Allinav has asked for the wisdom of the Perl Monks concerning the following question:

Hi, I've managed to write a regex which crashes perl and in diagnosing the issue found another which returns the wrong data in the capture groups. I wrote the code below to test out the regex and the returned capture groups. 
#! /usr/bin/perl
use warnings;
use strict;


my $String_to_test = $ARGV[0];
if ( !defined($String_to_test) || $String_to_test eq "")
{
        die "please supply a string to test qr code agaist as first ar
+guement\n";
}

my $qrtest = eval ( $ARGV[1] );
if ( $@ )
{
        die "invalid qr supplied at arg2. caused the following error\n
+$@\n";
}

print "qr=$qrtest\n";
print "string=$String_to_test\n";

if ( $String_to_test =~ $qrtest )
{
        print "$qrtest present\n";
        print "1=$1\n" if defined($1);
        print "2=$2\n" if defined($2);
        print "3=$3\n" if defined($3);
        print "4=$4\n" if defined($4);
        print "5=$5\n" if defined($5);
        print "6=$6\n" if defined($6);
        print "7=$7\n" if defined($7);
        print "8=$8\n" if defined($8);
}
else
{
        print "$qrtest NOT present\n";
}

[download]
To break the problem down I started with the example from perlre for the branch reset. The qr_test.pl contains the code above.
Case 3 is the broken one. Case 1 and 2 are trying to show the fundamental structure of the regex is OK and Case 4 is a work round.

Case 1: Using example regex from perlre branch reset 
./qr_test.pl atuvz 'qr/ ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) )
+ ( z ) /x'
qr=(?x-ism: ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) )
string=atuvz
(?x-ism: ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) ) presen
+t
1=a
2=t
3=v
4=z

[download]
This does what is expected.

Case 2: Added an additional term to 3rd alternative "(w)" 
./qr_test.pl atuvwz 'qr/ ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) 
+(w) ) ( z ) /x'
qr=(?x-ism: ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) (w) ) ( z ) )
string=atuvwz
(?x-ism: ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) (w) ) ( z ) ) pr
+esent
1=a
2=t
3=v
4=z

[download]
This does not do what is expected the 4th capture group should contain w not z and the 5th should contain z

Case 3: Added a fourth term to the 3rd alternative "(x)" 
./qr_test.pl atuvwxz 'qr/ ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v)
+ (w) (x) ) ( z ) /x'
qr=(?x-ism: ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) (w) (x) ) ( z
+ ) )
string=atuvwxz
(?x-ism: ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) (w) (x) ) ( z ) 
+) present
1=a
2=t
./qr_test.sh: line 11: 40158 Segmentation fault

[download]
This crashes perl

Case 4: Having see the odd behaviour in case 2 tried adding dummy groups to first alternative. 
./qr_test.pl 'atuvwxz' 'qr/ ( a )  (?| x ( y ) z | (p (q) r) | (t) u (
+v) (w) (x) ) ( z ) /x'
+ ./qr_test.pl atuvwxz 'qr/ ( a )  (?| x ( y ) z () () ()| (p (q) r) |
+ (t) u (v) (w) (x) ) ( z ) /x'
qr=(?x-ism: ( a )  (?| x ( y ) z () () ()| (p (q) r) | (t) u (v) (w) (
+x) ) ( z ) )
string=atuvwxz
(?x-ism: ( a )  (?| x ( y ) z () () ()| (p (q) r) | (t) u (v) (w) (x) 
+) ( z ) ) present
1=a
2=t
3=v
4=w
5=x
6=z

[download]
This works so I have a work around.

So my questions are these:
Is this a bug?
Or did I missing something in a manual that says the first alternation must have the most capture groups?
if I did missing something in a manual should perl have warned me rather than crashing? Version of perl I'm using is This is perl, v5.10.1 (*) built for x86_64-linux-thread-multi Fix: Upgrade to a perl 5.022 or later.

Replies are listed 'Best First'.
Re: regex with capture groups and branch reset crashes perl
by choroba (Cardinal) on Dec 11, 2015 at 13:04 UTC
    What Perl version are you using? In 5.22.0, I'm getting no crashes, and the "wrong" output seems to be correct. 
    ($q=q:Sq=~/;[c](.)(.)/;chr(-||-|5+lengthSq)`"S|oS2"`map{chr |+ord
}map{substrSq`S_+|`|}3E|-|`7**2-3:)=~y+S|`+$1,++print+eval$q,q,a,

    [download]
[reply]
[d/l]
Re: regex with capture groups and branch reset crashes perl
by mr_mischief (Monsignor) on Dec 11, 2015 at 14:58 UTC

    I tried this on a few versions.

    These passed (OS X, MacPorts):

    • 5.16.3
    • 5.18.4
    • 5.20.3
    • 5.22.0
    These also passed (CentOS):
    • 5.14.4
    • 5.16.3
    These failed:
    • 5.10.1 on CentOS 6 (stopped at 4, no crash though)
    • 5.8.8 on CentOS 5 (got confused by '(?|' syntax and threw an error)

    This really seems to be something caused by an older version as others have said. I would recommend updating to a newer version and testing.

[reply]
      • 5.8.8 on CentOS 5 (got confused by '(?|' syntax and threw an error)

      The '(?|' syntax was only introduced with Perl version 5.10.

      Give a man a fish:  <%-{-{-{-<

[reply]
[d/l]
Re: regex with capture groups and branch reset crashes perl
by Anonymous Monk on Dec 11, 2015 at 14:00 UTC
    Of course, segfault means there is a bug. 5.022 has this in ./re/re_tests: 
    # Used to crash, because the last branch was ignored when the parens
# were counted:
(?|(b)|()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()(
+)()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
+()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()(
+)()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
+()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()(
+)()()()()()()()()()()(a))

    [download]
    Which is kind of similar to your example. Looks like this bug was fixed at some point, I don't know when...
[reply]
[d/l]

Back to Seekers of Perl Wisdom