Varnish 3 not parsing regex pattern correctly

Question

I'm working on some code that generates PCRE patterns used to ban objects in varnish and having a problem with varnish telling me the pattern is invalid.

This is one of the patterns generated by my code

(?i)(((^| )page\-2[^ ]*($| )))

And this is what varnish is telling me

 0 Debug        - "REGEX: <missing )>"

13 RxRequest    c BAN
13 RxURL        c /
13 RxProtocol   c HTTP/1.1
13 RxHeader     c User-Agent: curl/7.34.0
13 RxHeader     c Accept: */*
13 RxHeader     c Host: www.test.local
13 RxHeader     c X-Tags: (?i)(((^| )page\-2[^ ]*($| )))

I've even tried simplifying the pattern during testing to this

page-2[^ ]
page\-2[^ ]

And I get this

 0 Debug        - "REGEX: <missing terminating ] for character class>"

A simple pattern like this works

page-2.*

I'm running varnish-3.0.2 revision cbf1284

The relevant VCL is

if (req.request == "BAN") {
    ban("obj.http.X-Tags ~ " + req.http.X-Tags);
    error 200 "Banned";
}

And, just so it's clear that the original pattern is valid

It may not like the space literal. Try replacing the spaces with `\s`? — RevanProdigalKnight, Jul 25 '14 at 14:54
You should put this as an answer as it is the problem/solution. Not sure why it doesn't like the spaces though. — Lee Saferite, Jul 25 '14 at 14:56
I've posted it as an answer. The only reason I even thought of that is because flex doesn't like space literals either. — RevanProdigalKnight, Jul 25 '14 at 15:02

score 2 · Accepted Answer · answered Jul 25 '14 at 15:01

2

So apparently, as I accidentally discovered in my comment on the original question, varnish doesn't like space literals in regex. The solution in this case is to replace the space literal with \s, which matches on any whitespace character, including spaces, tabs, and newlines.

answered Jul 25 '14 at 15:01

RevanProdigalKnight

1,316
1
14
23

While I don't like having to use \s as opposed to the space literal, this works for my uses right now. – Lee Saferite Jul 25 '14 at 15:24
Playing around I realized I could use `\x20` as well – Lee Saferite Jul 25 '14 at 15:28

Varnish 3 not parsing regex pattern correctly

1 Answers1