javafan,

Thanks for your comments

... I don't think @- and @+ are guaranteed to contain anything meaningful during a match, so it would be hard to say there's a bug.

Observe that the main anomaly, i.e., the fact that @- and @+ have different lengths occurs at the end of the regexp parsing, not during the intermediate process. See the following variant of the former example:

pl@nereida:~/Lperltesting$ cat -n ./offsetsin5_10_end.pl
     1  #!/usr/local/lib/perl/5.10.1/bin//perl5.10.1
     2  use v5.10;
     3
     4  my $input;
     5
     6  local $" = ", ";
     7
     8  my $parser = qr{
     9      ^
    10      ((?&expr))
    11      ((?&expr))
    12      \z
    13          (?{
    14               say "main:\n\@- = (@-)\t\t ".scalar(@-)." items\n
+\@+ = (@+)\t ".scalar(@+)." items\n";
    15          })
    16
    17      (?(DEFINE)
    18          (?<expr>
    19              (.) (.)
    20              (?{
    21                   say "expr:\n\@- = (@-)\t ".scalar(@-)." items
+\n\@+ = (@+)\t ".scalar(@+)." items\n";
    22              })
    23          )
    24      )
    25  }x;
    26
    27  $input = <>;
    28  chomp($input);
    29  if ($input =~ $parser) {
    30    say "matches: ($&)";
    31    say "At the very end:\n\@- = (@-)\t ".scalar(@-)." items\n\@
++ = (@+)\t ".scalar(@+)." items\n";
    32  }
[download]

The only new line is line 31. When executed says:

pl@nereida:~/Lperltesting$ ./offsetsin5_10_end.pl
abab
expr:
@- = (0, , , , 0, 1)     6 items
@+ = (2, , , , 1, 2)     6 items

expr:
@- = (0, 0, , , 2, 3)    6 items
@+ = (4, 2, , , 3, 4)    6 items

main:
@- = (0, 0, 2)           3 items
@+ = (4, 2, 4, , , )     6 items

matches: (abab)
At the very end:
@- = (0, 0, 2)   3 items
@+ = (4, 2, 4, , , )     6 items
[download]

Apologies for not being able to help sending a patch. I would like to have the required knowledge to do it.