SID-00098: FilterPlugin - Can a bad RE cause endless loop?

Status:	Answered	TWiki version:	4.2.3	Perl version:	5.008008 (linux)
Category:	FilterPlugin	Server OS:	Linux 2.6.15.7 (i486-linux-gnu-thread-multi)	Last update:	17 years ago

Some RE work fine others cause endless loops. It doesn't seem possible, but it's forced me to restart apache several times.

• the 1st line works, the 2nd line causes an infinte loop - eats up to 100% of the cpu with the view process

%EXTRACT{topic="%TOPIC%" expand="off" limit="100" pattern="[\n\r]+(\|[^\n\r]*Due Diligence & issue resolution.*?)[\n\r]+" format="$1 $n"}%

%EXTRACT{topic="%TOPIC%" expand="off" limit="100" pattern="[\n\r]+([^\n\r]*Due Diligence & issue resolution.*?)[\n\r]+" format="$1 $n"}%

-- BurtWelsh - 13 Feb 2009

Discussion and Answer

I've had other cases where each view would grab 25% of the capacity on a quad-processor machine: one click to a processor! It happens.

> Can a bad RE cause endless loop?

Not the RE per se, but the perl in the plugin that is parsing it. In other words, you probably have discovered a bug to report.

-- SeanCMorgan - 13 Feb 2009

Well I'm just the fastest bug finder in the world then because the first 5 REs I wrote all caused endless loops. By the way - in the above examples isn't the 1st RE just a subset of the 2nd RE? There were some RE that were even simpler that caused endless loops.

• 1st RE extracts a line starting with "|" and containing "Due Diligence & issue resolution"
• 2nd RE extracts a line containing "Due Diligence & issue resolution"

Anybody know of another way to extract an arbitrary number of occurrences of a string in a file?

-- BurtWelsh - 13 Feb 2009

The bug appears to be that FilterPlugin doesn't like .* (or at least, it works when i don't use that):

1st case:
%EXTRACT{topic="%TOPIC%" expand="off" limit="100"
  pattern="^\| Due Diligence & issue resolution([^\n]*)"
  format="$1 $n"
}%

2nd case:
%EXTRACT{topic="%TOPIC%" expand="off" limit="100" 
   pattern="Due Diligence & issue resolution([^\n]*)"
   format="$1 $n"
}%

FilterPlugin works on each line of %TOPIC%, so you don't need all that [\n\r] stuff in your original EXTRACTs (I removed them for legibility). I used [^\n]* simply to avoid the string .*

P.S. on my server, your EXTRACTs chewed up RAM instead of CPU. That would seem to indicate recursion run amuck.

-- SeanCMorgan - 17 Feb 2009

Thanks, I didn't realize

• FilterPlugin works line by line
• ^ is start of line $ is end of line

-- BurtWelsh - 20 Feb 2009

If you answer a question - or someone answered one of your questions - please remember to edit the page and set the status to answered. The status selector is below the edit box.