Very Strange Error(s)

ruby-talk
1

Hi,

I've been encountering strange behaviour with Ruby recently. I'm wondering if anybody has seen anything like this before or if anyone has any ideas.

Here's what's happening:

- something that should be a String turns out to be an Integer
- if I run the same method call again, say by using a begin...rescue in a loop, and I loop a few times it eventually works with identical input -- if I do run it again, and if an error occurs again, it is usually in a different place
- long running Ruby process (actually Rails running webrick in development mode) -- I *think* I've never seen this in a short running process
- the frequency of the error does not seem to increase (so once it happens it does not start happening all the time, the system seems to be behaving properly)
- error happens during a request by the user.
- error normally quite deep in an execution stack
- ruby 1.8.4 (2006-03-28) [powerpc-darwin8.5.0] -- that's OS X
- I've had this happen in XML parsers (REXML and xampl-pp), in rubyful soup, and in some code that simply prints a tree of objects
- I cannot reproduce it reliably

The going away on second or third try is particularly strange.

Any ideas?

Thanks,
Bob

···

----
Bob Hutchison -- blogs at <http://www.recursive.ca/hutch/>
Recursive Design Inc. -- <http://www.recursive.ca/>
Raconteur -- <http://www.raconteur.info/>
xampl for Ruby -- <http://rubyforge.org/projects/xampl/>

2

This is happening here as well sporadically, and I can't reproduce it either. Mac OS X 10.4.6, ruby 1.8.4 (2005-12-24) [powerpc-darwin8.3.0] compiled from tarball.

-- fxn

···

On Apr 18, 2006, at 22:10, Bob Hutchison wrote:

I've been encountering strange behaviour with Ruby recently. I'm wondering if anybody has seen anything like this before or if anyone has any ideas.

Here's what's happening:

- something that should be a String turns out to be an Integer

3

Are you absolutely sure that to_i isn't being called somewhere along
the way (maybe in some odd cases?)

Ryan

···

On 4/18/06, Bob Hutchison <hutch@recursive.ca> wrote:

Hi,

I've been encountering strange behaviour with Ruby recently. I'm
wondering if anybody has seen anything like this before or if anyone
has any ideas.

Here's what's happening:

- something that should be a String turns out to be an Integer

4

I've been experiencing this same thing on Mac OS X. I haven't been
able to come up with a small piece of code but it seems to happen most
often in my network intensive and threaded code (I have a custom
memcached client that seems to become unstable on heavy loads). I
should also note that this has been for sometime but on extremely rare
and hard to reproduce cases.

I am running on:

ruby 1.8.4 (2006-04-18) [powerpc-darwin8.6.0]

I can't share the code at this point unfortunately (too large and it
is up to my client that I am in contract with -- I will try to come up
with something I can though). Examples of traces I get (in my log
files):

!! undefined method `wakeup' for -517611318:Fixnum
/usr/local/lib/ruby/1.8/thread.rb:116:in `unlock'
/usr/local/lib/ruby/1.8/thread.rb:137:in `synchronize'
.. long trace ..

!! wrong instance allocation
memcache/memcache.rb:209:in `exception'
memcache/memcache.rb:209:in `incr'
.. long trace ..

One thing I can note at this point is that it seems to always be the
-517611318 magic number or a "wrong instance allocation" error. It
seems to be limited to Mac OS X (I've not tested anything else but
different Linux systems).

I would be happy to run any test code people might have to help narrow
the field down. If I get any further information I will post it to
ruby-core.

Brian.

···

On 4/18/06, Bob Hutchison <hutch@recursive.ca> wrote:

Hi,

I've been encountering strange behaviour with Ruby recently. I'm
wondering if anybody has seen anything like this before or if anyone
has any ideas.

Here's what's happening:

- something that should be a String turns out to be an Integer
- if I run the same method call again, say by using a begin...rescue
in a loop, and I loop a few times it eventually works with identical
input -- if I do run it again, and if an error occurs again, it is
usually in a different place
- long running Ruby process (actually Rails running webrick in
development mode) -- I *think* I've never seen this in a short
running process
- the frequency of the error does not seem to increase (so once it
happens it does not start happening all the time, the system seems to
be behaving properly)
- error happens during a request by the user.
- error normally quite deep in an execution stack
- ruby 1.8.4 (2006-03-28) [powerpc-darwin8.5.0] -- that's OS X
- I've had this happen in XML parsers (REXML and xampl-pp), in
rubyful soup, and in some code that simply prints a tree of objects
- I cannot reproduce it reliably

The going away on second or third try is particularly strange.

Any ideas?

5

Yeah, I've seen this plenty myself. I believe it is this:

http://ruby-talk.org/cgi-bin/scat.rb/ruby/ruby-core/7401

And if I understood the discussion (way over my head), it has been fixed:

http://ruby-talk.org/cgi-bin/scat.rb/ruby/ruby-core/7477

James Edward Gray II

···

On Apr 18, 2006, at 3:10 PM, Bob Hutchison wrote:

Hi,

I've been encountering strange behaviour with Ruby recently. I'm wondering if anybody has seen anything like this before or if anyone has any ideas.

6

I found this <http://blog.segment7.net/articles/2006/04/07/chasing-undefined-method-for-fixnum&gt; just now. Looks pretty relevant. Seems as though a re-compile with -O0 might be a work around. Some suggestion that using gcc 3.3 might work too.

···

On Apr 18, 2006, at 4:10 PM, Bob Hutchison wrote:

Hi,

I've been encountering strange behaviour with Ruby recently. I'm wondering if anybody has seen anything like this before or if anyone has any ideas.

Here's what's happening:

- something that should be a String turns out to be an Integer
- if I run the same method call again, say by using a begin...rescue in a loop, and I loop a few times it eventually works with identical input -- if I do run it again, and if an error occurs again, it is usually in a different place
- long running Ruby process (actually Rails running webrick in development mode) -- I *think* I've never seen this in a short running process
- the frequency of the error does not seem to increase (so once it happens it does not start happening all the time, the system seems to be behaving properly)
- error happens during a request by the user.
- error normally quite deep in an execution stack
- ruby 1.8.4 (2006-03-28) [powerpc-darwin8.5.0] -- that's OS X
- I've had this happen in XML parsers (REXML and xampl-pp), in rubyful soup, and in some code that simply prints a tree of objects
- I cannot reproduce it reliably

The going away on second or third try is particularly strange.

Any ideas?

Thanks,
Bob

----
Bob Hutchison -- blogs at <http://www.recursive.ca/hutch/&gt;
Recursive Design Inc. -- <http://www.recursive.ca/&gt;
Raconteur -- <http://www.raconteur.info/&gt;
xampl for Ruby -- <http://rubyforge.org/projects/xampl/&gt;

----
Bob Hutchison -- blogs at <http://www.recursive.ca/hutch/&gt;
Recursive Design Inc. -- <http://www.recursive.ca/&gt;
Raconteur -- <http://www.raconteur.info/&gt;
xampl for Ruby -- <http://rubyforge.org/projects/xampl/&gt;

7

I've not been having any of these problems myself. I've been running WEBrick for days now.

ruby 1.8.4 (2005-12-24) [powerpc-darwin8.6.0]

- Jake McArthur

8

Actually, from what I can tell the issue is separate (or possibly and
unsolved factor). You might note that my post above is using a check
out I made from stable today.

Brian.

···

On 4/18/06, James Edward Gray II <james@grayproductions.net> wrote:

Yeah, I've seen this plenty myself. I believe it is this:

http://ruby-talk.org/cgi-bin/scat.rb/ruby/ruby-core/7401

And if I understood the discussion (way over my head), it has been
fixed:

http://ruby-talk.org/cgi-bin/scat.rb/ruby/ruby-core/7477

9

Excellent! So just to double-check, is it an issue that has been seen only in OSX?

-- fxn

···

On Apr 18, 2006, at 23:42, Bob Hutchison wrote:

I found this <http://blog.segment7.net/articles/2006/04/07/chasing-undefined-method-for-fixnum&gt; just now. Looks pretty relevant. Seems as though a re-compile with -O0 might be a work around. Some suggestion that using gcc 3.3 might work too.

10

Oops, sorry. My bad.

James Edward Gray II

···

On Apr 18, 2006, at 4:09 PM, Brian Mitchell wrote:

On 4/18/06, James Edward Gray II <james@grayproductions.net> wrote:

Yeah, I've seen this plenty myself. I believe it is this:

http://ruby-talk.org/cgi-bin/scat.rb/ruby/ruby-core/7401

And if I understood the discussion (way over my head), it has been
fixed:

http://ruby-talk.org/cgi-bin/scat.rb/ruby/ruby-core/7477

Actually, from what I can tell the issue is separate (or possibly and
unsolved factor). You might note that my post above is using a check
out I made from stable today.

11

I found this <http://blog.segment7.net/articles/2006/04/07/chasing-undefined-method-for-fixnum&gt; just now. Looks pretty relevant. Seems as though a re-compile with -O0 might be a work around. Some suggestion that using gcc 3.3 might work too.

Excellent! So just to double-check, is it an issue that has been seen only in OSX?

It sure looks that way, but I don't know for sure. Someone thought that it might be happening on Windows too, but, to me, it sounded like a different problem. All references/hints that I've come across have been pretty recent compilations of Ruby, even though the Ruby version might be old. So that's a bit of support for the work around.

I hope that work around works. Though it is nice to know I'm not alone :slight_smile:

Cheers,
Bob

···

On Apr 19, 2006, at 3:11 AM, Xavier Noria wrote:

On Apr 18, 2006, at 23:42, Bob Hutchison wrote:

-- fxn

----
Bob Hutchison -- blogs at <http://www.recursive.ca/hutch/&gt;
Recursive Design Inc. -- <http://www.recursive.ca/&gt;
Raconteur -- <http://www.raconteur.info/&gt;
xampl for Ruby -- <http://rubyforge.org/projects/xampl/&gt;

12

An update...

I've re-compiled with the compiler's optimiser set to: -O0 and have not had that error come up again. However, Ruby is *very* slow.

Also I say this today: <http://blog.segment7.net/articles/2006/04/20/update-on-undefined-method-for-fixnum&quot;&gt;

Cheers,
Bob

···

On Apr 19, 2006, at 7:11 AM, Bob Hutchison wrote:

On Apr 19, 2006, at 3:11 AM, Xavier Noria wrote:

On Apr 18, 2006, at 23:42, Bob Hutchison wrote:

I found this <http://blog.segment7.net/articles/2006/04/07/chasing-undefined-method-for-fixnum&gt; just now. Looks pretty relevant. Seems as though a re-compile with -O0 might be a work around. Some suggestion that using gcc 3.3 might work too.

Excellent! So just to double-check, is it an issue that has been seen only in OSX?

It sure looks that way, but I don't know for sure. Someone thought that it might be happening on Windows too, but, to me, it sounded like a different problem. All references/hints that I've come across have been pretty recent compilations of Ruby, even though the Ruby version might be old. So that's a bit of support for the work around.

I hope that work around works. Though it is nice to know I'm not alone :slight_smile:

Cheers,
Bob

-- fxn

----
Bob Hutchison -- blogs at <http://www.recursive.ca/hutch/&gt;
Recursive Design Inc. -- <http://www.recursive.ca/&gt;
Raconteur -- <http://www.raconteur.info/&gt;
xampl for Ruby -- <http://rubyforge.org/projects/xampl/&gt;

----
Bob Hutchison -- blogs at <http://www.recursive.ca/hutch/&gt;
Recursive Design Inc. -- <http://www.recursive.ca/&gt;
Raconteur -- <http://www.raconteur.info/&gt;
xampl for Ruby -- <http://rubyforge.org/projects/xampl/&gt;