr.succ! }.should == nil
     @watcher.attach(Rev::Loop.default).should == @watcher
     nr.should == '0'
+    sleep interval
     Rev::Loop.default.run_once
     nr.should == '1'
   end
@@ -20,25 +23,28 @@ describe Rev::TimerWatcher do
         TMP.succ!
       end
     end
-    @watcher = MyTimerWatcher.new(0.01, true)
+    @watcher = MyTimerWatcher.new(interval, true)
     @watcher.attach(Rev::Loop.default).should == @watcher
     MyTimerWatcher::TMP.should == '0'
+    sleep interval
     Rev::Loop.default.run_once
     MyTimerWatcher::TMP.should == '1'
   end
 
   it "can have the on_timer callback redefined between runs" do
-    @watcher = Rev::TimerWatcher.new(0.01, true)
+    @watcher = Rev::TimerWatcher.new(interval, true)
     nr = '0'
     @watcher.on_timer { nr.succ! }.should == nil
     @watcher.attach(Rev::Loop.default).should == @watcher
     nr.should == '0'
+    sleep interval
     Rev::Loop.default.run_once
     nr.should == '1'
     @watcher.detach
     @watcher.on_timer { nr = :foo }.should == nil
     @watcher.attach(Rev::Loop.default).should == @watcher
     nr.should == '1'
+    sleep interval
     Rev::Loop.default.run_once
     nr.should == :foo
   end
-- 
Eric Wong

From normalperson at yhbt.net  Fri Dec  4 04:21:34 2009
From: normalperson at yhbt.net (Eric Wong)
Date: Fri, 4 Dec 2009 01:21:34 -0800
Subject: [Rev-talk] [PATCH (trivial)] build: use "install" for copying
	shared objects
Message-ID: <20091204092134.GA11390@dcvr.yhbt.net>

Unlike "cp", "install" will unlink the destination before
copying the file over so an already running process can continue
using the previous extension (instead of segfaulting a running
process because a shared object is being overwritten).
---

 I pushed this out to the "buildfix" branch on git://yhbt.net/rev since
 my master branch there has a commit for the timer_watcher spec still
 awaiting a second set of eyes...

 Rakefile |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/Rakefile b/Rakefile
index 3229fb6..60fbb27 100644
--- a/Rakefile
+++ b/Rakefile
@@ -67,7 +67,7 @@ def setup_extension(dir, extension)
 
   file ext_so => ext_files do
     make "#{ext}"
-    cp ext_so, "lib"
+    install ext_so, "lib"
   end
 end
 
-- 
Eric Wong

From youngh at caida.org  Fri Dec  4 20:08:11 2009
From: youngh at caida.org (Young Hyun)
Date: Fri, 4 Dec 2009 17:08:11 -0800
Subject: [Rev-talk] [PATCH] make openssl-nonblock more compatible with Ruby
	1.9.2 preview 2
Message-ID: <254C2C4E-8B16-443E-9D50-28E795666038@caida.org>

Here's a small patch to make openssl-nonblock more compatible with  
Ruby 1.9.2 preview 2.  Specifically, ReadAgain and WriteAgain now  
include IO::WaitReadable and IO::WaitWritable,
respectively.  This change allows a single code base to work with  
either Ruby 1.9.2 preview 2 or with openssl-nonblock on older Ruby  
versions.  I've only tested this on ruby 1.8.7-p174.

Rather than directly rescuing ReadAgain, one should rescue  
IO::WaitReadable (etc.) as described in the Ruby 1.9.2 preview 2 RDoc.

  --Young


-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Changed-to-be-more-compatible-with-Ruby-1.9.2-previe.patch
Type: application/octet-stream
Size: 1859 bytes
Desc: not available
URL: <http://rubyforge.org/pipermail/rev-talk/attachments/20091204/26f3508e/attachment.obj>
-------------- next part --------------


From tony at medioh.com  Fri Dec  4 20:45:22 2009
From: tony at medioh.com (Tony Arcieri)
Date: Fri, 4 Dec 2009 18:45:22 -0700
Subject: [Rev-talk] [PATCH] make openssl-nonblock more compatible with
	Ruby 1.9.2 preview 2
In-Reply-To: <254C2C4E-8B16-443E-9D50-28E795666038@caida.org>
References: <254C2C4E-8B16-443E-9D50-28E795666038@caida.org>
Message-ID: <c7e6b2b00912041745v2b688c4cic3b2a4a255e5aeb1@mail.gmail.com>

Thanks!  Awhile ago I was asking Ruby Core about what the final names for
these would be.  Guess they're decided now.

I actually spun Rev's OpenSSL nonblocking support off into its own gem:

http://github.com/tarcieri/openssl-nonblock

I've been meaning to switch Rev to depend on the openssl-nonblock gem (much
in the way I spun off iobuffer into its own gem), and in fact the names of
these exceptions were one of the things I was waiting on.  This gem
"monkeypatches" (in C code!) the Ruby SSL classes with nonblocking support.
If I adjust the names of these exceptions in this gem, it should present an
API which is fully compatible with Ruby 1.9.2 and is available to any
application running on any other version of Ruby.

I can make the same changes to this gem, or you can submit another patch, if
you like.  Given this development, I'd really like to switch Rev to use the
openssl-nonblock gem.

On Fri, Dec 4, 2009 at 6:08 PM, Young Hyun <youngh at caida.org> wrote:

> Here's a small patch to make openssl-nonblock more compatible with Ruby
> 1.9.2 preview 2.  Specifically, ReadAgain and WriteAgain now include
> IO::WaitReadable and IO::WaitWritable,
> respectively.  This change allows a single code base to work with either
> Ruby 1.9.2 preview 2 or with openssl-nonblock on older Ruby versions.  I've
> only tested this on ruby 1.8.7-p174.
>
> Rather than directly rescuing ReadAgain, one should rescue IO::WaitReadable
> (etc.) as described in the Ruby 1.9.2 preview 2 RDoc.
>

-- 
Tony Arcieri
Medioh/Nagravision
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://rubyforge.org/pipermail/rev-talk/attachments/20091204/38f6ee71/attachment.html>

From tony at medioh.com  Wed Dec 16 00:12:36 2009
From: tony at medioh.com (Tony Arcieri)
Date: Tue, 15 Dec 2009 22:12:36 -0700
Subject: [Rev-talk] [PATCH] make openssl-nonblock more compatible with
	Ruby 1.9.2 preview 2
In-Reply-To: <c7e6b2b00912041745v2b688c4cic3b2a4a255e5aeb1@mail.gmail.com>
References: <254C2C4E-8B16-443E-9D50-28E795666038@caida.org>
	<c7e6b2b00912041745v2b688c4cic3b2a4a255e5aeb1@mail.gmail.com>
Message-ID: <c7e6b2b00912152112v2e083a67q85cfd8b88e970472@mail.gmail.com>

Hi, I pushed your patch to github.

I'll try to release a new openssl-nonblock gem soon, but for now I'm having
issues with Gemcutter.

On Fri, Dec 4, 2009 at 6:45 PM, Tony Arcieri <tony at medioh.com> wrote:

> Thanks!  Awhile ago I was asking Ruby Core about what the final names for
> these would be.  Guess they're decided now.
>
> I actually spun Rev's OpenSSL nonblocking support off into its own gem:
>
> http://github.com/tarcieri/openssl-nonblock
>
> I've been meaning to switch Rev to depend on the openssl-nonblock gem (much
> in the way I spun off iobuffer into its own gem), and in fact the names of
> these exceptions were one of the things I was waiting on.  This gem
> "monkeypatches" (in C code!) the Ruby SSL classes with nonblocking support.
> If I adjust the names of these exceptions in this gem, it should present an
> API which is fully compatible with Ruby 1.9.2 and is available to any
> application running on any other version of Ruby.
>
> I can make the same changes to this gem, or you can submit another patch,
> if you like.  Given this development, I'd really like to switch Rev to use
> the openssl-nonblock gem.
>
>
> On Fri, Dec 4, 2009 at 6:08 PM, Young Hyun <youngh at caida.org> wrote:
>
>> Here's a small patch to make openssl-nonblock more compatible with Ruby
>> 1.9.2 preview 2.  Specifically, ReadAgain and WriteAgain now include
>> IO::WaitReadable and IO::WaitWritable,
>> respectively.  This change allows a single code base to work with either
>> Ruby 1.9.2 preview 2 or with openssl-nonblock on older Ruby versions.  I've
>> only tested this on ruby 1.8.7-p174.
>>
>> Rather than directly rescuing ReadAgain, one should rescue
>> IO::WaitReadable (etc.) as described in the Ruby 1.9.2 preview 2 RDoc.
>>
>
> --
> Tony Arcieri
> Medioh/Nagravision
>



-- 
Tony Arcieri
Medioh! A Kudelski Brand
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://rubyforge.org/pipermail/rev-talk/attachments/20091215/8fef60b4/attachment-0001.html>

From tony at medioh.com  Wed Dec 16 00:21:50 2009
From: tony at medioh.com (Tony Arcieri)
Date: Tue, 15 Dec 2009 22:21:50 -0700
Subject: [Rev-talk] [PATCH/RFC] timer_watcher_spec: ensure interval has
	passed before running
In-Reply-To: <20091203082251.GA17596@dcvr.yhbt.net>
References: <20091019023837.GA17116@dcvr.yhbt.net>
	<20091203082251.GA17596@dcvr.yhbt.net>
Message-ID: <c7e6b2b00912152121x54a041b7y887931d050f80e81@mail.gmail.com>

Sorry about the belated reply... I've been busy with other stuff, especially
Reia :)

On Thu, Dec 3, 2009 at 1:22 AM, Eric Wong <normalperson at yhbt.net> wrote:

> Eric Wong <normalperson at yhbt.net> wrote:
> > There was a timer_watcher event_callback defined for the
> > IOWatcher class in the timer_watcher.rb file.  I assume this is
> > a copy-and-paste error and a grep of the source shows no
> > "timer_watcher" methods ever being dispatched.  So change this
> > to allow the "on_timer" method to be overidden when given a
> > block.
> >
> > Also added a spec to show it works.
> > ---
>
> Actually, I think the spec I added at the time was always
> broken under 1.9...
>
> Or I'm understanding Rev::Loop.run_once incorrectly.
>

Can you better clarify the behavior your were expecting with
Rev::Loop.run_once and timers?

Rev::Loop.run_once will perform a single pass through libev's event loop.
Your spec sleeps until the timer interval should've passed, and checks if it
receives an event, which it does, at least when I run the spec.  So I'm
confused... what's the problem?


> I thought this was broken by my recent changes to make
> Rev::Loop.run_once work better under 1.8, but reverting back[1] to the
> original commit that introduced this spec reveals it was always broken
> under 1.9.
>
> I've pushed this out to git://yhbt.net/rev , but please take
> this with a grain of salt since it's probably too late for
> me at night to be thinking clearly.
>

Is this spec failing for you?

Anyway, merged and pushed...

-- 
Tony Arcieri
Medioh! A Kudelski Brand
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://rubyforge.org/pipermail/rev-talk/attachments/20091215/e51d5593/attachment.html>

From normalperson at yhbt.net  Wed Dec 16 01:23:52 2009
From: normalperson at yhbt.net (Eric Wong)
Date: Wed, 16 Dec 2009 06:23:52 +0000
Subject: [Rev-talk] [PATCH/RFC] timer_watcher_spec: ensure interval
	has	passed before running
In-Reply-To: <c7e6b2b00912152121x54a041b7y887931d050f80e81@mail.gmail.com>
References: <20091019023837.GA17116@dcvr.yhbt.net>
	<20091203082251.GA17596@dcvr.yhbt.net>
	<c7e6b2b00912152121x54a041b7y887931d050f80e81@mail.gmail.com>
Message-ID: <20091216062352.GC1654@dcvr.yhbt.net>

Tony Arcieri <tony at medioh.com> wrote:
> On Thu, Dec 3, 2009 at 1:22 AM, Eric Wong <normalperson at yhbt.net> wrote:
> > Eric Wong <normalperson at yhbt.net> wrote:
> > > There was a timer_watcher event_callback defined for the
> > > IOWatcher class in the timer_watcher.rb file.  I assume this is
> > > a copy-and-paste error and a grep of the source shows no
> > > "timer_watcher" methods ever being dispatched.  So change this
> > > to allow the "on_timer" method to be overidden when given a
> > > block.
> > >
> > > Also added a spec to show it works.
> > > ---
> >
> > Actually, I think the spec I added at the time was always
> > broken under 1.9...
> >
> > Or I'm understanding Rev::Loop.run_once incorrectly.
> 
> Can you better clarify the behavior your were expecting with
> Rev::Loop.run_once and timers?
> 
> Rev::Loop.run_once will perform a single pass through libev's event loop.
> Your spec sleeps until the timer interval should've passed, and checks if it
> receives an event, which it does, at least when I run the spec.  So I'm
> confused... what's the problem?

I was mistaken and thought Rev::Loop.run_once would sleep the calling
thread on its own, but now that I think about it, it makes sense:
run_once may run in a different thread under 1.9.

> > I thought this was broken by my recent changes to make
> > Rev::Loop.run_once work better under 1.8, but reverting back[1] to the
> > original commit that introduced this spec reveals it was always broken
> > under 1.9.

> > I've pushed this out to git://yhbt.net/rev , but please take
> > this with a grain of salt since it's probably too late for
> > me at night to be thinking clearly.
> 
> Is this spec failing for you?
> 
> Anyway, merged and pushed...

Without the patch, yes it was failing under 1.9.  All good now, thanks!

-- 
Eric Wong

From tony at medioh.com  Wed Dec 16 01:29:11 2009
From: tony at medioh.com (Tony Arcieri)
Date: Tue, 15 Dec 2009 23:29:11 -0700
Subject: [Rev-talk] [PATCH/RFC] timer_watcher_spec: ensure interval has
	passed before running
In-Reply-To: <20091216062352.GC1654@dcvr.yhbt.net>
References: <20091019023837.GA17116@dcvr.yhbt.net>
	<20091203082251.GA17596@dcvr.yhbt.net>
	<c7e6b2b00912152121x54a041b7y887931d050f80e81@mail.gmail.com>
	<20091216062352.GC1654@dcvr.yhbt.net>
Message-ID: <c7e6b2b00912152229r37505d87t1f3e9442e1826b79@mail.gmail.com>

On Tue, Dec 15, 2009 at 11:23 PM, Eric Wong <normalperson at yhbt.net> wrote:

> I was mistaken and thought Rev::Loop.run_once would sleep the calling
> thread on its own, but now that I think about it, it makes sense:
> run_once may run in a different thread under 1.9.
>

run_once does not sleep.  It performs a single pass through the event loop
and dispatches any pending events, then immediately returns control back to
the caller once all pending events are processed.  It's the "non-blocking"
alternative to run.


> Without the patch, yes it was failing under 1.9.  All good now, thanks!
>

Great.  Hopefully I can do a new release soon.

-- 
Tony Arcieri
Medioh! A Kudelski Brand
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://rubyforge.org/pipermail/rev-talk/attachments/20091215/23bebc2c/attachment.html>

From tony at medioh.com  Thu Dec 17 00:20:50 2009
From: tony at medioh.com (Tony Arcieri)
Date: Wed, 16 Dec 2009 22:20:50 -0700
Subject: [Rev-talk] [PATCH/RFC] timer_watcher_spec: ensure interval has
	passed before running
In-Reply-To: <e880bd570912162116j172320f6q598517f8f7d4a0c9@mail.gmail.com>
References: <20091019023837.GA17116@dcvr.yhbt.net>
	<20091203082251.GA17596@dcvr.yhbt.net>
	<c7e6b2b00912152121x54a041b7y887931d050f80e81@mail.gmail.com>
	<20091216062352.GC1654@dcvr.yhbt.net>
	<c7e6b2b00912152229r37505d87t1f3e9442e1826b79@mail.gmail.com>
	<e880bd570912162116j172320f6q598517f8f7d4a0c9@mail.gmail.com>
Message-ID: <c7e6b2b00912162120r34dd4d45g11c10483dcf5c331@mail.gmail.com>

On Wed, Dec 16, 2009 at 10:16 PM, Roger Pack <rogerdpack at gmail.com> wrote:

> Hooray for rev--the only evented ruby architecture with an
> on_write_complete LOL.
> -r
>

Honestly, this is the single most important event which is missing from the
EventMachine contract.

Perhaps I should've spent my time simply adding this callback to
EventMachine, but I preferred to build a true OO event API on top of Ruby.

-- 
Tony Arcieri
Medioh! A Kudelski Brand
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://rubyforge.org/pipermail/rev-talk/attachments/20091216/5efd5902/attachment.html>