Instiki users - May 2006 - [Instiki] WEBrick instable at multiple text revisions

Here is an intersection of Mark Scala''s problem "frequent
crashes"
(2006-05-05) and Damian Cugley''s problem "Resource leak..."  
(2006-05-09).

Instiki takes frequent revisions of the same page by the same author  
within a short time as multiple iterations of the same version, not as  
separate versions. Nice feature. But after a few iterations, my  
Instiki can''t handle it anymore, and I''m getting error
messages like
"private method `split'' called for 113435111:Fixnum",
"private method
`gsub!'' called for 101855024:Fixnum", or "undefined method
`+'' for
nil:NilClass", after submitting the updated text. Sometimes
''Cancel''
is an option, sometimes not, since it leads to a page only saying "You  
are being redirected" from which escape is impossible. So, often a  
second ''Submit'' is inevitable, which succeeds, be it after
having made
some trivial change in the text to avoid the "You have tried to save  
... whithout changing it''s content" message.

A minor annoyance? No. From that moment on the Ruby process is eating  
memory whenever a ''Submit'' is done on whatever edit page,
until all
memory is consumed and the game is over. Another not very unusual  
scenario: Ruby doesn''t swallow up the memory, but crashes immediately.
It may take a lot of time to get the web server down, once it has  
started seizing more and more memory.

I''m using Instiki 0.11.0 and Ruby 1.8.4 [2005-12-24] on Windows 2000  
or XP. WEBrick takes care for the webserving. Having no more than one  
(myself) or a maximum of 60 users (my department), WEBrick should be  
able to cope with it, I think. The number of text revisions that may  
give trouble, varies between 3 and 16. All the testing concerned one  
or two users being logged in.

Anyone any idea?

msm at uwnet.nl wrote:
> Instiki takes frequent revisions of the same page by the same author
> within a short time as multiple iterations of the same version, not as
> separate versions. Nice feature. But after a few iterations, my
> Instiki can''t handle it anymore, and I''m getting error
messages like
> "private method `split'' called for 113435111:Fixnum",
"private method
> `gsub!'' called for 101855024:Fixnum", or "undefined
method `+'' for
> nil:NilClass", after submitting the updated text.
Ah, *that* I''ve seen before: it was due to a hand-built Ruby on Mac  
OS X with incorrect threading compile options.  This caused Ruby''s  
internal state to get completely screwed up for some (but not all)  
applications.  A correctly built Ruby interpreter resolved those  
nnnn:Fixmnum problems and related bizarre issues entirely.

If you''re using a pure-Windows (i.e. not Cygwin) Ruby, you might want  
to check out the RubyInstaller (http://rubyinstaller.rubyforge.org/ 
wiki/wiki.pl) page for a one-click Ruby installer package for  
Windows.  See if the most recent package (1.8.4-17 release candidate  
2, built on May 3, 2006) works any better for you.
> I''m using Instiki 0.11.0 and Ruby 1.8.4 [2005-12-24] on Windows
2000
> or XP.

-- John

John, and others,

After 32 iterations: I give up, this Ruby version is too strong for  
me, how healthy its
internal state!

The new package also reports "ruby 1.8.4 (2005-12-24) [i386-mswin32]",
so I have to be
more precise: "1.8.4-16 preview 3" was behaving badly, the just  
installed "1.8.4-17
release candidate 2" works fine, until now. A first impression of  
course, but I''m rather
confident at this moment.

Thanks,
Meindert Meindertsma.

Quoting John Whitley <whitley at acm.org>:
> msm at uwnet.nl wrote:
>> Instiki takes frequent revisions of the same page by the same author
>> within a short time as multiple iterations of the same version, not as
>> separate versions. Nice feature. But after a few iterations, my
>> Instiki can''t handle it anymore, and I''m getting
error messages like
>> "private method `split'' called for
113435111:Fixnum", "private method
>> `gsub!'' called for 101855024:Fixnum", or "undefined
method `+'' for
>> nil:NilClass", after submitting the updated text.
>
> Ah, *that* I''ve seen before: it was due to a hand-built Ruby on
Mac
> OS X with incorrect threading compile options.  This caused Ruby''s
> internal state to get completely screwed up for some (but not all)
> applications.  A correctly built Ruby interpreter resolved those
> nnnn:Fixmnum problems and related bizarre issues entirely.
>
> If you''re using a pure-Windows (i.e. not Cygwin) Ruby, you might
want
> to check out the RubyInstaller (http://rubyinstaller.rubyforge.org/
> wiki/wiki.pl) page for a one-click Ruby installer package for
> Windows.  See if the most recent package (1.8.4-17 release candidate
> 2, built on May 3, 2006) works any better for you.
>
>> I''m using Instiki 0.11.0 and Ruby 1.8.4 [2005-12-24] on
Windows 2000
>> or XP.
>
>
> -- John
>
>
> _______________________________________________
> Instiki-users mailing list
> Instiki-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/instiki-users
>

Something is rotten in the state of my wiki. The Ruby upgrade wasn''t  
the panacea after all, although most pages are trouble-free now  
indeed. Some pages however still may go beserk at an unpredictable  
moment: weird error messages at submitting, and the subsequent memory  
leak. A quantum effect? Worn-out bits?

After many experiments with a troublesome page, I''m getting to suspect
Textile-tables containing WikiWords. When WikiWords (actually their  
square brackets) are removed, or when Textile table-formatting is  
replaced by pure HTML (preserving all WikiWords), there never seems to  
come an end at the error-free updating of that page. Here is the  
problematic one: [myPage]

  table(in).
  |(nn). Volledige naam |(nn). : |(n). Bxxxxxxxxxxxxx Axxxxxxxxx  
(xx-xxxxxxxxxx) |
  |(nn). Klantgroep |(nn). : |(n). *X* |
  |(nn). AS/IS-code |(nn). : |(n). *XX* |
  |(nn). Productie-omgeving |(nn). : |(n). *X* op mainframe XXXX |
  |(nn). Beheercluster |(nn). : |(n). [[Cluster 1]], [[Cluster 2]]  
(Fxxx Xxxxx xx Xxx; Fxxx Xxxx Xxxxxxxxxx), [[Cluster 3]], [[Cluster  
6]] |
  |(nn). Applicatiebeheer |(nn). : |(n). [[Meindert Meindertsma]] |
  |(nn). Techniek |(nn). : |(n). [[SAS]], [[JCL]], [[OPC]] |

The quick brown fox jumps over the lazy dog. The quick brown fox jumps  
over the lazy dog. The quick brown fox jumps over the lazy dog. The  
quick brown fox jumps over the lazy dog. The quick brown fox jumps  
over the lazy dog. The quick brown fox jumps over the lazy dog.

The quick brown fox jumps over the lazy dog. The quick brown fox jumps  
over the lazy dog. The quick brown fox jumps over the lazy dog. The  
quick brown fox jumps over the lazy dog.

Iteratie 1: anonimiseren.
Iteratie 2.
Iteratie 3.
Iteratie 4.
Iteratie 5.
Iteratie 6.
Iteratie 7.
Iteratie 8.
Iteratie 9.
Iteratie 10.
Iteratie 11 geeft:
undefined method `to_sym'' for nil:NilClass.

[/myPage] Sorry for the Dutch frases, but you get the idea. The first  
iteration consisted of creating a new page, pasting text from another  
page and covering parts of it with x-s and a little story about a  
jumping fox. All next updates I did no more than adding a line  
"Iteratie 2." and so on.

I don''t have to present the HTML version, do I? I assure you it looks  
awful, because the whole table must be written on one logical line,  
without any hard returns between or inside the tags (Markdown would do  
a better job here, I presume). O well, there is trick: a linebreak at  
the end of (and within!) each table cell is harmless, but still not  
making a great layout in the edit page. By the way: is it safe to put  
HTML tags into this mail -- what if it''s read through a web browser?  
That''s why I''m a bit cautious when dealing with HMTL.

Could my assumption that Textile-tables and WikiWords don''t go  
together be true? Has anyone similar, supplemental or contrary  
experiences? This could be a reason to switch to Markdown, but I want  
to be sure before taking such a drastic step.

Meindert Meindertsma.

Quoting Meindert Meindertsma <msm at uwnet.nl>:
> John, and others,
>
> After 32 iterations: I give up, this Ruby version is too strong for
> me, how healthy its
> internal state!
>
> The new package also reports "ruby 1.8.4 (2005-12-24)
[i386-mswin32]",
> so I have to be
> more precise: "1.8.4-16 preview 3" was behaving badly, the just
> installed "1.8.4-17
> release candidate 2" works fine, until now. A first impression of
> course, but I''m rather
> confident at this moment.
>
> Thanks,
> Meindert Meindertsma.
>
> Quoting John Whitley <whitley at acm.org>:
>
>> msm at uwnet.nl wrote:
>>> Instiki takes frequent revisions of the same page by the same
author
>>> within a short time as multiple iterations of the same version, not
as
>>> separate versions. Nice feature. But after a few iterations, my
>>> Instiki can''t handle it anymore, and I''m getting
error messages like
>>> "private method `split'' called for
113435111:Fixnum", "private method
>>> `gsub!'' called for 101855024:Fixnum", or
"undefined method `+'' for
>>> nil:NilClass", after submitting the updated text.
>>
>> Ah, *that* I''ve seen before: it was due to a hand-built Ruby
on Mac
>> OS X with incorrect threading compile options.  This caused
Ruby''s
>> internal state to get completely screwed up for some (but not all)
>> applications.  A correctly built Ruby interpreter resolved those
>> nnnn:Fixmnum problems and related bizarre issues entirely.
>>
>> If you''re using a pure-Windows (i.e. not Cygwin) Ruby, you
might want
>> to check out the RubyInstaller (http://rubyinstaller.rubyforge.org/
>> wiki/wiki.pl) page for a one-click Ruby installer package for
>> Windows.  See if the most recent package (1.8.4-17 release candidate
>> 2, built on May 3, 2006) works any better for you.
>>
>>> I''m using Instiki 0.11.0 and Ruby 1.8.4 [2005-12-24] on
Windows 2000
>>> or XP.
>>
>>
>> -- John
>>
>>
>> _______________________________________________
>> Instiki-users mailing list
>> Instiki-users at rubyforge.org
>> http://rubyforge.org/mailman/listinfo/instiki-users
>>
>
>
>
> _______________________________________________
> Instiki-users mailing list
> Instiki-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/instiki-users
>