FreeLists / dokuwiki / [dokuwiki] Re: snapshot errors?

[Chris Smith <chris@xxxxxxxxxxxxx>]

Andreas Gohr wrote:
> On Tue, 06 Feb 2007 17:32:55 +0000
> Chris Smith <chris@xxxxxxxxxxxxx> wrote:
>
>   
> > Todd Augsburger wrote:
> >     
> > > Just installed the 2007-02-06 snapshot and get these errors on many
> > > of the pages:
> > >
> > > Warning: preg_match() [function.preg-match]: Compilation failed:
> > > regular expression too large at offset 0 in
> > > [...] /inc/parser/lexer.php on line 115
> > >
> > > Anybody else getting this, or is it just me? Pages with a large
> > > quantity of links seem go give this error, but other pages seem
> > > slow too.
> > >   
> > >       
> > see fs#1067 at bugs.splitbrain.org
> >     
> Chris, did you have a look at the proposed patch, yet? I understand
> what the patch does, but I'm not quite sure why the parser is called
> this much suddenly. Is this related to the recent metadata changes with
> for FS#1010?
>
> Andi
>
>   
I haven't looked at the proposed patch. 

Re, FS#1010.  Yes and no.  Its related to the change that #1010 fixed. 
Before that fix there would be no attempt to parse metadata if it didn't 
exist.  Ben's patch started parsing metadata for firstheadings when 
metadata didn't exist. My patch continues with that, but removes the 
possibility of looping.
I expect the underlying problem existed before, but the most number of 
calls to the parser was probably something like three or four (sidebar 
with sidebar template, edit page localised text, preview for edit page, 
metadata for edited page) and even that was unlikely to be reached as 
sidebar and localised text are most likely supplied from cache.
I took a look at the code in the parser, as I understand it, the bug 
says second and subsequent instantiations of the parser use existing 
patterns for acronyms, smileys, wordblock, etc, but don't recognise that 
they are already built and go through the same pattern building as 
before ... but adding to the existing pattern.  So we need to either:
- add code to realise the pattern already exists (the patch provide)
- add code further upstream to avoid further pattern building once its 
already complete.
The second alternative should be the best.  I guess we make the final 
pattern static and reuse it if its there.
Cheers,

Chris

--
DokuWiki mailing list - more info at
http://wiki.splitbrain.org/wiki:mailinglist